Anthropic and OpenAI Head Toward $3 Trillion Trillion IPOs as Trump Accounts App Launches
All-In Podcast
The token revenue market has consolidated into an Anthropic and OpenAI duopoly, while the new 'Trump Accounts' app marks the most monumental transformation of the American social contract since 1935 by converting young Americans into capital shareholders.
## Background
Published on YouTube by the All-In Podcast on July 11, 2026, this episode features guest host Brad Gerstner filling in for David Friedberg. The show description outlines discussions on OpenAI vs. Anthropic IPOs, Meta's model pricing war, potential Chinese export controls on AI, and the launch of the 'Trump Accounts' initiative to get young Americans bought into capitalism.
## TL;DR
The episode explores the consolidation of the frontier AI market into an Anthropic and OpenAI duopoly, detailing the staggering financial metrics, infrastructure challenges, and the shifting dynamics between open-source and closed-source enterprise models. The hosts also examine China's intensifying regulatory control over its domestic AI sector, including restrictions on overseas model access and corporate shifts to closed architectures. Finally, the discussion highlights the official launch of 'Trump Accounts,' a monumental financial initiative designed to provide every American child with an investment seed to promote widespread equity ownership and long-term wealth generation.
## Atmosphere & style
The atmosphere is energetic, fast-paced, and highly collaborative, driven by the 'hot software summer' dynamic of major tech and policy shifts. The hosts display a sharp, elite-tier peer dynamic, blending high-stakes venture capitalism insights with pragmatic technical critiques. There is a palpable sense of historic scale and optimism, particularly during the deep dive into the nationwide rollout of the Invest America Act, which contrasts against the sober, hard-nosed mathematical realities discussed regarding enterprise AI token costs.
## Key takeaways
- [03:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=189s) SpaceX successfully executed a landmark public offering at a $2 trillion valuation, raising $5 billion on approximately $35 billion of forward revenue.
- [03:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=189s) Anthropic and OpenAI are driving a massive trillion-dollar IPO pipeline, with Anthropic filing confidentially and OpenAI rumored to release GPT-6 amid soaring revenue trends.
- [13:02](https://www.youtube.com/watch?v=PHL1j2ti420&t=782s) Enterprise AI ROI currently faces a stark reality, with token costs escalating rapidly while realized productivity gains for the S&P 493 remain heavily capped.
- [17:41](https://www.youtube.com/watch?v=PHL1j2ti420&t=1061s) Uber provides an operational blueprint for AI efficiency, deploying over 2,500 distinct agentic skills and integrating AI tools across 99% of its engineering workforce.
- [20:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1200s) The total addressable market for artificial intelligence is fundamentally unprecedented, as the technology scales into a bottom-up adoption model across every corporate tier globally.
- [25:05](https://www.youtube.com/watch?v=PHL1j2ti420&t=1505s) Rapidly falling inference and token costs, driven by open-source frameworks and hardware innovations, are fueling Jevons Paradox by radically scaling aggregate token consumption.
- [27:39](https://www.youtube.com/watch?v=PHL1j2ti420&t=1659s) Corporate share of wallet remains heavily concentrated within elite frontier labs like Anthropic and OpenAI, creating a dominant revenue duopoly despite open-source alternatives.
- [35:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=2100s) Mark Zuckerberg and Meta have ignited a fierce price war by entering the token distribution market directly with heavily discounted pricing via the Meta Model API.
- [54:29](https://www.youtube.com/watch?v=PHL1j2ti420&t=3269s) The Chinese Communist Party is tightening national security regulations, pushing top domestic AI companies from open-source to closed-source architectures to restrict overseas access.
- [01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) The official 'Trump Accounts' app launched under the Invest America Act, generating over 1.5 million accounts and $1 billion in total deposits within its first 24 hours.
- [01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) The Invest America Act establishes a privately owned investment account at birth for every American child, creating a monumental shift in the national social contract.
- [01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) Prominent tech leaders and philanthropists, including Michael Dell, Gwynne Shotwell, and Brad Gerstner, have contributed billions to jumpstart accounts for lower-income children.
- [01:13:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=4389s) The initiative offers major tax advantages and compound growth mechanics designed to systematically transform young Americans into lifelong capital shareholders.
- [01:21:40](https://www.youtube.com/watch?v=PHL1j2ti420&t=4900s) The programmatic rollout aims to directly combat the unstable K-shaped economic recovery and re-engage younger generations with the core mechanisms of capitalism.
- [01:29:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5340s) Strict withdrawal limits and milestone gates prevent immediate liquidation at age 18, forcing the vast majority of wealth to roll over into long-term retirement accounts.
- [01:36:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5760s) Unlike traditional Social Security, Trump Accounts guarantee legal title and direct asset transferability to a participant's legal heirs.
The podcast opens with Jason Calacanis introducing All-In episode 280, noting that David Friedberg is away on vacation. Brad Gerstner fills in for Friedberg, joining Chamath Palihapitiya and David Sacks. The hosts exchange banter about their respective summer travels; Gerstner calls in from his "patriotic" flag room on the East Coast after spending time in Washington, D.C. celebrating America's 250th upcoming milestone, while Calacanis reports from Paris, France, where he has been conducting interviews at the Raise Conference.
Chamath describes his current routine as a "hot software summer," operating out of Europe to meet with enterprise software clients. He shares details of a dinner in Geneva, Switzerland, which led to his appointment alongside tech figures like Jensen Huang, Brad Smith, and Grab CEO Anthony Tan to a new United Nations Commission for AI. The commission is co-chaired by Salesforce CEO Marc Benioff, whom Chamath praises as the "impresario of impresarios" for his business-building mastery. When asked by the hosts what the true calling of this UN commission is, Chamath quips "open source," though he admits the presence of Anthropic co-founder Tom Brown, who was advocating passionately for safety and their specific vision.
---
[03:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=189s) The Trillion-Dollar IPO Rush: SpaceX, Anthropic, and OpenAI
The conversation transitions to the main tech narrative: a massive rush toward trillion-dollar IPOs, a phenomenon previously dissected at the All-In Liquidity Summit. Calacanis highlights that SpaceX recently led the charge, executing a textbook public offering that priced at a $2 trillion valuation. Although the stock surged to $200 per share post-IPO, it has since settled back to around $150 per share, which is right around its initial public offering price. The besties view this as a highly successful, "priced to perfection" blueprint that other mega-cap AI companies are studying.
Attention shifts to Anthropic, which filed confidentially for an IPO on June 1st. Prediction markets on Polymarket reflect a 65% chance that Anthropic will debut publicly before the end of 2026. Tech investor Gavin Baker previously speculated that Anthropic could exit 2026 with an annualized revenue run rate eclipsing $100 billion, suggesting it could trade at a $3 trillion valuation. Meanwhile, rumors on Twitter indicate OpenAI has regained its momentum, preparing to release its GPT-6 models within the next 30 days, with revenue trending toward $70 billion for the year.
* **SpaceX IPO Metrics:** Raised $5 billion at a $2 trillion valuation, trading on roughly $35 billion of forward revenue. Peak-to-trough drawdown post-IPO hit roughly 50% before stabilizing.
* **Anthropic Revenue Rumors:** Trending to exit the year at over $100 billion in revenue, potentially out-pacing OpenAI's recent trajectory.
* **OpenAI Revenue Rumors:** Trending around $70 billion for the year, doubling the scale of SpaceX's forward revenue baseline.
* **Polymarket Odds:** 65% probability of an Anthropic IPO occurring within the calendar year.
---
[06:45](https://www.youtube.com/watch?v=PHL1j2ti420&t=405s) Market Mechanics and the Retail Investor Dilemma
Gerstner breaks down the structural innovations pioneered by Elon Musk, Brett Johnsen, and Gwynne Shotwell during the SpaceX offering. He details how they navigated early index inclusion and structured a "staged release" time-and-milestone-based lockup to mitigate the extreme volatility typical of mega-IPOs. While institutions like Altimeter, Fidelity, and T. Rowe Price are eager to buy into Anthropic and OpenAI at scale, Gerstner cautions that these are no longer "get-rich-quick schemes" for retail investors.
The hosts engage in a philosophical debate regarding modern accredited investor laws, which Calacanis and Gerstner agree are "insane" because they shut everyday citizens out of early wealth-generation windows. However, because these AI giants are going public at trillion-dollar baselines, future returns will be dictated by steady compounding rather than immediate 100% pops. Gerstner projects that both Anthropic and OpenAI can realistically compound their revenues at over 30% annually for the foreseeable future, an extraordinary rate for companies operating at such massive financial scale.
---
[13:02](https://www.youtube.com/watch?v=PHL1j2ti420&t=782s) The Enterprise AI ROI Reckoning
A critical debate emerges over the commercial sustainability of AI token spend. Chamath presents a stark warning based on a real-time assessment from his own CTO: their company's token costs are doubling every 45 days, yet downstream productivity gains have effectively hit an asymptote, yielding a maximum 5% improvement. He argues that while the market is currently "astonished" by raw top-line revenue growth, an enterprise reckoning is inevitable.
To test the broader economic impact of AI, Chamath tasked Anthropic's new model, Fable 5 High, with calculating how much AI has boosted S&P 500 earnings per share (EPS) growth since 2024. While the model initially claimed a 50% lift, Chamath adjusted the query to exclude Nvidia's chip manufacturing windfall. The resulting data revealed that the corporate EPS growth for the remaining S&P 493 was just 9%—driven overwhelmingly by inflationary pricing power and a 3% lift from stock buybacks. This places the actual, realized enterprise ROI from AI tools somewhere between a meager 0% and 2%. Chamath warns that consumer-facing models provide a "safe harbor" from these aggressive ROI calculations because individual subscribers don't demand hard financial metrics, whereas brittle enterprise budgets eventually will.
> "Right now, our token costs are doubling every 45 days... our upside is essentially flat... we've effectively already asymptoted. I suspect that if you can get out now, you should get out now before all of that starts to seep into the water table."
---
[17:41](https://www.youtube.com/watch?v=PHL1j2ti420&t=1061s) Uber's Blueprint for Algorithmic Efficiency
The conversation shifts to how elite operations-driven enterprises are fighting back against runaway token costs. Calacanis cites a public corporate breakdown shared on X by Uber's CTO, Praveen Anidil. After burning through their initial token allocations in Q1, Uber restructured its approach away from unstructured "vibe coding" toward specialized, department-level engineering alignment.
Uber's infrastructure model offers a potential path forward for corporate capital allocation. Rather than letting isolated employees experiment aimlessly with tools like Perplexity, Uber is embedding forward-deployed software engineers directly into core operational departments—such as legal, marketing, HR, and procurement. These engineers partner with department heads who possess deep "systems thinking" expertise to architect targeted agentic skills. Calacanis suggests that public companies should begin explicitly reporting the EPS gains directly attributable to AI implementations to restore market confidence in corporate tech spending.
* **Uber Engineering Adoption:** 99% of Uber's engineering workforce currently utilizes integrated AI developer tools.
* **Code Architecture:** Over 70% of all pull requests at Uber are now explicitly attributed to local or cloud-based AI agents.
* **Agent Deployment:** Uber's technical staff has successfully engineered over 2,500 distinct "agentic skills" deployed across the enterprise.
[20:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1200s) The Unprecedented TAM of Intelligence
The current scale of money being poured into artificial intelligence is undeniably experimental, and direct return on investment (ROI) is still a question mark for many. However, the market is so early in terms of enterprise adoption that the typical rules of revenue concentration do not apply. The total addressable market (TAM) here encompasses every small, medium, and large company on the planet, which explains why we have never witnessed revenue growth quite like this before. This growth isn't concentrated within a handful of massive clients; instead, millions of customers are independently making rational economic decisions every day to invest in tokens. While tech-adjacent figures like Praveen at Uber are actively working on cost takeouts to justify these investments, the industry is on the verge of breakthroughs in machine intelligence that will dramatically transform the revenue side of the equation. This includes major advancements in life sciences and product innovation.
Companies are finding it impossible to divorce themselves from this trajectory. For example, Jensen Huang has noted that all next-generation chip design work at NVIDIA is now being assisted by AI—meaning the machine is literally building the machine. For a leader at the frontier like Jensen, using anything less than the absolute best frontier models is out of the question because even tiny advantages in intelligence are vital. This phenomenon mirrors the trajectory of Snowflake; despite significant infrastructure and query optimization over the years, their revenue growth continued unabated because they consistently penetrated deeper use cases across the enterprise. If the top frontier labs cross the $100 billion revenue mark by the end of the year, they could easily see a three-to-five-times growth trajectory next year. A jump from $100 billion to $300 billion or $400 billion represents an incremental $200 billion to $300 billion in revenue, a feat that is entirely incomprehensible in the history of Silicon Valley and the world.
- **Enterprise Footprint:** Millions of customers are independently choosing to buy into AI tokens based on daily economic utility.
- **NVIDIA Case Study:** Jensen Huang has integrated AI into all internal chip architecture workflows, demonstrating that frontier labs rely on the highest-performing models available.
- **Unparalleled Growth Scale:** A potential leap from $100 billion to over $300 billion in a single year would represent unprecedented revenue expansion in economic history.
---
[22:36](https://www.youtube.com/watch?v=PHL1j2ti420&t=1356s) Bottom-Up Adoption and Token Maxing
What makes intelligence-on-demand uniquely disruptive compared to historical software shifts is its universal applicability across organizations. When a classic tool like Microsoft Excel was first introduced, its immediate benefits accrued almost exclusively to specific departments like accounting, leaving human resources or marketing largely unaffected. In contrast, generative AI tools are actively utilized by every single person across every tier of an enterprise simultaneously. If a 1,000-person organization has employees spending $200 a month on these tools, which then doubles to $400 a month, the total expenditure amounts to roughly $5,000 a year per person. When compared to an average corporate salary of $100,000 to $150,000, that is merely an incremental 3% to 5% cost. Given that these tools can make an individual multiple times more effective at their job, the corporate spend is an easy mathematical justification.
This represents a classic bottom-up product adoption cycle. Employees can simply put a $20-a-month subscription on their corporate credit cards without triggering the scrutiny of a Chief Information Officer (CIO) or Chief Technology Officer (CTO). This frictionless, widespread corporate penetration explains the steep revenue ramp that observers are struggling to internalize. It raises the question of whether any department in a modern organization can legitimately claim that intelligence-on-demand isn't for them. However, barriers still exist; compliance and regulatory frameworks mean that sectors dealing with sensitive financial data, HIPAA regulations, or private HR data are not yet fully permitted to deploy these tools at scale.
- **Universal Utility:** Unlike legacy software like Excel, generative AI tools instantly impact every department from HR to engineering.
- **Low Friction Spend:** The $20-a-month entry point bypasses typical CIO/CTO procurement hurdles via corporate credit card expenses.
- **The Productivity Equation:** An incremental corporate spend of 3% to 5% per employee is easily offset by multi-fold improvements in workplace efficiency.
---
[25:05](https://www.youtube.com/watch?v=PHL1j2ti420&t=1505s) The 95% Price Drop and Agentic Arbitrage
Once a user experiences the productivity gains of AI, the workflow becomes highly addictive. This shifts behavior toward exploring open-source alternatives and sophisticated developer frameworks, such as the open-source Hermes agent by Nous Research, and leveraging platforms like OpenRouter to acquire API keys for models like GLM. On the infrastructure side, decentralized crypto projects like Bittensor (TAO) are spinning up specialized subnets that make models like GLM 5.2 available at incredibly low prices. Accessing these alternative API pipelines can slash token costs by up to 95%. This massive deflationary shift in inference costs is happening concurrently across multiple vectors: rapid software optimizations, advancing open-source models, distributed computing networks, and hardware leaps from companies like Groq and Cerebras.
"When you have unlimited tokens as an exercise... it really does change your thinking when costs go down." Dropping inference costs by 95% completely alters user behavior. Instead of running automated AI agents on a daily cadence, a user can comfortably shift them to hourly schedules. Furthermore, a single monolithic task can be broken up into three distinct specialized agents running hourly side-by-side. For example, a system can be set up to ingest every single episode of the *All-In Podcast* and *This Week in Startups* via scheduled cron jobs, operating a continuous trend-spotting agent that surfaces the top technological shifts every hour based on raw keyword inputs. This massive spike in volume driven by falling prices is a textbook example of Jevons Paradox: as a resource becomes more efficient and cheaper to use, the aggregate consumption of that resource increases dramatically.
- **Inference Price Meltdown:** Token costs are experiencing roughly 90% price reductions year-over-year, heavily driven by hardware innovators like Groq and Cerebras.
- **Bittensor (TAO):** The crypto network's subnets are hosting models like GLM 5.2, enabling extreme cost-cutting for heavy API developers.
- **Agentic Workloads:** Drastically lower costs allow users to scale single agent workflows into multi-agent systems running hourly cron jobs.
---
[27:39](https://www.youtube.com/watch?v=PHL1j2ti420&t=1659s) Open Source vs. Frontier Share of Wallet
The central debate in AI circles revolves around the structural durability of the frontier model companies. Ever since the "DeepSeek moment"—which triggered a temporary 40% market correction in certain tech sectors—critics have argued that frontier labs are structurally vulnerable. The prevailing theory was that open-source models would rapidly close the intelligence gap, and intelligent model routing would allow enterprises to easily divert their workflows to commoditized, cheap tokens. Yet, despite 18 months of these arguments, the empirical data on the ground shows the exact opposite. As highlighted in online discussions by figures like Jesse Zhang, the actual economic value and "share of wallet" are increasingly concentrating within the top frontier labs, even while the total volume of commodity tokens grows elsewhere.
This paradox can be understood through the lens of consumer technology cycles. When the iPhone was a novelty, consumers eagerly upgraded every year because the performance leaps justified the price. Eventually, the technology reached a baseline level of maturity where users decided their current phone was "good enough" and stopped upgrading. AI models are bound to hit a similar inflection point. Currently, choosing a model like Claude 3.5 Sonnet over a cheaper alternative is driven by the fact that lower-tier models often feel "nerfed" or refuse to answer complex, specialized research queries. However, once a CFO gets fully integrated into corporate AI budgeting, the pressure to optimize costs will spark a completely different conversation.
- **The DeepSeek Shockwaves:** A pivotal market event that led many analysts to predict open-source models would completely commoditize frontier AI labs.
- **Economic Concentration:** Despite the availability of cheap open weights, enterprise financial commitment ("share of wallet") is still heavily favoring top frontier providers.
- **The "Good Enough" Threshold:** A future consumer shift where users may stop chasing frontier models once mid-tier models satisfy standard daily tasks.
---
[30:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1800s) Sovereign AI Strategies and Enterprise Implementation
The geopolitical landscape is shifting as global leaders realize they cannot rely solely on proprietary American models. Following recent United Nations commissions featuring tech leaders like Marc Benioff, Jensen Huang, and Brad Smith, it has become clear that virtually every country is actively formulating its own "sovereign AI strategy." Nations are highly hesitant to subject their core infrastructure to foreign technical or regulatory risks. Instead, they are eager to leverage open-source foundations—often standing up custom full-stack solutions from the silicon up using NVIDIA hardware—to serve their domestic companies and citizens. If an open-source model performs at 95% or 99% the level of a frontier American model, many sovereign nations will gladly accept that trade-off to ensure total operational independence.
```
+-------------------------------------------------------------------------+
| Sovereign AI Stack Model |
+-------------------------------------------------------------------------+
| 1. Domestic Regulatory Compliance & Infrastructure Guardrails |
| 2. Custom Enterprise Implementations (e.g., DoorDash Coding Benchmarks) |
| 3. Open-Source Foundations & Model Routing (e.g., Claude + Kimmy 2) |
| 4. Sovereign Hardware & Infrastructure Layer (e.g., NVIDIA Clean Slate) |
+-------------------------------------------------------------------------+
```
Simultaneously, enterprise implementation of open-source architectures remains highly complex compared to the frictionless deployment of a pre-approved enterprise tool like Anthropic's Claude. Configuring a decentralized network, setting up OpenRouter, and establishing dynamic fallback routing can take hours of specialized engineering time. While elite tech firms have the talent to pull this off, many traditional corporations simply lack the technical capacity. A prime example of sophisticated execution comes from Andy Fang, the CTO of DoorDash. DoorDash publically shared its internal software engineering benchmarks, revealing a hybrid approach: they route their most complex, high-level code reviews to Anthropic's frontier models while successfully offloading lower-level, volume-heavy tasks to secondary models like Kimmy 2, maintaining strict code quality without blowing out their budget.
- **Sovereign Tech Independence:** Nation-states are rejecting closed American models in favor of building domestic AI stacks to mitigate geopolitical risks.
- **DoorDash Hybrid Architecture:** CTO Andy Fang implemented an internal routing system that couples Anthropic's frontier capabilities with Kimmy 2 for lower-tier tasks.
- **The Impending Earnings Test:** The true resilience of corporate AI budgets will be tested during macro-economic downturns, where cutting software token spend is far easier for executives than executing staff layoffs.
[35:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=2100s) Zuck's Price War and the Premium Token vs. Commodity Debate
The battle lines in the AI landscape are shifting from raw intelligence benchmarks to economic warfare, highlighted by Meta’s aggressive new positioning. Mark Zuckerberg is pivoting his strategy away from purely open-sourcing models to "scorch the earth" and is instead triggering a direct price war. Using his old college social media handle, `@finkd`, Zuckerberg posted an unprecedented flurry of tweets to announce Meta's Spark 1.1 model—a powerful agentic coding tool priced at "one one hundredth of the cost" of comparable frontier models. Meta aims to aggressively enter the token distribution space through its new Meta Model API.
However, the Besties argue that this cost reduction may not matter for high-stakes enterprise tasks. While a lagging or commodity model is perfectly fine for basic tasks like summarizing a document—which might take 20,000 cheap tokens—high-consequence workloads present a totally different mathematical equation.
> "If you're talking about replacing a software engineer for two hours, that may take two million expensive tokens. And the consequence of using something that's 95% as good is really high... if the task breaks early, or it breaks in the middle, or it breaks at the end, there's a huge cost to that."
If an AI agent is replacing a specialized consultant who bills at $200 an hour, a company competing on quality will not care about the trivial cost difference between paying $3 for a cheap, error-prone model or $15 for a bulletproof, premium frontier model. Millions of independent market actors are validating this with their wallets every day; despite speculation that the intelligence gap between commodity and frontier models is collapsing, enterprise revenue for premium closed models continues to skyrocket.
Meanwhile, global sovereign entities are aggressively funding their own alternatives to escape American tech dominance:
* **United Arab Emirates (UAE):** Shipping its own open models via the Abu Dhabi Technology Innovation Institute's Falcon project.
* **Saudi Arabia:** Investing heavily in specialized Arabic Large Language Models (LLMs).
* **Japan:** Investing $6 billion into a massive consortium named "Neo Terra," specifically aiming to skip ahead directly to physical AI and robotics.
---
[40:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=2400s) The Reality of Enterprise Token Routing and the Hybrid Architecture Illusion
David Sacks weighs in on how enterprise Chief Technology Officers (CTOs) and Chief Financial Officers (CFOs) are reacting to skyrocketing compute bills. While there is a massive collective desire to shift token consumption to cheaper open-source options, a stark gap remains between corporate ambition and technical execution. Highly sophisticated tech companies like Coinbase and DoorDash have successfully built custom middleware and token-routing layers to automatically triage simple tasks to mundane models while escalating complex workflows to frontier labs. For the vast majority of legacy enterprises, however, "the spirit is willing, but the flesh is weak." They simply do not possess the internal engineering talent to build these intelligent routing systems.
Consequently, open source's share of total enterprise wallet spend actually decreased from 19% last year down to 11% this year. Though raw open-source usage is skyrocketing, actual dollar spend remains overwhelmingly concentrated in closed frontier labs because they offer ultimate convenience, a refined product, and a headless architecture that requires no heavy lifting from enterprise IT departments.
Furthermore, a framework proposed by the founder of Decagon highlights that open-source models are highly effective for mature, predictable use cases where data sets are clearly defined and post-training can be precisely targeted. For example, an AI agent handling customer support doesn't need to know advanced physics, so it can eventually be optimized down to a small, hyper-trained model. However, for immature use cases and general technological discovery, companies have no choice but to rely on the raw general intelligence of frontier labs.
Compounding this stickiness, Palo Alto Networks CEO Nikesh Arora points out that true model fungibility—the dream of seamlessly "hot-swapping" one model out for a cheaper one on the fly—is currently blocked by a major technical hurdle: nobody has figured out how to abstract away memory, context, and conversation history from the underlying model itself.
---
[45:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=2700s) The Self-Recursive Lead: Will the Intelligence Gap Ever Close?
Ali Ghodsi, the founder of Databricks, recently revealed that optimization doesn't just happen at the model level, but through the surrounding architecture. Databricks discovered that by keeping the exact same underlying model but swapping out the operational "harness" (using tools like GLM 5), enterprises can cut their token costs cleanly in half. This hyper-optimization mirrors the experiences of developers who notice up to an 80% reduction in token usage once they instruct their custom agentic applications to run analytics, manage memory states, and optimize their daily automated routines.
While an entire ecosystem of inference clouds and hyperscalers (such as BaseTen and Fireworks AI) is emerging to provide tools for model fungibility, a deeper philosophical question divides the tech industry: Is artificial intelligence actually converging?
While standard benchmarks suggest commodity models are catching up to frontier labs, the highly asymmetrical revenue distribution suggests otherwise. The non-consensus, contrarian view is that frontier intelligence will not converge at all. Instead, superintelligence could become fully self-recursive.
As models take on increasingly complex, multi-hour agentic tasks, the top-tier frontier labs might permanently extend their lead rather than letting the gap close.
Illustrating the dizzying speed of real-world adoption, Anton, the CEO of Lovable (an AI service allowing users to "vibe code" software applications), reported that his company's revenue exploded from $0 to $350 million in its first two years, and has now scaled from $100 million to $600 million in revenue over a 30-month period. Both Lovable and ElevenLabs (led by CEO Mati) are spending tens of millions of dollars directly with frontier labs. Despite lingering corporate anxieties regarding data leakage and the threat of frontier labs eventually building competing native features, these hyper-growth startups remain heavily dependent on premium closed models to power their core engines.
Here is a vivid and detailed walkthrough of this portion of the transcript, capturing the technical debate, the geo-political chess match, and the historic policy rollout with its accompanying banter.
---
Verticalized AI Models and the Rise of the Duopoly
The conversation opens with a discussion about the shifting landscape of AI adoption for major enterprises. Both companies mentioned—major customers of the frontier labs—are actively working on their own in-house models. Their explicit goal is to transition away from frontier models in favor of proprietary, specialized solutions. Creating verticalized models (tailored for specific tasks like voice or code generation) is becoming easier every six months. Consequently, eight- and nine-figure enterprise customers are expected to migrate away from frontier labs.
However, a strong counterpoint is raised regarding ElevenLabs. Given the highly competitive marketplace, it is questionable whether a company striving to provide the best voice agent in the world can afford to use an inferior, self-built model just because it is cheaper. If they manage to build something genuinely superior, the strategy holds, which is exactly what ElevenLabs believes it is achieving.
This raises the question of whether model capabilities will truly converge. The ease of building proprietary models may ultimately depend on how discrete and predictable the use case is. For example, Decagon, a customer support AI company, now routes 90% of its usage to open models. They can do this because they have the opportunity to post-train and heavily customize those open models using their own data and learnings.
This highlights a potential pattern based on the maturity of the use case:
* **Immature Use Cases:** Enterprises use the most powerful frontier models available while they are still figuring out the workflow.
* **Mature/Defined Use Cases:** Workloads are shifted to purpose-built, post-trained open models to achieve independence and "AI sovereignty."
The market today heavily reflects a duopoly measured by revenue, with only two companies capturing meaningful market share based on token revenue. Anthropic is hovering around $60-something billion in ARR, while OpenAI is at $40-something billion in ARR. Because every token they serve represents a real-world use case, these top two labs are learning continuously, potentially widening the gap between themselves and everyone else. While there appeared to be five major labs a year ago, it has rapidly consolidated into a top two, fitting the historical tech trend of shifting toward monopoly or duopoly.
Conversely, looking strictly at revenue can be misleading. Open source utilization represents "dark tokens" that do not show up as traditional revenue. For instance, DoorDash's spending on Anthropic or OpenAI is visible, but their extensive use of open-source models deployed on their own commoditized hardware remains hidden. This utilization shows up instead on the balance sheets of companies like NVIDIA, Cerebras, and alternative cloud providers (NeoClouds, Caruso Cloud) hosting the infrastructure.
---
The Geopolitical Chessboard: China's Open-Source Crackdown
The discussion shifts to global AI sovereignty following a pair of anonymously sourced Reuters reports regarding AI regulations in China. According to the scoop, Chinese Communist Party (CCP) regulators met with major tech players Alibaba, ByteDance, and Zhipu AI (the creators of the GLM 5.2 model). The regulators discussed restricting overseas access to China's top open and closed AI models. Additionally, the CCP is making any theft or leaks of AI research a national security offense and tightening control over who can fund Chinese AI labs. A stark example of this control occurred when the CCP pulled employees of Manus, a Chinese AI company, back to China after they attempted to relocate to Singapore.
The primary driver behind these restrictions is a deep concern over "mythos." Chinese authorities are worried that external forces will exploit software vulnerabilities and that Washington might deploy an AI model against Chinese national interests.
This marks a complete reversal from previous debates about whether the United States should ban open models. The incentive structure for Chinese AI firms appears to follow a distinct pattern: stay open-source until you catch up or get close to the Western frontier, and then transition to closed-source to capture all the commercial value for yourself. This mimics the exact trajectory Sam Altman executed at OpenAI when transitioning from a non-profit with open models to a for-profit with closed models. Similarly, Google utilized this playbook with Android—keeping it open to build a massive developer community and drive reinforcement learning, only to tighten licensing restrictions and bundle Google services once dominance was achieved.
The current status of top Chinese models reflects this shift:
* **ByteDance:** Their top model (the ChatGPT equivalent in China) has always been closed-source.
* **Alibaba (Qwen):** Originally open-source, now reportedly moving to a closed-source model.
* **Zhipu AI (GLM):** Caught up significantly to American frontier capabilities in specific tasks while open, but is now moving to a closed model.
In Washington D.C., there is absolute, unified agreement across the political spectrum—all the way up to the President—about the absolute necessity of staying ahead of China in the AI race. However, concerns remain regarding lower-level bureaucrats who might inadvertently implement counterproductive regulations or bans due to political pressure from the AI "doomer" community.
Furthermore, evidence shows that Chinese models have been distilling American frontier models, as GLM 5 reportedly contains distinct watermarks from Western systems. The U.S. government is expected to take steps to penalize and prevent this distillation. Paradoxically, the ultimate gating factor for American token generation and inference might not be software or chips, but energy. Team analysis shows that expected load growth between now and 2050 leaves the U.S. about three entire California's worth of energy short. This energy vulnerability extends to Taiwan, which relies heavily on Liquified Natural Gas (LNG). A Chinese blockade could cause Taiwan to run out of energy within two to three weeks, completely disrupting global chip supply lines.
> "The absolute best thing that could happen for America in terms of winning the AI race against China is if China somehow sprouted their own doomer community... We need to get their P-Doom up."
---
The Launch of "Trump Accounts" and the Invest America Act
Brad Gerstner shares a massive updates from his recent trip to Washington D.C. Celebrating a four-year policy mission, the official "Trump Accounts" app (associated with the Invest America Act passed via reconciliation a year prior) officially went live on July 4th. To mark the launch, a historic joint bell-ringing between the NYSE and NASDAQ was conducted directly from the Oval Office, attended by the President, 100 CEOs, and affected families.
The app immediately became the number one downloaded app in the world, generating massive traction within its first day on the market:
* Over **1.5 million accounts** created within the first 24 hours.
* Over **$1 billion in total deposits** recorded in the initial launch window.
The underlying mechanics of a Trump Account are straightforward: every child receives a privately owned investment account at birth tied to their Social Security number, seeded with a $1,000 contribution. If matched and supplemented by a modest savings rate of $10 a week, the account compounds in the S&P 500 to roughly $50,000 by the time the child reaches age 18. The accounts are entirely free for the lifetime of the recipient. Upon turning 18, the youth can deploy the capital toward higher education or roll it over into a retirement account like a Roth IRA. The President has even called to accelerate the program to auto-create accounts for all 50 million to 70 million American children under the age of 18.
The initiative has drawn massive corporate and philanthropic backing. Michael and Susan Dell anchored the launch, contributing over $1 billion to fund accounts for 25 million lower- and middle-income children. SpaceX President Gwynne Shotwell joined the effort by contributing $350 million worth of her personal SpaceX shares to target children in lower-income communities. Micron added $250 million, offering up to $1,000 per employee as an employer contribution match, and Brad personally funded accounts for every eligible child in the state of Indiana.
The hosts engage in some classic poker-game ribbing, laughing at the irony of Brad—who famously rage-quits poker games over a $6,000 loss—willingly dropping $100 million of his own money into this program. Brad defends the scale of the move, noting that this platform cuts out the traditional charitable middlemen, allowing billionaires to distribute wealth directly to America's children at an unprecedented scale. They project raising $10 billion in direct philanthropy within the first 12 months.
The segment concludes with a critique of the political pushback the initiative has faced online. Because the accounts bear Donald Trump's name, certain critics suffering from "Trump Derangement Syndrome" (TDS) on platforms like BlueSky have expressed a refusal to open them for their children—a stance the hosts label as patently insane and irresponsible given that these accounts are projected to compound to millions of dollars by the time the recipient reaches adulthood.
---
[01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) The Launch of Trump Accounts and Overcoming TDS
The conversational focus shifts squarely onto the "Invest America Act," colloquially and officially known as "Trump accounts." Host Jason Calacanis brings up the immediate online blowback, noting that some people are highly skeptical simply because of the branding. Guest Brad Gerstner responds directly, clarifying that parents on the ground are completely ignoring the political noise. Sign-ups are spanning the entire economic and political spectrum because parents simply care about securing their kids' connection to the American dream.
Brad reveals that during a recent lunch with Donald Trump and the Treasury Secretary, the President was intensely focused on speed and inclusivity. Trump specifically ordered Brad to find a way to automatically create these accounts using under-18 Social Security numbers so that busy, working-class parents wouldn’t miss out. Brad states that their goal is to automatically establish 50 to 70 million accounts over the next 90 days. When asked if left-wing politicians like Bernie Sanders or Ro Khanna are trying to block the initiative out of spite, Brad points out that prominent Democrats—including Senator Cory Booker, Governor Gavin Newsom, Governor Wes Moore, and Senator John Fetterman—have already broken ranks to support it.
- **A Tale of Two Worldviews:** Brad frames the political divide between the dependency model of Bernie Sanders (taxing corporations to redistribute money through Washington) versus the universal ownership model supported by the Trump administration and centrist Democrats.
- **The Core Philosophical Antidote:** As Brad told the President directly: "The antidote to more socialism is more capitalism."
- **Equity Nation:** Calacanis points out that if this rollout succeeds, US equity ownership could realistically surge from the current 50% to 75% of the population.
[01:13:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=4389s) The Tax Advantages and Wealth Generation Mechanics
David Sacks weighs in, calling the program an incredible new philanthropic platform and a vital antidote to populist anger against billionaires. However, Sacks believes the "freebie" philanthropic aspect is actually overshadowing the staggering tax advantages, which have set "CPA Twitter" ablaze. Experts are comparing Trump accounts to a mandatory financial playbook, ranking them alongside employer-matched 401ks, Health Savings Accounts (HSAs), and Roth IRAs as absolute "must-dos."
Sacks breaks down the exact financial mechanics of the accounts, with Brad validating the numbers:
- **Contribution Limits:** Parents, friends, and family can contribute up to $5,000 annually per child under the age of 18.
- **Employer Matching:** Corporate employers can contribute up to $2,500 completely tax-free. Sacks advises employees to immediately ask their bosses to divert $2,500 of their existing salary into a Trump account to instantly lower their taxable income.
- **The College Student Loophole:** Because the account compounds tax-free until age 18, it can eventually be rolled over into a traditional or Roth IRA. Sacks highlights a clever strategy circulating among CPAs: wait to do the Roth conversion until the child is no longer a dependent and sits in a 0% tax bracket (e.g., during or right after college). This allows families to shift the wealth into a Roth IRA incredibly cheaply.
- **The Compound Effect:** With $200,000 to $300,000 accumulated by age 18, standard market compounding would easily turn these children into millionaires by age 60, completely bypassing the need to rely on Social Security. "We're gonna have a lot of rich kids with trust funds," Sacks jokes.
[01:16:51](https://www.youtube.com/watch?v=PHL1j2ti420&t=4611s) Voluntary Corporate Donations and the Main Street Agenda
Before abruptly leaving the podcast to hop on an enterprise software sales call, Chamath Palihapitiya throws out a parting idea: Brad should convince massive AI frontrunners like OpenAI and Anthropic to donate corporate equity directly into the Trump accounts of everyday kids. The hosts joke about Chamath's hustle, reciting the classic sales mantra, "Always Be Closing."
Brad builds on Chamath's idea, confirming he has already spoken to tech leaders like Dario Amodei, Sam Altman, and Elon Musk about corporate donations. Crucially, Brad emphasizes that these contributions must be voluntary rather than forced into a government "slush fund" controlled by politicians like Bernie Sanders or Alexandria Ocasio-Cortez. Because individual accounts have a $5,000 annual limit, massive corporate donations would be held in a pooled account and systematically distributed to existing children and the roughly 3.5 to 3.7 million infants born in the US each year. Within a decade, the platform is projected to host over 100 million private, compounding accounts.
- **Gamifying Financial Literacy:** The app—originally designed by Brad and his sons, with backend implementation help from Vlad Tenev at Robinhood—allows middle and high school students to see their personal portfolios on their phones. "Imagine opening up that account... and seeing, oh man, I'm in the game. I have ownership," Brad notes.
- **The Philanthropic Pipeline:** Billionaires like Michael Dell and SpaceX's Gwynne Shotwell are using the tool for direct giving. For instance, if Shotwell distributes 2 million shares of SpaceX stock to 2 million kids, each child instantly receives a stock asset worth $150. Sacks notes this is infinitely better than routing money through the traditional "NGO industrial complex," where organizations siphon off 40% of donations for executive salaries and nice offices.
- **Historic Policy Shift:** Brad calls the Invest America Act the most monumental transformation of the American social contract since the introduction of Social Security in 1935, specifically targeting the 50% of Main Street families who have historically felt completely left behind by Wall Street.
[01:21:40](https://www.youtube.com/watch?v=PHL1j2ti420&t=4900s) Overcoming the K-Shaped Recovery
Calacanis steps in to level up the conversation, stating that while he is frequently critical of the Trump administration's policy missteps, the partisan whining over the name "Trump accounts" needs to stop. He references a previous poker game where he explicitly told Brad to lean into the branding. Calacanis argues that regardless of an individual's personal "Trump Derangement Syndrome" (TDS), the country is currently facing a severe, unstable K-shaped recovery. The deep-seated tension between the haves and the have-nots has left working-class Americans feeling entirely alienated by modern capitalism, making structural innovations like this absolutely vital for national cohesion.
[01:25:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5100s) Reclaiming Capitalism and the "Trump Accounts"
The conversation shifts into an impassioned defense of capitalism over alternative ideologies, framed around the 250th anniversary of the United States. The hosts argue that young generations are losing faith in the American experiment, making the launch of the "Trump Accounts" a vital mechanism to re-engage them in the free market. Rather than looking for government handouts like free rides or free pizza, or attacking wealthy figures like Ken Griffin, this initiative allows every child to actively participate in entrepreneurship. It is described as the ultimate evolution of the Giving Pledge; instead of billionaires merely donating their wealth at the end of their lives, philanthropy is strategically deployed so that every single young American becomes a shareholder in the nation's economic success.
The hosts compare this framework to Australia's highly successful superannuation system, which mandates that individuals contribute a percentage of their income into a directed retirement account. This forced savings model creates a deep sense of safety and financial security for families, ensuring that children have a guaranteed head start regardless of their parents' socioeconomic status. Brad Gerstner is widely praised for his masterful execution in navigating the divisive political landscape to build a foundational architecture that could eventually supplement or transform structural platforms like Social Security.
* **The Giving Pledge Alternative:** Seamlessly channels wealth from elite donors—including Michael and Susan Dell—directly into accounts for the broader public, turning America into an "equity nation."
* **Australia's Superannuation:** Noted as a model for high societal happiness and safety due to compulsory 12% to 14% retirement contributions.
* **The Promise of Upward Mobility:** Designed to alleviate the deep financial anxiety experienced by single parents and minimum-wage workers by securing their children's future early.
[01:29:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5340s) Preventing the "YOLO" Effect and State-Level Action
Addressing potential criticisms regarding financial responsibility, the Besties debate what happens when a recipient turns 18 and faces the temptation to squander the funds. While initial design ideas favored locking the capital until age 30 to allow for greater emotional and financial maturity, political consensus dictated that individuals old enough to vote and fight in wars must have agency over their money. However, structural safeguards are embedded: users can only withdraw up to 25% of the funds for specific, productive milestones—such as purchasing a primary home, starting a business, or financing higher education. The remaining 75% automatically rolls over into a traditional IRA, which carries strict early-withdrawal penalties to disincentivize reckless spending.
This platform is also positioned as a revolutionary tool for public education and financial literacy. Because 37 states already mandate financial literacy in their curricula, having real money in an account shifts the subject from an abstract concept to an engaging reality for a seventh grader tracking shares in companies like Nike, Apple, or Nvidia. Furthermore, the momentum is rapidly expanding past federal policy into state-level action, with multiple states preparing to reallocate underperforming block grants directly into children's accounts.
* **Early Withdrawal Guardrails:** Capital usage is restricted at age 18 to a maximum of 25% for a home down payment, higher education, or business creation.
* **State-Level Adoption:** Roughly 25 states, including Oklahoma, West Virginia, and Indiana, are planning to supplement these accounts by redirecting ineffective state spending directly to kids.
* **The "Snowball" Effect:** Capitalizes on the zero-to-25 age window, capturing the most critical and historically neglected phase of compound interest.
[01:32:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5520s) Designing a World-Class Consumer Product for Government
Brad Gerstner reflects on the deeply personal journey of bringing this initiative to life, tracing its origin back to a kitchen-table sketch with his two sons in the fall of 2020. The project culminated in an exceptional public-private design collaboration within the administration's first 18 months. Major credit is given to Airbnb co-founder Joe Gebbia for stepping up to lead the product design, alongside a core team that included Treasury Secretary Luke Pettit, Vlad Tenev, and Michael Dell. Their explicit goal was to reject the standard, friction-heavy quality associated with bureaucratic state software and instead build one of the premier consumer products on the market, achieving stellar ratings and massive day-one engagement.
The achievement is highlighted as a rare, genuinely bipartisan victory for American youth. Figures from both sides of the political spectrum, such as Democratic Governor Wes Moore of Maryland, acknowledged that both parties had been attempting to construct a wealth-building vehicle of this nature for 40 years. The success serves as a direct rebuttal to critics who express a desire to dismantle traditional economic structures, demonstrating that the ultimate path forward lies in evolving and doubling down on the 250-year-old formula of American capitalism.
* **The Core Product Team:** Composed of Brad Gerstner, Michael Dell, Airbnb's Joe Gebbia, Robinhood's Vlad Tenev, and Luke Pettit from the Treasury.
* **Bipartisan Validation:** Prominently praised by Maryland Governor Wes Moore, signaling cross-ideological support for early-childhood equity accounts.
* **Product Standard:** Engineered to match the engagement metrics, interface quality, and user retention of top-tier Silicon Valley technology applications.
[01:36:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5760s) Expanding Equity to the Middle Class and Future Horizons
The final segment outlines the vast runway ahead for expanding this equity framework to older demographics. While Social Security remains a sacred, untouchable political promise for current recipients, the Besties criticize its mechanics as a "black hole of government" where citizens lack title, ownership, control, or the ability to pass assets down to heirs. In contrast, expanding the Trump Accounts to allow individuals between the ages of 20 and 40 to make supplemental, tax-advantaged contributions would create a transparent, wealth-building tool that citizens genuinely own.
Certified Public Accountants (CPAs) and financial planners are already highlighting the account as an extraordinary tool for middle-class family planning. By effectively creating an "all-child IRA" that allows contributions from parents, employers, and philanthropists alike before a individual even secures their first official job, the initiative successfully reclaims the first third of the compounding hill. The episode concludes with the hosts celebrating this milestone on America's 250th birthday, reiterating the tech industry's duty to build pathways that lift up those at the bottom of the economic ladder.
* **Structural Difference from Social Security:** Trump Accounts guarantee direct legal title and asset transferability to a participant's heirs, unlike standard payroll tax contributions.
* **The Market Gap Solution:** Bypasses the traditional requirement that an individual must have active W-2 wage earnings to open and utilize a tax-advantaged retirement account.
* **Future Outlook:** Conversations are underway to scale the platform into a lifetime supplemental IRA for citizens up to age 40 to participate in national technology and manufacturing gains.
## Notable quotes
> [13:02](https://www.youtube.com/watch?v=PHL1j2ti420&t=782s) Right now, our token costs are doubling every 45 days... our upside is essentially flat... we've effectively already asymptoted. I suspect that if you can get out now, you should get out now before all of that starts to seep into the water table. — Chamath Palihapitiya
> [25:05](https://www.youtube.com/watch?v=PHL1j2ti420&t=1505s) When you have unlimited tokens as an exercise... it really does change your thinking when costs go down.
> [35:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=2100s) If you're talking about replacing a software engineer for two hours, that may take two million expensive tokens. And the consequence of using something that's 95% as good is really high... if the task breaks early, or it breaks in the middle, or it breaks at the end, there's a huge cost to that.
> The absolute best thing that could happen for America in terms of winning the AI race against China is if China somehow sprouted their own doomer community... We need to get their P-Doom up.
> [01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) The antidote to more socialism is more capitalism. — Brad Gerstner
> [01:13:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=4389s) We're gonna have a lot of rich kids with trust funds — David Sacks
> [01:16:51](https://www.youtube.com/watch?v=PHL1j2ti420&t=4611s) Imagine opening up that account... and seeing, oh man, I'm in the game. I have ownership — Brad Gerstner
## Mentioned
Jason Calacanis, David Friedberg, Brad Gerstner, Chamath Palihapitiya, David Sacks, Raise Conference, United Nations Commission for AI, Jensen Huang, Brad Smith, Anthony Tan, Grab, Marc Benioff, Salesforce, Tom Brown, Anthropic, SpaceX, All-In Liquidity Summit, Polymarket, Gavin Baker, OpenAI, GPT-6, Elon Musk, Brett Johnsen, Gwynne Shotwell, Altimeter, Fidelity, T. Rowe Price, Fable 5 High, Nvidia, S&P 500, Praveen Anidil, Uber, Perplexity, Snowflake, Microsoft Excel, Nous Research, Hermes, OpenRouter, GLM, Bittensor, TAO, GLM 5.2, Groq, Cerebras, Jevons Paradox, DeepSeek, Jesse Zhang, Claude 3.5 Sonnet, Andy Fang, DoorDash, Kimmy 2, Mark Zuckerberg, Spark 1.1, Meta Model API, Falcon project, Abu Dhabi Technology Innovation Institute, Neo Terra, Coinbase, Decagon, Nikesh Arora, Palo Alto Networks, Ali Ghodsi, Databricks, GLM 5, BaseTen, Fireworks AI, Anton, Lovable, ElevenLabs, Mati, NeoClouds, Caruso Cloud, Alibaba, ByteDance, Zhipu AI, Manus, Sam Altman, Google, Android, Qwen, Bernie Sanders, Ro Khanna, Cory Booker, Gavin Newsom, Wes Moore, John Fetterman, Donald Trump, Alexandria Ocasio-Cortez, Vlad Tenev, Robinhood, Michael Dell, Susan Dell, Micron, BlueSky, Ken Griffin, Nike, Apple, Joe Gebbia, Airbnb, Luke Pettit
The All-In Podcast examines the financial sustainability of frontier AI labs preparing for multi-trillion-dollar public listings amid emerging enterprise ROI questions, alongside the nationwide rollout of the Invest America Act's universal child investment accounts.
Trillion-Dollar AI IPOs and Frontier Market Concentration (3:09)
SpaceX's public offering—raising $5 billion at a $2 trillion valuation on approximately $35 billion in forward revenue—has established an operational template for mega-cap tech listings. Attention has shifted to Anthropic, which filed confidentially for an IPO on June 1, with Polymarket displaying a 65% probability of a debut before the end of 2026. Tech investor Gavin Baker projects Anthropic could exit 2026 with over $100 billion in annualized revenue, supporting a potential $3 trillion valuation, while OpenAI is reportedly pacing toward $70 billion in annual revenue with next-generation models imminent. Brad Gerstner projects that both Anthropic and OpenAI can compound revenue at more than 30% annually, though he cautions that entering public markets at multi-trillion-dollar baselines shifts investor upside from speculative spikes toward long-term compounding.
The podcast hosts observe that token revenues are consolidating into an effective duopoly led by Anthropic and OpenAI. While open-source model volume is growing, visible commercial software spend remains concentrated in top closed providers, whose continuous real-world use cases create a self-reinforcing intelligence advantage.
Enterprise AI ROI and Compute Cost Discipline (13:02)
Chamath Palihapitiya warns that enterprise token expenses are doubling every 45 days while organizational productivity gains have flattened near 5%. Citing a query conducted with Anthropic's Fable 5 High model that stripped out Nvidia's earnings, Palihapitiya notes that S&P 493 EPS growth was only 9%—driven primarily by pricing power and buybacks—placing actual enterprise AI ROI between 0% and 2%. While consumer applications face minimal scrutiny, enterprise IT budgets are expected to undergo rigorous financial audits.
In response, companies are restructuring AI usage away from ad-hoc employee prompting toward structured engineering pipelines. Uber CTO Praveen Anidil revealed that 99% of Uber's engineering workforce uses developer AI tools, generating over 70% of pull requests across 2,500 specialized agentic skills. Despite hardware optimizations from Cerebras and Groq that cut token inference costs by up to 95%, the hosts argue that enterprises will still pay premium rates for frontier models in mission-critical workflows where execution failures carry high financial costs.
Open-Source Commoditization and Sovereign AI Geopolitics (35:00)
Even as Meta aggressively prices its Spark 1.1 model at one-hundredth the cost of frontier alternatives, David Sacks notes that open source's share of enterprise wallet spend declined from 19% to 11% year-over-year. Sacks attributes this to the reality that legacy enterprises lack internal software talent to architect custom model routing. Nikesh Arora adds that seamless model hot-swapping is restricted because memory and conversation state cannot yet be decoupled from underlying models.
Geopolitically, nation-states are developing independent sovereign AI strategies to avoid reliance on closed US infrastructure, including initiatives in Saudi Arabia, the UAE, and Japan's $6 billion Neo Terra robotics consortium. Reports also indicate Chinese regulators are moving to restrict foreign access to models from Alibaba, ByteDance, and Zhipu AI, following an established playbook of leveraging open source to reach parity before closing access to commercialize. In the US, the hosts project that power availability, rather than silicon, will serve as the primary bottleneck, with national load growth facing an estimated shortfall equivalent to three Californias by 2050.
The Invest America Act and Universal Equity Accounts (70:00)
Brad Gerstner reports on the federal launch of 'Trump Accounts' under the Invest America Act, which recorded over 1.5 million account creations and $1 billion in deposits within 24 hours. The program opens a privately owned account at birth tied to a child's Social Security number, seeded with $1,000 and projected to compound to roughly $50,000 by age 18 with supplemental savings of $10 per week. Initial philanthropic commitments include $1 billion from Michael and Susan Dell for 25 million children, $350 million in SpaceX equity from Gwynne Shotwell, and $100 million from Gerstner.
Under program guidelines, parents and relatives can contribute up to $5,000 annually, alongside tax-free employer matches of up to $2,500. Structural guardrails restrict withdrawals at age 18 to a maximum of 25% for home purchases, higher education, or business formation, while the remaining 75% automatically rolls into an IRA. Gerstner and Sacks position the policy as a mechanism to expand equity ownership from 50% to 75% of the US population, bypassing traditional nonprofit overhead through direct capital transfers.
The discussion leaves unresolved whether frontier AI labs can justify multi-trillion-dollar valuations if enterprise productivity gains continue to lag compute costs, and whether universal equity accounts can maintain cross-partisan durability to reshape the American social safety net.
## Background
本期节目由知名商业科技播客《All-In Podcast》于 2026 年 7 月 27 日发布在 YouTube 平台上。视频简介指出，本期由 Altimeter Capital 创始人 Brad Gerstner 代班休假的 Friedberg，核心议题涵盖 OpenAI 与 Anthropic 的万亿级 IPO 竞逐、扎克伯格挑起的模型价格战与开源闭源分歧、中国或限制顶级模型出海的动向，以及旨在让年轻一代重回资本主义的“特朗普账户”（Trump Accounts）正式落地。
## TL;DR
节目首先对标 SpaceX 模版，深入剖析了 Anthropic 与 OpenAI 冲刺万亿级 IPO 的路径，并尖锐指出了企业端面临的 Token 成本飙升与真实 ROI 考核压力；随后探讨了开源模型与前沿闭源双寡头的博弈，认为即便 Meta 等挑起价格战，前沿实验室在复杂工程中的不可替代性依然锁定了核心钱包份额；在分析了中美 AI 竞赛中的模型管控与电网能源瓶颈后，全场重点聚焦由 Brad Gerstner 主导推动的“特朗普账户”（《投资美国法案》）在白宫敲钟落地，详述其如何通过免税复利、雇主配资与去中介化穿透慈善，为全美未成年人打造覆盖一生的股权雪坡。
## Atmosphere & style
全场对话兼具顶级创投宏观洞察的极高信息密度与老友牌局般的嬉笑怒骂，节奏明快且极富张力。几位主持人既有对高估值泡沫与企业财报暴雷的冷峻推演，又充满对技术爆发与全民资本主义制度创新的澎湃激情；尤其当众人调侃 Brad 平日牌局抠门却豪捐一亿美元资助家乡儿童，以及回顾白宫椭圆形办公室敲钟仪式时，真挚的个人情怀与硅谷精英的现实算度交织在一起，给听众带来既生动接地气又宏大硬核的沉浸式体验。
## Bulletin
- [02:58](https://www.youtube.com/watch?v=PHL1j2ti420&t=178s) SpaceX 完成约 2 万亿美元估值的教科书式 IPO，为 Anthropic 和 OpenAI 的万亿美元级上市铺平了机制道路；传闻 Anthropic 年化营收冲刺 1000 亿美元并具备意外盈利能力，且已秘密提交上市申请。
- [27:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1620s) 企业在 AI 上的 Token 支出正面临 ROI 大清算：尽管开源模型在调用量上激增，但由于复杂长链路任务容错率低，闭源前沿实验室在实际付费总额（Share of Wallet）中占比逆势扩大并形成双寡头格局。
- [54:29](https://www.youtube.com/watch?v=PHL1j2ti420&t=3269s) 针对中国考虑限制前沿 AI 模型出海的传闻，嘉宾指出这符合“落后时开源圈地、追平后闭源变现”的经典商业周期，中美 AI 竞赛真正的瓶颈已从纯算法转向能源电力短缺与官僚监管壁垒。
- [01:03:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=3789s) 由 Brad Gerstner 历时四年推动的《投资美国法案》（公开品牌为“特朗普账户”）正式落地，为 18 岁以下儿童设立免税复利投资账户，上线 24 小时内全美建户超 150 万个、吸金逾 10 亿美元。
## Key takeaways
- [05:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=300s) SpaceX 的万亿美元上市确立了分阶段渐进式解禁与提前纳入指数的机制模版，直接为后续 Anthropic 和 OpenAI 的万亿级 IPO 扫清制度障碍。
- [10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=600s) Anthropic 聚焦企业级代码与推理而意外实现盈利，相较于业务发散且现金消耗高昂的 OpenAI，更可能先一步以超万亿美元估值叩开公开市场。
- [13:30](https://www.youtube.com/watch?v=PHL1j2ti420&t=810s) 剔除英伟达等美股七巨头后，AI 带来的标普其余 493 家企业实际 EPS 提振仅在 0% 到 2% 之间，企业采购预算随时面临断崖式收缩的风险。
- [17:15](https://www.youtube.com/watch?v=PHL1j2ti420&t=1035s) Uber 通过建立前向部署工程师与部门架构师深度绑定的 Agent 战队，将代码自动生成扩展至法务与运营全流程，提供了一条企业验证 AI 真实 ROI 的范式。
- [20:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1200s) 前沿模型实验室面临史无前例的万亿级 TAM，机器正在制造机器（如英伟达用 AI 设计新芯片），推理与查询优化反而加速了全业务渗透（Snowflake 悖论）。
- [22:38](https://www.youtube.com/watch?v=PHL1j2ti420&t=1358s) 自下而上的按需智能普及使得员工每月仅需花费相当于薪资 3% 到 5% 的成本，即可撬动数倍人效提升，绕过了传统缓慢的企业软件销售周期。
- [25:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1500s) 开源生态与专用 ASIC 芯片击穿推理成本 95%，推动自动化 Agent 从低频单次定时任务向每小时并发多 Agent 协同巡检质变。
- [27:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1620s) 大模型领域正在验证“杰文斯悖论”：虽然开源模型分食了廉价长尾 Token，但在高阶任务与知识替代中，前沿模型的绝对可靠性使其牢牢掌控了客户的核心钱包份额。
- [30:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1800s) 各国政府因技术受制风险加速推进“主权 AI”，倾向于依托开源模型自建端到端技术栈，95% 的“足够好”能力即可满足国家级基础设施需求。
- [42:47](https://www.youtube.com/watch?v=PHL1j2ti420&t=2567s) 多模型动态热插拔遭遇工程锁钥：由于历史对话、记忆与长上下文无法跨架构无损解耦迁移，限制了企业将任务无缝切换至廉价开源模型。
- [45:10](https://www.youtube.com/watch?v=PHL1j2ti420&t=2710s) “智能正在收敛”是一个假象：领先实验室借助通用智商优势攫取最高商业收入，从而反哺更大算力集群，自递归超智能飞轮正在持续拉大代差。
- [54:29](https://www.youtube.com/watch?v=PHL1j2ti420&t=3269s) 中国模型团队早期拥抱开源是落后阶段换取生态与反馈的常规策略，一旦基准逼近世界前沿走向闭源独享商业利益是技术巨头的历史必然周期。
- [01:00:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=3600s) 中美 AI 竞赛的终极物理决胜点已转向电力能源：到 2050 年美国电力常规负荷缺口相当于整整三个加州，而台湾赖以运转的液化天然气储备仅够支撑两到三周。
- [01:13:50](https://www.youtube.com/watch?v=PHL1j2ti420&t=4430s) “特朗普账户”通过每年最高 5000 美元私人捐赠与 2500 美元雇主免税注资，全额追踪标普指数免税复利，被财务界视为中产财富规划的顶级避税工具。
- [01:30:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5400s) 为防止年轻人在 18 岁盲目挥霍，该账户设定 18 岁时最多只能提取 25% 用于购房、创业或高等教育，其余 75% 强制滚入个人退休账户（IRA）继续复利。
- [01:36:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5760s) 特朗普账户补齐了传统 IRA 砍掉的人生前 25 年“黄金复利雪坡”，以直接所有权取代福利依赖，是自 1935 年社保法案以来美国社会契约最深刻的重构。
## Detailed walkthrough
### [00:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=0s) 开场寒暄：Friedberg 缺席与各自主场的“火热夏天”
节目一开场，Jason Calacanis 宣布这是 2026 年 7 月的全网第一播客《All-In》第 280 期。常规主持人之一 David Friedberg 正在休假，由老朋友、知名投资人 Brad Gerstner（Altimeter Capital 创始人兼 CEO）代班救场。Brad 调侃 Friedberg 可能在爱达荷州或其他地方度假，而自己则身处美国东北部的“国旗室”（Flag Room）里工作，刚在华盛顿特区度过了庆祝“建国 250 周年”（America 250）的一周。
Jason 则透露自己目前在巴黎，刚刚在 RAISE 峰会上密集录制了 8 场访谈，随后将上线播客主信息流。另一边，Chamath Palihapitiya 正在欧洲深入一线跑业务，自嘲正在经历属于自己的“SaaS 与企业软件火热之夏”（hot software summer）。大家打趣 Chamath 过去在做投资人时对 SaaS CEO 们要求苛刻，如今自己亲自带队在工厂般的前线卖软件，才真正体会到销售企业级软件的艰辛与不易。
```mermaid
flowchart TD
accTitle: 嘉宾动向与联合国 AI 委员会关系图
accDescr: 展示节目主持与嘉宾的地理分布及 Chamath 参与的日内瓦联合国 AI 委员会阵容
subgraph Locations["嘉宾当前动态"]
J["Jason Calacanis<br/>巴黎 RAISE 峰会录制"]
B["Brad Gerstner<br/>美国东北部 Flag Room"]
C["Chamath Palihapitiya<br/>欧洲奔波销售企业软件"]
end
subgraph Geneva["日内瓦活动：联合国 AI 委员会"]
MB["Marc Benioff<br/>(Salesforce 联合主席/组局人)"]
Members["委员会成员:<br/>• Chamath Palihapitiya<br/>• Jensen Huang (NVIDIA)<br/>• Brad Smith (Microsoft)<br/>• Anthony Tan (Grab)<br/>• Tom Brown (Anthropic 联合创始人)"]
MB --- Members
end
C --> Geneva
```
Chamath 分享了他在日内瓦的行程：他参加了由 Salesforce 创始人兼 CEO Marc Benioff 组织的晚宴。Benioff 担任联合主席组建了一个联合国 AI 委员会（UN Commission for AI），出席的大佬云集，包括英伟达 CEO 黄仁勋（Jensen Huang）、微软总裁 Brad Smith、Grab 联合创始人 Anthony Tan 以及 Anthropic 联合创始人 Tom Brown 等。Chamath 盛赞 Benioff 是“大师中的大师、活动家之王”（the impresario of impresarios），让人直观感受到他是如何将 Salesforce 打造成庞然大物的。在被调侃联合国委员会是否真能干成实事、Anthropic 人员是否在宣扬“世界末日论”时，Chamath 替 Tom Brown 澄清，称 Tom 是个真诚、出色的创业者，虽然将安全使命写在脸上，但全场并没有渲染末日恐惧；而全球大部分国家在 AI 议题上的真实诉求和未来趋势，普遍倒向了开源（Open Source）。
---
### [02:58](https://www.youtube.com/watch?v=PHL1j2ti420&t=178s) 万亿级 IPO 潮起：SpaceX 样板与 Anthropic、OpenAI 的冲刺
话题切入核心焦点：万亿美元体量的超级独角兽排队上市浪潮。Jason 指出，这曾是上个月流动性峰会（Liquidity Summit）的核心议题。今年已经见证了 SpaceX 这一人类历史上空前的万亿美元级 IPO：其股价曾一度冲上 200 美元，目前回落至 150 美元，恰好稳定在 IPO 发行价附近，以约 2 万亿美元的市值稳居全球第七大上市公司之列。
紧随其后的是生成式 AI 两大领头羊——Anthropic 和 OpenAI。Anthropic 已于 6 月 1 日提交了保密上市申请文件（Jason 调侃这种“保密申请”几乎在提交瞬间就人尽皆知）。根据预测市场 Polymarket 的数据，Anthropic 在年内完成 IPO 的概率约为 65%。此前知名科技投资人 Gavin Baker 曾预测，Anthropic 到 2026 年底的年化营收有望突破 1000 亿美元，且盈利能力极强，一旦上市当前市值就能冲击 3 万亿美元。
Chamath 分析指出，这些公司无疑都是伟大的企业，但关键在于“出清价”（market clearing price）以及二级市场吸收这种超大体量新股的承受能力：
- **OpenAI**：业务线较为发散，对 C 端消费级市场的依赖度高，上次披露财务状况时其现金消耗（Cash burn）依然处于非常高昂的状态。
- **Anthropic**：聚焦于企业级与代码推理，更有可能如 Brad 此前所言处于“意外盈利”（accidentally profitable）的状态。
为了说明企业端对 AI 支出的警惕，Chamath 举了一个极为生动的实操案例：他当天刚与自己的 CTO 讨论公司的 Token 消耗，CTO 抛出了一个令人震惊的现实——公司的 Token 成本每 45 天就翻一倍，然而换来的下游实际生产力提升最高只有 5%。CTO 坦言，底层大模型的能力已经遭遇平台期、出现渐近线效应（asymptoted），如今必须灌入指数级暴涨的 Token 数量，才能榨出微弱的技术迭代。Chamath 据此警告：整个产业在未来三到四年都会迎来这一场成本与产出失衡的“大清算”（reckoning）。因此，大模型公司若能在这种幻象彻底渗入资本底盘之前趁早高估值上市融走巨资，绝对是最理智的生存策略。
---
### [05:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=300s) 解码 SpaceX IPO 范式：指数提前纳入与解禁博弈
针对 Jason 提出的“在未来 6 到 9 个月内、于明年一季度日本滑雪季前，Anthropic 与 OpenAI 是否能 100% 上市”的问题，手握两家早期股份的 Brad Gerstner 给出了高度肯定的预期，认为概率极高（除非出现封锁台湾海峡等不可抗力的黑天鹅事件）。Brad 深入拆解了 SpaceX 上市给所有万亿级巨头树立的教科书级模版（Textbook IPO）：
SpaceX 上市核心指标与机制：
- **融资与估值**：在约 350 亿美元的前瞻营收（Forward revenue）支撑下，SpaceX 在 IPO 中募资约 200 亿美元，上市后市值飙涨 25%，稳定在约 2 万亿美元左右。
- **管理层操盘**：Elon Musk、Gwynne Shotwell 以及 Bret Johnsen 团队开创了多项针对超级体量公司的上市创新机制。
- **财务与体量对标**：若 Gavin Baker 对 Anthropic 年底营收冲破 1000 亿美元的传言属实，其次年 GAAP 营收将远超千亿规模，比 SpaceX 的 350 亿前瞻营收更为惊人，完全具备支撑 3 万亿估值超级 IPO 的底气。
针对市场争议极大的“提前纳入指数”（Early index inclusion）与“锁定解禁期设计”（Lockup structure），Brad 作出了深度剖析：
| 机制创新领域 | 传统 IPO 规则与常态 | SpaceX 开创的万亿级超级 IPO 实践 |
|:--- |:--- |:--- |
| **锁定解禁机制** (Lockup Structure) | 通常为上市后 180 天一刀切全部解禁，极易引发老股东集中踩踏套现。 | 采用**分阶段渐进式解禁**（Staged release），将股份释放与特定时间窗口及公司关键业务里程碑严格绑定。 |
| **主要指数纳入** (Index Inclusion) | 要求新股经历数个季度至数年的二级市场严苛检验，防范早期剧烈波动。 | 获交易所与指数编制机构特批**提前破格纳入**（Early inclusion），允许被动巨量资金池迅速提供流动性承接。 |
| **制度逻辑演变** | 过去新上市公司多为年轻、未经周期考验、收入及利润体量较小的小体量成长股。 | 巨头在私募市场已完成巨额体量积累（营收超数百亿、甚至具备盈利能力），其系统重要性倒逼指数修改陈旧规矩。 |
Brad 坦承，市场对提前纳入指数确实存在合理担忧。统计数据显示，传统 IPO 在上市后 6 个月内的波峰到波谷（Peak-to-trough）平均回撤幅度高达 50%；若在投机情绪高涨的峰值将新股塞入指数，被动追踪者将直接硬抗 30% 以上的惨烈回撤。但 SpaceX 的独特之处在于其规模实在过于庞大且举足轻重，迫使各大交易所和指数编制机构最终打破惯例、修改规则，这也为后续 Anthropic 和 OpenAI 的万亿级上市铺平了制度道路。
### [10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=600s) 万亿美元估值狂想：OpenAI 与 Anthropic 的 IPO 路线图与复合增长预期
Brad Gerstner 指出，SpaceX 的成功上市为 Anthropic 这类超级独角兽提供了一份绝佳的参照蓝本。无论从盈利前景、模型迭代速度还是营收增长势头来看，像 Altimeter、Fidelity（富达投资）以及 T. Rowe Price 这样的顶级机构投资者，都会毫不犹豫地争抢进入 Anthropic 的领投名册（top of the book），“所有人都会疯狂蜂拥而上（pig piling in）”。市场此前传言 Anthropic 计划在年内启动上市；尽管众所周知 Anthropic 在营收增长曲线上曾一度反超 OpenAI，但 OpenAI 如今也重新找回了势头与底气（swagger and mojo）——不仅近期密集发布新模型，市场上更有强烈传闻称 GPT-6 将在未来 30 天内亮相。据 Twitter（现 X）上的最新流言，OpenAI 今年的营收规模预计可达 700 亿美元左右；虽然略低于外界传闻中 Anthropic 超过 1000 亿美元的惊人数字，但这仍是 SpaceX 现有营收的两倍之多。基于这种爆炸性增长，作为全球最顶尖的两大前沿模型实验室（frontier premier labs），两家公司完全有底气以超过 1 万亿美元的估值叩开公开市场的大门。
在探讨谁将先拔头筹时，Brad 认为两家公司之间其实并不存在一场你死我活的“抢跑竞赛”，各自会在合适的时间点登场。不过，OpenAI 由于牵涉复杂的非营利转向营利性实体的企业重组架构（corporate restructuring），流程上显然更加繁琐，因此如果 Anthropic 先于 OpenAI 上市并不会令人意外。当 Jason Calacanis 追问 Altimeter 是否会在高达 3 万亿美元的估值水平下大举买入时，Brad 坦言，任何一旦跨过 1 万亿美元估值门槛的标的，靠短期炒作“一夜暴富”（get rich quick schemes）的时代就彻底终结了。正如他们此前在流动性论坛和播客中反复强调的，SpaceX 等超级独角兽即使上市，初期流通盘（float）通常也仅有 4% 到 5%，短期股价难免上下波动，普通散户不必在上市当天恐慌性追高，市场在一年后很可能仍会在合理区间提供上车机会。但从长期看，这些公司依然具备极强的复合增长潜力（compounders）：
```mermaid
flowchart TD
accTitle: 超级 AI 实验室 IPO 驱动逻辑与资本预期
accDescr: 展示 OpenAI 与 Anthropic 在营收规模、万亿市值与长期复合增长方面的核心驱动要素
subgraph Labs["超级前沿实验室 (Frontier Labs)"]
OA["OpenAI: 传今年营收 ~$70B<br/>GPT-6 预期 / 重组复杂"]
ANT["Anthropic: 传今年营收 >$100B<br/>企业级渗透 / SpaceX 式上市蓝本"]
end
subgraph Valuation["资本市场估值逻辑"]
TRILLION["万亿美元级 IPO 上市<br/>机构投资者排队抢单 (Top of the book)"]
FLOAT["低流通盘 (~4%-5%)<br/>摒弃短期暴富幻想"]
end
subgraph Growth["长期回报预期"]
CAGR["年化 30%+ 复合营收增长<br/>在百亿甚至千亿高基数上的超常规扩张"]
end
OA --> TRILLION
ANT --> TRILLION
TRILLION --> FLOAT
FLOAT --> CAGR
```
- Brad 猛烈抨击了美国现行的合格投资者法案（accredited investor laws），直言该制度荒唐透顶（insane），剥夺了广大普通民众在早期分享这些顶尖创新红利的机会。
- 估值超万亿美元后，IPO 当天出现 50% 到 100% 的暴力持续拉升反倒意味着发行定价出现了严重失误；投资者应当着眼于其以超过 30% 的年化复合增长率持续创造长期价值——尤其是在年营收已经高达数十亿乃至数百亿美元的庞大基数上，维持 30% 以上的高速扩张在公开市场极为罕见。
---
### [13:30](https://www.youtube.com/watch?v=PHL1j2ti420&t=810s) 商业定位分化与企业端 ROI 隐忧：消费级与企业级的角力
Jason 抛出了一个核心疑问：一两年后公众与资本市场会如何看待这两大巨头的竞争版图？在此之前，ChatGPT 牢牢占据了消费级市场的心智认同（brand recognition），成了普通人做作业、父母找家用电器维修方案的代名词；而 OpenAI 随后似乎被 Sora、迪士尼合作项目以及与 Jony Ive 联手打造硬件设备等庞杂的消费端企图分散了精力。反观 Claude，则精准卡位企业级市场并大获成功，使得 OpenAI 恍然大悟后才掉头加速杀向企业服务。这是否意味着 OpenAI 犯下了将当年 Google 的高增长黄金生态位拱手让给 Anthropic，自己反而沦落为当年 Yahoo 局面的战略失误？
Chamath Palihapitiya 则给出了截然不同、极具冷峻视角的分析。他指出，企业端业务看似光鲜亮丽，但其根本死穴在于：买单的企业客户迟早必须看到真金白银的投资回报率（ROI）。Chamath 分享了他用 Anthropic 最新发布的模型做的一项实证压力测试。他首先提问“自 2024 年以来，AI 到底为标普 500 指数的每股收益（EPS）增长贡献了多少提振？”，模型最初给出的回答是惊人的 50%。但 Chamath 仔细拆解底层数据后发现，该数据将英伟达（NVIDIA）等芯片巨头出售硬件赚取的暴利全数计入了其中。当他剔除“美股七巨头”，追问其余“标普 493 家企业”真实的 EPS 增长时，答案瞬间骤降至 9%；进一步剥离通胀带来的价格传导能力（定价权）以及大约 3% 的股票回购影响后，所有公开数据能支撑的“AI 实际带来的 ROI 提振仅在 0% 到 2% 之间徘徊”。
Chamath 警告称，企业级市场的繁荣在目前阶段可能极具脆弱性（brittle）。当像 Brad 或 Gavin 这样的敏锐投资人开始逼问各家跨国企业的 CEO 与 CTO “你们投在 AI 上的钱到底拉动了多少 EPS”时，一旦企业拿不出令人信服的数据且自身缺乏持续提价的能力，企业采购预算随时可能出现断崖式收缩；相比之下，消费级市场虽然单客价值低，但坐拥数千万至上亿用户（相差整整两个数量级），反而成了避风港（safe harbor），完美规避了严苛的 ROI 财务审查。
---
### [17:15](https://www.youtube.com/watch?v=PHL1j2ti420&t=1035s) 从极客刷 Token 到业务落地：Uber 的 Agent 战队与企业 ROI 救赎
针对企业端投入产出比无法交账的尖锐矛盾，Jason 指出，近几周播客讨论的焦点正是企业在 Token 上动辄每年砸下数百万美元，且账单还在翻倍暴增，如果无法交出超越无风险收益率（risk-free rate of return）的回报，管理层必将面临愤怒股东的集体问责。好在这个行业的核心决策层——CTO、CEO 与顶级资本配置者们，已经开始在 X 平台上正面迎击这个质疑。Jason 特别引用了 Uber 首席技术官 Praveen 近期的长篇复盘：Uber 曾经在第一季度就因为内部狂欢式的使用而耗尽了全季度的 Token 预算，但他们并未就此止步，而是探索出了一套极具借鉴意义的企业级落地体系。
```mermaid
flowchart LR
accTitle: Uber 内部 Agentic 落地架构与部门赋能
accDescr: 展示 Uber 从研发端工具扩展到法务、运营与采购等核心部门的 Agentic Pods 协作机制
subgraph Eng["工程与研发底座"]
CODING["99% 工程师使用 AI 工具 / Vibe Coding"]
AGENTS["70%+ Pull Requests 由 Agent 生成"]
SKILLS["沉淀 2,500 个 Agentic Skills"]
end
subgraph Pods["跨部门赋能: Agentic Pods"]
FDE["前向部署工程师 (Forward Deployed Engineers)"]
SYS["业务部门系统架构专家 (Systems Thinking)"]
FDE --- SYS
end
subgraph Biz["赋能核心业务中后台"]
DEPT["法务 / 运营 / 市场 / 客服 / HR / 采购"]
EPS["降本增效，寻求真实的 EPS 提振"]
end
Eng --> Pods
Pods --> Biz
Biz --> EPS
```
Praveen 披露，目前 Uber 内部 99% 的工程师都在日常工作中深度使用 AI 编程辅助工具（vibe coding），超过 70% 的 Pull Request（代码合并请求）由本地或云端 Agent 主导完成，团队累计构建了多达 2500 项专属 Agent 技能（Agentic skills）。更为关键的一步是，Uber 将 Agentic AI 从纯粹的研发部门溢出，渗透到了法务、运营、市场、客户支持、人力资源（HR）以及供应链采购等全业务链条。Uber 打造了所谓的“Agent 战斗小组”（agentic pods），其实质就是将前向部署工程师（forward deployed engineers）直接派驻到各个职能部门，与深谙底层流程与系统性思维（systems thinking）的部门负责人并肩作战。
Jason 评价道，这种打法让 AI 脱离了盲目“拉老虎机式刷 Token（hitting the slot machine of token maxing）”的初级阶段，告别了各部门员工随意用 Claude Code 或 Perplexity 自嗨的状态，而是上升为由资深系统架构师主导的自顶向下的战略重塑。Chamath 顺势调侃称，既然 Uber 是由 Dara Khosrowshahi 这台“超级运营机器”掌控的技术企业，他们就应该在财报中单独开辟一项，白纸黑字地披露究竟有多少 EPS 增长归功于 AI；Brad 也随即准备就企业如何跨越这一阶段展开深入阐述。
- Uber 的核心突破在于跨越了开发者“单兵自嗨”的代码生成，将工程能力封装进各职能部门的核心系统流程中，真正开始追求成本的大幅下挫与效能的硬性提升。
- 这种前向部署工程团队与部门业务专家的融合模式，被在场嘉宾一致认为是传统企业在面临 AI 资本开支质疑时，自证实际 ROI 的唯一可行路径。
### [20:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1200s) 史无前例的万亿级 TAM：当机器开始制造机器
关于企业在人工智能上的大额资本开支究竟是不是泡沫，Brad Gerstner 指出，虽然不可否认目前有大量资金花在“实验性项目”上，短期内看不到直接的 ROI（投资回报率），但整个行业仍处于极其早期的阶段，早期到“根本没人在意当下的浪费”。原因在于，AI 面对的潜在市场规模（TAM）是这颗星球上的每一家小型、中型以及大型企业。以往企业级软件的收入往往高度集中在少数几个巨头客户手中，而如今是全球数以百万计的客户每天在独立做出经济理性的决策。例如 Uber 的 Praveen 等技术领袖虽然在积极寻求降本，但企业削减成本正是为了省下预算去购买更多 Token。
更关键的转折点在于，底层智能正在突破临界点，从单纯的“成本削减工具”转变为“收入驱动引擎”。无论是在生命科学突破还是产品创新领域，企业即便想脱离 AI 也已不可能。Brad 举了英伟达 CEO 黄仁勋（Jensen Huang）的例子：黄仁勋多次公开强调，英伟达下一代芯片的架构设计工作已经全面交由 AI 完成，“机器正在制造机器”（The machine is building the machine）。在前沿技术竞争中，哪怕只有极其微弱的智力优势都是生死攸关的，黄仁勋绝不可能退而求其次去选用二流模型。
Brad 用 Snowflake 的历史进行了类比：Snowflake 早期也经历过客户疯狂做底层查询优化、单次查询成本大跌的阶段，但这不仅没有打垮 Snowflake，其营收反而因为渗透进更多企业业务场景而持续爆发式增长。在此基础上，Brad 提出了一个极具冲击力的预测：如果前沿模型实验室今年年底能达到 1000 亿美元的年化收入跑道，明年完全有可能再翻 3 到 5 倍。过去硅谷风险投资人看到一家公司从 1 亿美元增长到 3 亿美元就会惊掉下巴，而现在大家讨论的竟是从 1000 亿美元跨越到 3000 亿甚至 5000 亿美元，单年产生 2000 亿到 4000 亿美元的净增量收入。在整部人类商业史上，从未出现过这种体量的扩张速度。
- **史无前例的收入跳跃**：从 100B 美元向 300B–500B 美元迈进，带来单年数千亿美元的纯增量。
- **Snowflake 优化悖论**：底层查询与推理优化并不会终结营收增长，反而会加速用例向全业务渗透。
- **不可替代的顶级算力需求**：在前沿设计领域（如英伟达用 AI 设计新芯片），细微的模型智商差距直接决定产品代际优势。
---
### [22:38](https://www.youtube.com/watch?v=PHL1j2ti420&t=1358s) “按需智能”的全员渗透与自下而上的购买浪潮
Jason Calacanis 从技术采用形态的角度剖析了这种爆发的根源。他将 AI 的普及与历史上的企业级软件进行了鲜明对比：过去推出一款划时代的软件（例如 Excel），其价值通常只会沉淀到公司的某一个或两个垂直部门。Excel 面世时，财务会计部门兴奋狂欢，但人力资源或市场部门并不会因此改变日常工作流，价值需要漫长的时间才能向下渗透。然而“按需智能”（Intelligence on Demand）的颠覆性在于，它从第一天起就击中了组织内的每一个岗位，全员都在同时试用并尝试落地。
这种全员试用迅速引发了底层的乘数效应：一家千人规模的企业，员工最初可能每人每月在各类模型订阅上花费 200 美元；几个月后用量翻倍到 400 美元，甚至一年达到 5000 美元。若以科技公司员工平均 10 万至 15 万美元的年薪计算，这笔开销仅相当于员工工资的 3% 到 5%。如果一个工具能让员工的综合生产效率提升 3 到 5 倍，企业为之买单就是显而易见的最优解。
更为凶猛的是其“自下而上”（Bottom-up）的传播特性。员工掏出自己的企业信用卡，每个月刷 20 美元使用 ChatGPT 或 Claude，没有哪个 CTO 或 CIO 会为此设卡阻拦。当这种自下而上的低门槛工具在极短时间内砸向企业组织内的所有人时，所形成的营收陡峭斜率彻底超出了传统软件分析师的认知框架。几乎没有哪个部门敢断言自己完全不需要这种即取即用的智力支持，除非受到极严格的合规与数据安全监管限制。
```mermaid
flowchart TD
accTitle: 自下而上的 AI 渗透与投资回报逻辑
accDescr: 展示 AI 从员工个人信用卡采用逐步渗透并以极低薪资占比换取数倍人效提升的闭环。
subgraph Adoption ["自下而上的采用通道"]
A["个人员工刷企业卡<br/>20 美元/月试用"] --> B["全员跨部门铺开<br/>（HR、市场、工程、运营）"]
B --> C["单人月 Token 消耗<br/>攀升至 200 - 400 美元"]
end
subgraph Financials ["企业 ROI 算账逻辑"]
C --> D["年化工具成本<br/>约 2,400 - 5,000 美元"]
D --> E["仅占 10-15 万美元年薪的<br/>3% - 5% 增量预算"]
E --> F["换取 3 - 5 倍的人效与产能释放<br/>带来企业采购正循环"]
end
```
- **全面覆盖而非部门孤岛**：不同于 ERP 或 Excel 等专用工具，大语言模型是通用认知接口，横扫所有职能岗。
- **微小的薪资增量成本**：人均几百美元的月度消耗在 10 万+ 美元薪资结构下仅占微小比例，极易通过管理层 ROI 审核。
- **信用卡渗透路径**：无须漫长的企业销售（Enterprise Sales）周期，完全依托员工日常工作流的自发拉动。
---
### [25:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1500s) Token 成本骤降 95% 与自主 Agent 的质变
在探讨数据安全合规（如金融数据、HIPAA 医疗隐私、敏感 HR 档案）尚无法随意接入前沿模型的同时，Jason 分享了自己近期在自动化 Agent 领域的亲身实践。他谈到自己沉迷于开源工具，深入体验了 Nous Research 开发的 Hermes Agent，并通过 OpenRouter 平台获取各家 API Key，还在去中心化 AI 网络 Bittensor（代币为 TAO）的子网上接入了便宜至极的 GLM 5.2 等模型。
这一系列技术栈让 Jason 体验到了模型推理成本暴跌 95% 的极致冲击。他指出，随着开源生态的进化、类似 TAO 这样的分布式算力网络兴起，以及 Groq 和 Cerebras 等新型推理专用芯片的落地，推理成本正在被多维度同时击穿。而成本暴跌带来的不是“省钱”，而是“使用频次与任务架构的彻底重构”：
- **频次升维**：自动化 Agent 从过去的每天执行一次定时任务（Cron job），直接升级为每小时轮询一次。
- **架构拆解**：原先让一个 Agent 负责所有事情，降本后被拆解成 3 个独立的专业 Agent，每小时各司其职并发执行。
- **全自主运转**：清晨醒来时，后台系统已经在夜间无声无息地完成了 14 项复杂的深度研究与数据检索。
Jason 举例说明，他搭建了一个自动化趋势追踪 Agent，将所有的《All-In Podcast》和《This Week in Startups》节目录音与行业最新动态输入其中，Agent 每小时自动扫描全网并整理出当前科技界最核心的 3 到 4 个最新趋势。这种由廉价 Token 驱动的工作模式彻底改变了人类处理信息和决策的节奏。
- **基础设施降本三箭头**：开源软件架构优化、TAO 等去中心化推理子网、Groq/Cerebras 推理 ASIC 芯片突破。
- **任务执行模式跃迁**：从“天级别查询”转向“小时级持续巡检”，单体 Agent 进化为并发多 Agent 协同。
---
### [27:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1620s) 杰文斯悖论、前沿钱包份额与 iPhone 式“够用就好”时刻
Brad Gerstner 接过话题指出，在过去两年半的时间里，每一年的 Token 价格都以约 90% 的幅度断崖式下跌，这正是典型的“杰文斯悖论”（Jevons Paradox）——某种资源的利用效率越高、成本越便宜，人类反而会成倍甚至数量级式地消耗更多这种资源。
回顾过去 18 个月的技术路线之争，自 DeepSeek 横空出世引发美股科技板块单日暴跌 40% 以来，整个硅谷都在激烈争论：许多人宣称前沿闭源实验室死期将至，开源大模型正在迅速抹平智商代差，而多模型路由（Model Routing）能极其轻松地把任务甩给廉价开源模型。然而 Brad 指出，虽然自己与 Bill Gurley 一样是开源的坚定支持者，但现实数据却狠狠扇了看空前沿模型的人一耳光。正如风险投资人 Jesse Zhang 本周发布的图表所示，并得到了 Palo Alto Networks 首席执行官 Nikesh Arora 等人的讨论验证：尽管廉价开源模型在绝对 Token 消耗数量上的份额确实在激增，但前沿闭源实验室（OpenAI、Anthropic）在客户实际付费总额（Share of Wallet）中的占比不仅没有下降，反而持续攀升。市场普遍以为“90% 性能足够好且价格便宜 90%”的模型会杀死 Anthropic 和 OpenAI，但真实商业世界并没有按照这个剧本走。
对于这一现象，Chamath Palihapitiya 给出了针锋相对的反思。他认为前沿模型的高份额只是智能手机早期的阶段性产物。在 iPhone 刚问世的前几年，由于每一代提升巨大，所有人都会狂热追新，觉得付出高昂溢价是理所当然的；但随着技术趋于成熟，总会出现一个转折点——消费者和企业开始觉得手里的老款设备“已经足够好”（Good enough），根本看不出代际升级的实质差异，换机动力就此瓦解。大语言模型同样难逃这个宿命。更何况现在的顶级前沿模型（例如 GPT-5 / Fable 5 级别的前沿版本）存在严重的“过度对齐与阉割”（nerfed）问题，用户在做日常调研时经常遭遇模型拒绝回答的尴尬局面。Chamath 认为，每个用户和企业早晚都会到达属于自己的拐点：只要一个模型能够给出合理合规的答案让人推进工作，底层到底是谁的模型、智商是否绝对领先，根本无足轻重。
- **杰文斯悖论的应验**：每年 Token 价格下跌 90%，但全社会总计算量与采购开支却呈指数级膨胀。
- **钱包份额（Share of Wallet）的割裂**：开源与廉价模型分食了海量长尾 Token 数量，但高净值商业预算依然高度集中在前沿实验室。
- **“iPhone 拐点”反思**：Chamath 警示，一旦模型能力满足基础需求且前沿模型受困于过度对齐与安全阉割，企业势必会滑向“够用即可”的实用主义。
### [30:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1800s) 主权 AI 浪潮与企业的成本削减算盘
Chamath 谈到了他近期与 Salesforce 首席执行官 Marc Benioff、NVIDIA 首席执行官黄仁勋（Jensen Huang）以及微软总裁 Brad Smith 一同参加联合国委员会（UN Commission）的见闻。他指出，放眼全球，没有任何一个国家不在紧锣密鼓地制定自己的“主权 AI 战略”（Sovereign AI strategy），而且这些国家绝不相信依赖受制于人的美国闭源商业模型会是长久之计。当前全球 AI 面临三重趋势：一是人类在地理分布上的渗透率，普及使用对所有人都是巨大的上行机遇；二是企业内部从早期的尝鲜实验向高频、可重复的常规化生产应用过渡；三是这套体系必须全面嵌入到各个国家与社会现有的监管基础设施之中。
```mermaid
flowchart TD
accTitle: Sovereign AI Stack Adoption vs Enterprise Cost Pressures
accDescr: Diagram showing how national sovereignty concerns and enterprise cost pressures drive the shift from closed American frontier models to open-weight models and hybrid routing.
subgraph Sovereignty["国家主权驱动 (Sovereign AI)"]
Gov["各国政府 / 主权机构"]
Gov -->|"规避技术受制风险"| OpenStack["本土端到端技术栈\n(开源模型 / NVIDIA 架构)"]
OpenStack -->|"95%-99% 性能已'足够好'"| NationalControl["自主可控基础设施"]
end
subgraph Enterprise["企业内部驱动 (Enterprise ROI)"]
CFO["CFO 利润与 ROI 施压"]
CTO["CTO / 工程团队落地实施"]
CFO -->|"财报暴跌 / 削减预算"| CTO
CTO --> Router{"动态模型路由 (Router)"}
Router -->|"高难度核心逻辑"| Frontier["前沿闭源模型\n(Anthropic Claude / OpenAI)"]
Router -->|"低阶常规代码与审查"| OpenWeight["低成本开源 / 开放权重\n(Kimi 2 / GLM-5.2 等)"]
end
```
这种全球博弈的结局必然是选择的极端多元化。许多国家完全不愿将国家命脉置于外部的技术风险之下，因此极其愿意砸重金在本土搭建一套从头到脚（soup to nuts）完全自主的技术栈。他们更倾向于以类似 NVIDIA 支持的开源模型为基础，在国门之内为本国企业和民众服务。哪怕开源模型的表现只有顶尖模型的 95% 或 99%，许多国家也完全会主张这“已经足够好”。而在商业公司层面，如果企业缺乏持续爆发的盈利增长，就无法长期承受高昂的模型调用支出，最终只能走向漫长痛苦的成本削减。Chamath 提到，除了少数极具决断力的公司（正如 Brad 曾写公开信向扎克伯格施压削减开支，Meta 才最终动手），绝大多数公司往往任由成本问题恶化。一旦财报遭遇重创，企业势必会寻找替代方案来转移成本；如果市场上出现“百事可乐与可口可乐”式的竞争，而百事的成本只有可口可乐的千分之一，资本市场和承销商就绝不能无视这一巨大的潜在风险。
---
### [32:15](https://www.youtube.com/watch?v=PHL1j2ti420&t=1935s) 开源实施门槛与 DoorDash CTO 的混合路由实践
Jason 随即从实际操作层面补充了反对意见：开源模型的部署难度极高。相比之下，在企业内部申请采购并直接调取 Claude 的 API 几乎是开箱即用。Jason 以自身为例，即便他技术功底扎实，为了配置好 Bittensor 网络和 OpenRouter 动态路由，依然折腾了数个小时——让系统在诸如智谱 GLM-5.2 等模型间自动路由，如果失败再无缝回退（fallback）至 Claude 的特定版本。普通企业根本不具备这种工程能力和技术团队。
但顶尖科技企业已经在迅速突破这一门槛。DoorDash 首席技术官 Andy Fang（联合创始人兼 CTO）在 X 平台上公开展示了其实践方案，正面回应了本播客此前讨论的焦点：
- **基准测试保证质量**：DoorDash 建立了内部代码评测基准，在保证代码审查质量丝毫不下降的前提下，大胆引入开源与开放权重（open-weight）模型。
- **高低分工协同**：将极其复杂的难题交给 Anthropic 的前沿旗舰模型处理，而将低层级的常规代码审查工作直接下放给月之暗面（Moonshot AI）的 Kimi 2。
- **CFO 与 CTO 的权力传导**：CFO 要求控制成本、兑现 ROI，技术部门便通过精细的模型分流实现降本增效。
Chamath 强调，行业现在必须前瞻性地预测“一旦企业财报不及预期（earnings miss）会发生什么”。在遭遇外部经济冲击或盈利滑坡时，管理层往往很难痛下决心直接裁员，最容易被大砍一刀的就是各项外部支出与技术外包成本。如果高昂的模型调用在短时间内未能证明其真实的投资回报率，这种采购风险将随着企业规模膨胀而急剧放大。
---
### [34:45](https://www.youtube.com/watch?v=PHL1j2ti420&t=2085s) 扎克伯格挑起价格战与前沿代币的不可替代性
Brad Gerstner 指出，全球主权技术栈与前沿智能并非非此即彼，未来的代币生态必然是开源与前沿闭源并存。事实上，大量日常代币（tokens）已经在流向各家实验室推出的轻量级、低延迟或滞后代币（如 OpenAI 和 Anthropic 的较小版本，以及 SpaceX/xAI 和 Google Gemini 的最新进展）。Meta 的动作尤其引人注目：扎克伯格在大学时代的老推特账号 `@finkd` 上异常活跃，发布了全新的 New Spark 1.1 模型，并推出 Meta Model API。扎克伯格公开宣称，新模型具备顶尖的 Agent 代理与代码编写能力，但价格仅为前沿闭源对手的百分之一，直接用极低成本挑起了激烈的价格战。
| 任务场景 | 代表性模型选择 | 代币消耗量与成本逻辑 | 失败与容错代价 |
|:--- |:--- |:--- |:--- |
| **基础文本处理**<br>（如文档摘要、日常格式转换） | 开源模型 / 滞后轻量模型<br>（Llama / Spark / Kimi 等） | 约 20,000 Cheap Tokens<br>（边际推理成本极低） | **极低**：内容若有微小瑕疵，人工快速扫一眼即可修正，完全可用低价换取性价比。 |
| **高阶复杂工程**<br>（如替代高级软件工程师 2 小时工作） | 顶尖前沿模型<br>（Anthropic Claude / OpenAI 旗舰） | 约 2,000,000 Expensive Tokens<br>（代币单价较高） | **极高**：长链路推理只要在早期、中期或尾声崩溃一次，之前烧掉的数百万代币与算力时间全数报废。 |
| **专业咨询替代**<br>（如替代时薪 $200 的专业顾问） | 绝对可靠的前沿模型<br>（Bulletproof 级前沿智能） | 约 $15 前沿成本 vs $3 廉价成本<br>（相较于 $200 账面价值差异极小） | **无法承受**：相对于收取的 $200 时薪，多花 $12 购买绝对靠谱的准确性完全微不足道，企业绝不会为了省 $12 冒险交付残次品。 |
Brad 强调，市场上所谓“开源模型达到顶尖模型 95% 性能就足够、客户将不再为前沿智能付费”的论调，在实际商业数据中根本站不住脚。Anthropic 和 OpenAI 迅猛爆发的营收增长，证明数以百万计的独立付费客户每天都在用真金白银投票。只要大模型是用来取代高薪知识工人的复杂工作，前沿模型带来的无瑕疵可靠性（bulletproof）就能让微小的推理成本差距显得无关紧要。商品化代币与前沿代币之间的智力断层短期内绝不会闭合。
---
### [38:40](https://www.youtube.com/watch?v=PHL1j2ti420&t=2320s) 全球主权 AI 实例与 David Sacks 的加入
Jason 为主权 AI 的具体落地补充了更多国际现实：
- **阿联酋（UAE）**：阿布扎比技术创新研究院（TII）持续主导并开源 Falcon（猎鹰）系列大模型。
- **沙特阿拉伯**：成立并重金支持 Humain，专注研发服务阿拉伯世界本土语境的专用阿拉伯语大语言模型。
- **日本**：本周宣布由政府及行业财团联合出资 60 亿美元，组建名为“Neo Terra”（NEOTRA）的产业财团，旨在避开纯文本大模型红海，直接跨越式布局物理 AI（Physical AI），即具身机器人与自动化实体系统。
在讨论进行到高潮之际，主持人 Jason 欢迎第四位合伙人 David Sacks（“Saxxy Poo”）正式切入对话，准备将关于代币 ROI、价值沉淀归属（开源还是前沿闭源）以及硅谷顶级 CTO 群体在社交媒体上的大辩论推向更深层次的政策与商业剖析。
### [40:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=2400s) 企业的 Token 成本困境与混合路由架构的技术壁垒
随着企业各部门纷纷引入大模型，首席财务官（CFO）们开始拿着账单兴师问罪，要求技术团队证明高昂支出的投资回报率（ROI），并追问是否存在更便宜的替代方案。企业首席技术官（CTO）们迫切希望将 Token 消耗转移到成本更低的模型上，毕竟算力和 Token 费用正在坐火箭般飙升。除了成本考量，企业还要面对 Alex Karp 此前强调过的“AI 主权”问题——各大公司极度担忧将自身核心业务的“独门秘方”或超额收益（Alpha）拱手让给前沿实验室（Frontier Labs），因为这些实验室随时可能在未来变成直接的竞争对手。因此，在商业动机和数据安全上，企业绝对有动力推进模型多元化，逃离被闭源模型锁定的命运。
然而现实情况却是“心有余而力不足”（The spirit is willing, but the flesh is weak）。绝大多数普通企业根本没有足够的技术能力实现这一设想。在科技圈内，像 Coinbase 和 DoorDash 这样拥有顶级工程能力的头部公司，确实成功自研了 Token 路由系统和中间件层，能够巧妙地将高难度任务分发给前沿大模型，而将日常机械的基础任务下放给普通模型。但对于普通传统企业而言，搭建并维护这套复杂的混合架构犹如天方夜谭。
这种技术能力的断层直接反映在市场数据上：尽管大家都在高呼拥抱开源，但闭源模型的钱包份额（Share of wallet）反而持续扩大。数据显示，开源模型在企业支出中的占比从去年的 19% 骤降至当下的 11%。这并不代表开源模型的使用量在减少——事实上，闭源和开源的实际调用量都在疯狂暴涨，而且开源模型只需支付底层算力费用，无需向模型实验室交过路费，因而很难仅凭美元支出衡量实际规模。但不可否认的是，“闭源模型大势已去”的论调完全被现实证伪。正如 Brad 所言，前沿实验室的收入正经历几何级暴增。对广大开发者和企业而言，直接调用顶级实验室打磨完美的开箱即用产品，依然是阻力最小、体验最极致的首选方案。
```mermaid
flowchart TD
accTitle: 混合 Token 路由与成本分流架构
accDescr: 展示企业如何通过中间件将请求分发至前沿模型或开源低成本模型
subgraph InputStage["客户端交互"]
Req["企业业务请求"]
end
subgraph RoutingLayer["中间件路由层 (Coinbase / DoorDash 方案)"]
Router{"任务难度评估"}
end
subgraph Models["模型处理层"]
Frontier["前沿闭源模型 (OpenAI / Claude)<br/>处理高复杂度、前沿探索任务"]
OpenSrc["开源 / 本地小模型<br/>处理结构化、成熟常规任务"]
end
Req --> Router
Router -->|"复杂前沿任务"| Frontier
Router -->|"低复杂度常规任务"| OpenSrc
```
- 开源模型在企业 AI 总支出中的份额从去年的 19% 下滑至今年的 11%。
- 仅有极少数具备顶尖自研能力的企业（如 Coinbase、DoorDash）成功落地了多模型动态分流的中间件系统。
- 闭源前沿实验室凭借极其打磨的商业化 API 与极高的便捷性，牢牢锁定了绝大多数企业的核心预算。
---
### [42:47](https://www.youtube.com/watch?v=PHL1j2ti420&t=2567s) 成熟用例 vs. 探索阶段：模型热插拔的上下文迁移难题
AI 智能客服平台 Decagon 创始人在一篇广为流传的深度博文中，道破了开源与闭源模型在应用周期中的本质差异：开源模型只有在“目标极其明确”的成熟用例下才能发挥最大威力。开源小模型虽然单价低廉，但前提是你必须拥有专有数据集并进行深度后训练（Post-training），清晰掌握输入输出的每一个环节。然而，在企业仍在摸索新可能性的“不成熟用例”（Immature use cases）阶段，业务人员根本不知道底层工作流该如何演化，此时最迫切需要的是通用智能最强大的前沿模型。正如讨论中所比喻的，成熟的客服场景根本不需要模型精通大学物理学，届时完全可以精简模型；但在前期的探索与试错期，顶尖前沿模型是无可替代的最佳抓手。
与此呼应，Palo Alto Networks 首席执行官 Nikesh Arora 也指出了企业界对于“模型可替代性”（Model Fungibility）或“无头化架构”（Headless）的强烈渴望。企业管理层最理想的构想是将所有大模型彻底商品化，变成可以随意即插即用、根据即时价格热插拔（Hot swap）的通用计算底座——哪个模型在这一秒能完成任务且报价最低，就将流量切换到哪里。
但这一美好设想遭遇了巨大的工程壁垒：记忆（Memory）、上下文（Context）和历史对话（History）该如何处理？目前全行业没有任何人能真正将上下文状态和对话记忆从特定的模型底层架构中完全抽象剥离出来。如果要实现无缝热插拔，就必须确保所有历史信息能够毫秒级无损迁移到替补的廉价模型中。这种中间件的技术难度极其苛刻，成了阻碍大模型彻底商品化的关键锁钥。
- **成熟业务场景**：边界清晰、输入输出格式固定，适合利用专有数据对开源小模型进行后训练，压低推理成本。
- **新兴业务探索**：工作流未定型，企业严重依赖通用智能最强的头部闭源模型进行能力摸底。
- **热插拔困境**：由于记忆、长上下文和交互历史无法跨架构无损解耦，企业很难在运行期自由替换底层模型提供商。
---
### [45:10](https://www.youtube.com/watch?v=PHL1j2ti420&t=2710s) Harness 外壳优化与非共识猜想：智能真的在收敛吗？
面对模型调用的高昂成本，处于行业前沿的极客和头部基础架构公司正在探索底层优化的新路径。Databricks 创始人 Ali Ghodsi 发现，即使完全不更换底层模型——不搞闭源换开源，也不在 OpenAI 与 Claude 之间跳跃——仅仅通过更换和重构外围的“Harness”（任务脚手架/应用运行外壳），针对记忆、技能编排等周边组件进行精简与调优，就能在相同模型下直接实现高达 2 倍的成本削减。Databricks 在 GLM 5 上测试发现，配合高效自研的 Harness，任务 Token 消耗硬生生被砍掉了一半。Jason 也深有共鸣地分享了自己的亲身经历：他此前编写了一个每小时自动运行一次用来挖掘前沿趋势的 Agent 监控脚本，后来命令系统进行专项优化，其 Token 消耗量惊人地降低了 80%。如今许多企业系统内部已经拥有非常精细的后台分析面板，能按小时、按具体任务和模型追踪消耗指标。诸如 Baseten、Fireworks 等推理云以及各大超大规模云服务商（Hyperscalers），都在全力以赴提供此类工具以推进模型可替代性。
随后，Brad Gerstner 抛出了一个直击灵魂且反主流共识的深刻思考：当前科技界普遍存在一个默认假设，即“大模型的智能正在收敛”（Convergence of intelligence），从各大跑分基准（Benchmark）来看，大家的分数确实越来越接近。然而，如果看实际的商业收入分布，市场非但没有收敛，反而呈现出极端的马太效应。
```mermaid
flowchart TD
accTitle: 递归超智能飞轮模型
accDescr: 展示更聪明模型获取更多收入进而拉大差距的正反馈循环
subgraph DataFlywheel["超智能递归飞轮 (Self-Recursive Flywheel)"]
A["顶尖通用智能优势"] --> B["获取最丰沛的商业收入"]
B --> C["采购更大规模的算力集群"]
C --> D["构建更具代差的下一代模型"]
D -->|"能力递归强化"| A
end
```
Brad 指出，非共识的现实很可能是：智能根本没有收敛，随着超级智能具备自我递归（Self-recursive）能力，领先者的优势不仅不会缩小，反而会无限放大。模型越聪明，获取的商业收入就越庞大；收入越庞大，就能采购更恐怖的算力；算力越强，训练出的下一代模型就更具代差。在未来两到三年，随着智能体（Agentic）开始处理极其复杂的端到端任务，前沿实验室与追赶者之间的距离很可能不是缩小，而是拉得更开。
这引发了生态位上初创企业的深深焦虑。Jason 提到他近期分别深度访谈了 Vibe-coding 明星产品 Lovable 的创始人 Anton（该公司在短短两三年内收入呈现出令人瞠目结舌的爆发式增长，年化收入从数百万直接冲向数亿美元体量）以及 ElevenLabs 的首席执行官 Mati。这些炙手可热的应用层巨头每年都要向前沿实验室进贡数千万美元的 API 调用费。当被单刀直入地问及“是否担心数据泄露、以及前沿大厂未来随时可能发布同类竞品把自己吃掉”时，这种寄生于前沿模型之上的生存焦虑表露无遗。
- **Harness 效率革命**：Databricks 证实，优化任务脚手架与记忆机制，能在完全不降级模型的前提下将推理成本直降 50%。
- **“智能收敛”假说受质疑**：Benchmark 表现的拉近掩盖了前沿实验室在长任务、自递归和工程落地上的真实代差。
- **前沿飞轮效应**：更强大的通用智能带来更高收入，支撑更大规模的算力扩展，使得领先优势可能长期处于发散而非收敛状态。
- **应用层生存悖论**：如 Lovable、ElevenLabs 等快速崛起的 AI 原生独角兽，一边向闭源实验室大额“纳贡”，一边时刻提防巨头下场自研类似功能形成降维打击。
### [50:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=3000s) 垂类专用模型崛起 vs 前沿双寡头垄断
讨论首先聚焦于头部应用公司是否会抛弃前沿模型实验室（Frontier Labs），转而自研私有专属模型。随着微调与垂直化技术的门槛每六个月就大幅下降一次，针对语音交互、代码构建等特定场景的“垂类专用模型”变得愈发容易搭建。行业中那些每年向 OpenAI 或 Anthropic 支付八位数甚至九位数 API 账单的巨头客户，都在积极谋求摆脱对前沿模型的依赖，实现模型层面的自主可控。
不过现场也出现了针锋相对的质疑：以做语音生成的 ElevenLabs 创始人 Mati 举例，在残酷的市场竞争中，面对前沿实验室供给的世界顶级语音能力，创业公司真敢为了省钱去冒险采用效果稍逊的自研便宜货吗？除非自研版本能做到更强，否则在尚未成熟的赛道上妥协就是自掘坟墓。对此，嘉宾以客服 AI 独角兽 Decagon 给出反证：Decagon 目前已有高达 90% 的调用量流向了开源模型，因为在明确具体的业务场景下，团队可以利用积累的大量私有数据和经验对开源底座进行深度后训练（post-training）。这一分歧最终归结为“业务成熟度”：在摸索与模糊阶段，企业必须依赖最强的前沿大模型；一旦业务闭环和输入输出完全确定，工作流便会向高性价比的定制化专用模型迁移。
与此同时，以 Brad 为代表的观点指出了当前大模型市场愈发明显的“双寡头垄断”（Duopoly）格局。仅按 Token 收入来看，除 Anthropic 与 OpenAI 外几乎没有其他玩家具有实质规模，且随着两家巨头处理的用例呈指数级爆发，其基于真实场景反馈的强化学习飞轮越转越快，与其他跟随者的差距正在被拉大。针对这一说法，另一方提醒道，不能单凭财报收入来断定胜负：企业在内部或 NeoCloud（如 Cerebras 等新型算力云）上私有化部署开源模型所产生的海量“暗 Token”（Dark Tokens）是完全免费且不计入软件厂商收入的，这部分价值直接被转嫁为了硬件基础设施与云厂商的算力消耗，并非不存在。
```mermaid
flowchart TD
accTitle: 大模型工作流选型与成熟度演进
accDescr: 展示企业从探索期使用前沿闭源模型到成熟期迁移至定制化开源模型的演进路径
subgraph Phase1 ["探索与早期阶段 (业务未定型)"]
F1["采用前沿闭源实验室模型\n(OpenAI / Anthropic)"] --> F2["依赖最强泛化能力快速验证 PMF"]
end
subgraph Phase2 ["成熟与定型阶段 (业务完全清晰)"]
F2 --> O1["沉淀专有业务数据与经验"]
O1 --> O2["在开源底座上进行后训练与定制"]
O2 --> O3["自建垂直模型/专属工作流\n(如 Decagon 90% 转开源)"]
end
```
- **Token 营收表现**：目前前沿模型市场已形成清晰的双寡头格局，Anthropic 与 OpenAI 的年度经常性收入（ARR）体量完全碾压同类实验室。
- **暗 Token 效应**：DoorDash 等企业在开源模型和专用硬件上跑的业务不体现为大模型公司的 ARR，而体现为英伟达、NeoClouds 及数据中心的算力托管支出。
---
### [54:29](https://www.youtube.com/watch?v=PHL1j2ti420&t=3269s) 中国或限制前沿模型出海：开源诱敌，闭源收网
路透社接连爆料的两篇重磅独家报道引发了全场热议。报道称中国官方正考虑限制其顶级 AI 模型的海外访问权限，两家监管机构已就此约谈了阿里巴巴、字节跳动以及智谱 AI（Zhipu AI，GLM-5.2 的开发者），探讨收紧顶尖闭源乃至开源大模型出海的可能。与此同时，任何大模型技术的外泄与窃取被上升至国家安全层面，Manus 团队从新加坡被召回国内亦被视作风向转变的信号。官方核心担忧之一在于所谓“Mythos”级别漏洞的利用，担心模型技术被外部用于网络攻击或损害自身利益。
针对“中国是否会彻底封锁开源模型”这一猜想，David Sacks 指出该报道存在夸大成分，更底层的商业规律是科技博弈中的经典周期：“落后时开源，赶超后闭源”。字节跳动对标 ChatGPT 的拳头模型从来都是闭源；阿里通义千问（Qwen）与智谱 GLM 在早期之所以坚定拥抱开源，是因为在落后于前沿基准时，闭源无法获取有意义的商业付费，开源反而能迅速拉拢开发者生态、换取实机调用量以及强化学习反馈。而一旦其评测分数追赶至世界梯队（例如 GLM-5.2 在某些任务逼近美国现有商业前沿），公司就会顺理成章地走向闭源以独吞商业价值——这与 Sam Altman 带领 OpenAI 从非营利/开源走向营利/闭源，以及早期 Meta 调整 Llama 开源边界的操作完全如出一辙。
刚从华盛顿特区交流归来的嘉宾进一步透露，当前白宫、财政部直至总统本人，跨党派最强烈的共识就是“不惜一切代价在 AI 上保持对中国的领先”。美国内部甚至掌握了充分痕迹，证实海外实验室的大模型输出中存在大量蒸馏自美国顶尖模型（如 Mythos 水印）的证据，美方后续势必会就模型蒸馏行为出台打击政策。大家一致认为，即便中国模型全面转向封闭，对美国生态的实际杀伤力也极为有限：美国本土包括英伟达及 Reflection 等开源力量随时可以承接需求，而开源生态的退潮本质上更像是中国企业从扩张期向商业变现期收网的常规手段，正如当年谷歌先用免费开源的 Android 席卷全球市场，随后强行捆绑 Google 搜索、Chrome 和 Google Drive 一样。
- **监管动向**：监管部门重点关注阿里、字节与智谱 AI 等头部梯队，旨在强化本土 AI 知识产权管控与防范地缘反制。
- **蒸馏争议**：美方监管层密切监控对前沿大模型的逆向蒸馏，部分海外模型输出被指直接带有特定前沿模型的隐形水印。
- **历史周期的重演**：科技巨头由弱变强普遍沿用“开源圈地获得飞轮，追平前沿转入闭源收割”的路径，Android 便是最典型的先例。
### [01:00:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=3600s) 中美 AI 竞赛的真正隐患：官僚监管与能源短缺
全场首先探讨了中美在人工智能领域的角逐。嘉宾们调侃道，如果中国也能诞生自己的“AI 末日论（Doomer）群体”，甚至出一个中国版的埃利泽·尤德科夫斯基（Eliezer Yudkowsky，知名 AI 风险论学者），不断推高当地的“P-Doom”（人类毁灭概率预估），让更多人对失业或技术失控惊慌失措并游说打压实验室，那将是美国赢得 AI 竞赛的“天赐良机”。言归正传，大伙一致认为，从总统到内阁高层，赢下 AI 竞赛的决心十分坚定，一年前的重磅 AI 政策演讲主题便是明确宣布美国进入 AI 竞赛。然而，真正的制度风险往往隐藏在联邦官僚机构的底层执行中：底层官僚可能在未充分理解产业逻辑的情况下盲目推行禁令，自以为在遏制对手，实则扼杀本土创新；同时，国会议员更容易受到激进末日论游说集团的政治施压，从而做出短视立法。
随后，讨论转向了更为严峻的硬性物理瓶颈——能源。即便软件算法不断演进、算力芯片持续迭代，真正的绊脚石却是电力供应。查马斯·帕利哈皮蒂亚（Chamath Palihapitiya）分享了其团队的一项惊人测算：到 2050 年，美国仅满足常规设备、电动汽车、家电与计算机的正常负荷增长，电力缺口就高达“三个加利福尼亚州的整体用电量”。此外，全球芯片代工核心台湾同样命悬一线，其运转极度依赖液化天然气（LNG），实际库存仅够维持两到三周；一旦台海面临封锁，电力将在短时间内断绝。无论中美还是台湾，AI 推理与算力扩展的终极决胜点已落在了电子供应上，美国必须在繁琐的监管阻力中加速审批核能、太阳能和电池储能建设。
- 能源缺口测算：到 2050 年，全美常规用电负载增长所面临的能源短缺折合相当于整整三个加州的电力规模。
- 台湾能源脆弱性：台湾芯片制造业赖以运转的液化天然气储备仅维持约 2 至 3 周，封锁风险对全球算力基础设施构成巨大威胁。
```mermaid
flowchart TD
accTitle: AI 产业发展的双重瓶颈模型
accDescr: 展现底层监管干预与物理能源短缺如何制约 AI 竞赛推进
subgraph Governance ["政策与监管维度"]
Top["总统及白宫高层: 坚定推进 AI 竞赛"]
Bureaucracy["底层官僚机构: 误判行业出台盲目禁令"]
Congress["国会: 屈从 Doomer 政治游说压力"]
Top -.-> Bureaucracy
Congress -.-> Bureaucracy
end
subgraph PhysicalLimits ["物理与基础设施瓶颈"]
Grid["美国电网承载力短缺 (缺口达3个加州)"]
TaiwanLNG["台湾能源储备极度脆弱 (LNG仅存2-3周)"]
EnergyGating["能源电子成为推理与算力的终极瓶颈"]
Grid --> EnergyGating
TaiwanLNG --> EnergyGating
end
Bureaucracy --> Impact["制约 AI 算力与模型研发"]
EnergyGating --> Impact
```
---
### [01:03:09](https://www.youtube.com/watch?v=PHL1j2ti420&t=3789s) “特朗普账户”重磅落地：白宫敲钟与全民资本家构想
话题随后转向重磅成果——由布拉德·格斯特纳（Brad Gerstner）历时四年全力推动的儿童投资账户（法案层面为《投资美国法案》Invest America Act，公开品牌为“特朗普账户”Trump Accounts）正式落地。布拉德刚从华盛顿特区归来，该应用一经推出便登顶全球 App Store 免费下载榜首。现场好友纷纷向他表达祝贺，回顾当年布拉德刚提出该构想时，大家甚至觉得他“疯了”、“纯属浪费时间”；然而该方案在去年通过预算和解程序被签署成法，并于今年 7 月 4 日正式上线，实现了数以百万计的账户创建与资金注入。
为庆祝这一历史性时刻，白宫椭圆形办公室上演了史无前例的盛况：纽约证券交易所（NYSE）与纳斯达克（NASDAQ）破天荒首次联合在椭圆形办公室举行敲钟仪式。现场汇集了 100 位知名企业 CEO 以及受助儿童家庭。总统在讲话中明确指出，这绝不仅仅是一个普通的福利派发项目，而是要“让每个孩子都成为资本家（making every child a capitalist）”。白宫甚至提出进一步构想：未来将为全美 5000 万至 7000 万名 18 岁以下未成年人自动开户，确保没有一个孩子被资本复利体系抛下。
布拉德详细拆解了该账户的核心运作机制：每个婴儿在出生获得社会安全号（SSN）的同时，即自动获得一个注资 1000 美元的私有投资账户。资金全额免管理费追踪标普 500 指数（S&P 500），终身免费持有；若辅以社会匹配捐赠以及家庭每周定投 10 美元，到孩子 18 岁成年时，账户本息将滚存至约 5 万美元。该笔资金在孩子 18 至 20 岁时可解锁用于高等教育，或直接转入罗斯个人退休账户（Roth IRA）继续享受复利红利。
- 上线 24 小时爆发力：应用上线首个 24 小时内，全美家长自发创建账户超 150 万个，直接充值存款超 10 亿美元。
- 便捷社交转账体验：每个儿童账户均配有专属二维码，亲朋好友无需繁琐手续，只需用手机双击 Apple Pay 扫描二维码，即可直接给孩子存入 25 或 50 美元。
---
### [01:06:04](https://www.youtube.com/watch?v=PHL1j2ti420&t=3964s) 亿级资金涌入与科技巨头慈善：Brad 的“破天荒”一亿美元
伴随产品上线，多位科技巨擘与大型企业带来了极其震撼的慈善认捐。戴尔科技创始人迈克尔·戴尔与苏珊·戴尔夫妇（Michael & Susan Dell）作为核心基石捐赠人，承诺提供超 10 亿美元资金，精准覆盖 2500 万名中低收入社区儿童；SpaceX 总裁兼首席运营官格温·肖特维尔（Gwynne Shotwell）紧随其后，捐出价值 3.5 亿美元的 SpaceX 私有股票，专项定向扶持低收入家庭未成年人；半导体巨头美光科技（Micron）则出资 2.5 亿美元设立雇主匹配激励，为每位员工子女匹配最高 1000 美元账户启动金。此外，社会各界的资助可以非常精准地通过“邮政编码（Zip code）与年龄”直接下沉到目标群体。
随后，现场好友爆出了本场最戏剧性的一幕：平时在私人扑克局里极其“抠门”、晚上 10 点后被要求加买 1 万美元筹码就大声抱怨、输掉 6000 美元便气得“怒退牌局（rage quit）”、去牌局最多只带一瓶红酒的布拉德·格斯特纳，这次竟然一口气自掏腰包捐出了整整 1 亿美元，专门全额资助他家乡印第安纳州的所有儿童。现场顿时爆发出一阵哄笑、鼓掌与高尔夫掌声，大家纷纷调侃这位“铁公鸡”创投大佬罕见地完成了从牌桌省钱到世纪慈善的华丽转身。
布拉德对此回应道，这个平台将彻底颠覆传统慈善运作模式，成为全美乃至人类历史上规模最大的点对点直达慈善平台（direct philanthropic platform）。他透露已向白宫立下军令状，目标是在系统运行的前 12 个月内撬动 100 亿美元的社会慈善捐赠。
- 关键捐款明细：
- 迈克尔与苏珊·戴尔夫妇：超 10 亿美元，定向覆盖全美 2500 万中低收入家庭孩子。
- 格温·肖特维尔（SpaceX 总裁）：捐出 3.5 亿美元价值的 SpaceX 股票。
- 美光科技（Micron）：投入 2.5 亿美元，每名员工子女最高提供 1000 美元雇主匹配金。
- 布拉德·格斯特纳：捐赠 1 亿美元，覆盖印第安纳州全境新生儿。
---
### [01:08:44](https://www.youtube.com/watch?v=PHL1j2ti420&t=4124s) 去中介化慈善与党派偏见的荒诞抵制
布拉德强调，该机制之所以能吸引数百亿级别资金，核心优势在于彻底消灭了传统慈善信托或基金会等“中间抽成中介（charitable middleman）”。过去坐拥百亿乃至千亿身家的顶级富豪，常常苦于缺乏高效、无损耗的大规模财富回馈渠道；而现在通过该底层系统，资金可以无缝穿透、毫发无损地直达全美儿童的个人账户，在孩子年满 18 岁前任何人都无法提前支取，最大化确保了资金安全与长周期指数复利。
然而，这一纯粹造福下一代财富积累的项目，却在高度撕裂的政治语境中遭遇了荒谬阻力。由于项目被冠以“特朗普账户（Trump Accounts）”，部分深陷“特朗普精神错乱综合征（TDS）”的极端党派人士在社交平台蓝天（Bluesky）及各大舆论场上发起了猛烈抵制，甚至声称拒绝为自己的孩子认领该账户。据测算，如果从小拥有这一启动资金并依托复利，当孩子年满 50 岁时账户规模有望成长至 1300 万美元的惊人数字。播客主理人们对此类意识形态极端化嗤之以鼻，直言“仅仅因为讨厌某个政治人物的名字，就亲手剥夺孩子未来可能拥有的 1300 万美元巨额财富，这种行为简直荒谬绝伦到了极致”。
### [01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) 特朗普账户与《投资美国法案》：全自动建户与两党博弈
讨论转向了全美热议的“特朗普账户”（Trump Accounts）。布拉德·格斯特纳（Brad Gerstner）澄清说，这项政策的法定授权法案其实是《投资美国法案》（Invest America Act），许多民主党人称其为“投资美国账户”，但官方目前的正式名称确实挂着特朗普的名字。面对部分选民和网络上“谁会相信特朗普账户”的质疑，布拉德指出，线下的实际情况完全不同，不同收入阶层、不同政治立场的父母都在积极注册，因为对父母而言，首要责任是为孩子积累储蓄、接通“美国梦”。布拉德还透露，在过去两年中，没有人比特朗普本人更关心如何让每一个孩子都拥有这个账户。在两人共进午餐时，特朗普一直在追问如何更快、更大规模地推行自动建户，坚决不希望有任何一个孩子因为父母打两份工太忙、或者因为对“特朗普”这个名字心存芥蒂而被落下。
杰森·卡拉卡尼斯（Jason Calacanis）随即逼问布拉德：“特朗普总统已经直接给你和财政部长下达了命令，要求利用 18 岁以下人口的社会安全号（SSN），在 90 天内直接自动为所有 5000 万到 7000 万孩子创建账户。布拉德，你会执行总统的指令，还是会违抗他？”布拉德回应表示，团队的目标确实是在接下来的 90 天内利用现有数据推动这 5000 万至 7000 万账户的设立，但这需要与财政部、白宫以及社保署等多方通力协作。杰森又追问伯尼·桑德斯（Bernie Sanders）和罗·康纳（Ro Khanna）等进步派民主党人是否会因为不想“把功劳拱手送给特朗普”而从中作梗。布拉德客观地表示，其实已经有不少民主党领袖站出来公开支持，包括参议员科里·布克（Cory Booker）、加州州长加文·纽森（Gavin Newsom）、马里兰州州长韦斯·摩尔（Wes Moore）以及宾夕法尼亚州参议员约翰·费特曼（John Fetterman）。
布拉德在 CNBC 上阐述了两种截然不同的美国发展观：一端是伯尼·桑德斯和佐兰·马姆达尼（Zohran Mamdani）式的路线，试图对大企业重税盘剥，将所有资金集中在华盛顿政客手中进行再分配，建立一种高度依附于联邦政府的模型；而另一端则是特朗普政府以及许多温和派民主党人所认同的理念——为每个美国孩子建立独立的私人账户并注入本金，让他们在经济上独立于政府，依靠自己的财商和复利去积累财富，从而更有可能完成高中学业、创办企业和购置房产。布拉德直言：“医治更多社会主义的解药，就是更多的资本主义。”
- **账户覆盖目标**：计划在未来 90 天内，通过跨部门数据整合，全自动为全美 5000 万至 7000 万名 18 岁以下未成年人开立账户。
- **全民持股跃升**：若全自动建户彻底落地，全美持有股票资产的人口比例将从目前的 50% 跃升至 70% 甚至 75%，使美国真正迈入“全民股东国家”（Equity Nation）。
- **两党合作阵营**：尽管法案以特朗普命名引发争议，但包括科里·布克、加文·纽森、韦斯·摩尔和约翰·费特曼在内的多位民主党重量级政要均已表态支持。
---
### [01:13:50](https://www.youtube.com/watch?v=PHL1j2ti420&t=4430s) CPA 推特热议的避税神器：从“慈善免费领钱”到顶级财富管理方案
大卫·萨克斯（David Sacks）对该计划给予了极高评价。他认为，在公众对亿万富翁的民粹主义愤怒日益高涨、普通人愈发觉得体制被操纵且自己无法分享国家繁荣成果的当下，这一举措是极佳的社会解药。不过，萨克斯敏锐地指出，外界对该账户的“慈善施舍”或“免费发钱”属性给予了过多关注，却忽视了它在专业财务界引发的轰动。在注册会计师推特圈（CPA Twitter）上，精明的财务规划师们早已炸开了锅，将其视为前所未有的顶级遗产规划与财富管理架构。
萨克斯将其与 401(k) 雇主匹配、健康储蓄账户（HSA）以及 Roth IRA 并列，称其为理性人绝对不可错过的“财务必选项”。他详细拆解了该账户令人惊叹的双重税收红利：
```mermaid
flowchart TD
accTitle: 特朗普账户（投资美国账户）资金流转与避税生命周期
accDescr: 展示未成年时期存入与免税复利、成年后转换为 Roth IRA 以及退休后免税提取的完整财务规划路径。
subgraph Phase1["阶段一：18 岁前免税积累期"]
A["每年最高 $5,000 父母/亲友捐赠"] --> C["特朗普账户<br>(Invest America Account)"]
B["每年最高 $2,500 雇主税前匹配/捐赠"] --> C
C --> D["全额免税复利投资<br>(指数基金/股票股权)"]
end
subgraph Phase2["阶段二：18 岁后转换与灵活运用"]
D -->|年满 18 岁| E["滚存至传统 IRA / Roth IRA"]
E --> F["CPA 避税妙计：<br>在子女大学或低收入期（0% 税率）转换为 Roth IRA"]
F --> G["账户资金达 20万-30万美元"]
end
subgraph Phase3["阶段三：终身免税复利与退休"]
G --> H["免税首套房首付 / 紧急医疗支出 / 创业基金"]
G --> I["持续免税复利至 60 岁退休<br>资产规模突破 100万+ 美元"]
end
```
根据法案设计，每年任何亲友都可以向 18 岁以下孩子的账户中免税赠予最高 5,000 美元；更诱人的是，雇主每年可以免税为员工子女的账户注入最高 2,500 美元。萨克斯建议，普通打工人即便要求雇主直接从自己的工资中扣除 2,500 美元打入孩子的特朗普账户都是血赚，因为这部分资金在雇佣双方两端全部免除个税与薪酬税。资金在孩子 18 岁之前享有完全的免税复利增长（Tax-Free Compounding），机制类似于传统的个人退休账户（IRA）。
更为精妙的操作在于成年后的转换阶段。萨克斯分享了推特上会计师提出的“神级避税操作”：当孩子年满 18 岁后，可以将该账户滚存（Rollover）为 Roth IRA。虽然从传统 IRA 转为 Roth IRA 时需要按当时税率缴纳税款，但聪明的家长完全可以等待孩子脱离受抚养人身份、进入大学或刚大学毕业且没有收入的时期（处于 0% 个税区间），此时进行转换几乎不需要缴纳任何税费。完成转换后，这笔在 18 岁时可能已累积至 20 万到 30 万美元的资金，将在 Roth IRA 中进行终生完全免税的复利投资。此外，根据 IRA 现行规则，这笔钱还可以免罚金用于购买首套房的首付款、支付突发医疗紧急费用或用于开启创业。萨克斯算了一笔账：如果在 18 岁时拥有 20 万至 30 万美元本金，只需按市场平均回报复利到 60 岁退休，最终资产规模将轻松突破 100 万美元以上。查马斯·帕里哈皮蒂亚（Chamath Palihapitiya）在一旁打趣道：“我们这简直是在批量制造‘裙带宝贝’（Nepo Babies），未来全美到处都是坐拥信托基金的富二代了！但这绝对比联邦社会保障（Social Security）靠谱得多。”
- **亲友免税额度**：每年每名 18 岁以下儿童最高可接受 5,000 美元的私人捐赠（父母、亲友均可参与）。
- **雇主税收优惠**：企业雇主每年可免税向员工子女账户注入最多 2,500 美元，双向抵税。
- **退休复利估算**：18 岁时滚存入 20 万至 30 万美元，放置在标普指数中复利至 60 岁，退休资产将毫无悬念地达到 100 万至数百万美元。
---
### [01:16:55](https://www.youtube.com/watch?v=PHL1j2ti420&t=4615s) 滚动复合至百万资产：科技巨头捐赠与未来一亿账户图景
就在大家深入讨论时，查马斯表示自己必须提前退席去给客户打企业软件销售电话（"Always Be Closing"），但他离开前给布拉德留下了一个激进的构想：既然布拉德预测人工智能的未来市场规模将达到数千亿甚至数万亿美元，他就应该去游说 OpenAI 和 Anthropic 等 AI 巨头，将这些初创公司的私人股权直接捐赠并按比例配发到每一个美国孩子的账户中。
查马斯离开后，大卫·萨克斯向布拉德抛出了执行层面的现实疑问：如果每名儿童每年的接收上限只有 5,000 美元，科技巨头数以百亿计的巨额股权或资金该如何注入并合规分配给孩子们？
布拉德披露，他确实已经在与 Anthropic 的达里奥·阿莫代伊（Dario Amodei）、OpenAI 的萨姆·奥尔特曼（Sam Altman）以及埃隆·马斯克（Elon Musk）等人商讨捐赠股份的可能性。布拉德重申，他坚决反对由政府强制“敲诈”企业股份，并将其扔进由伯尼·桑德斯或 AOC（亚历山大·奥卡西奥-科尔特斯）等人掌控的华盛顿政府“小金库”资金池；相反，捐赠必须是完全自愿的，且必须直接落入属于公民个人的私人账户中，伴随他们的一生进行复利增长。
关于如何消化巨额捐赠并规避 5,000 美元年度限额的问题，布拉德给出了两重机制方案：
1. **庞大基数直接分摊**：当全美 5000 万个账户建立完毕后，即便是几十亿美元的捐赠，平均摊派到每个孩子头上也不会突破 5,000 美元的年上限。
2. **设立滚动分发资金池（Pooled Account）**：大额股权或现金可以先进入一个托管资金池，在当年按法规上限分配给在册的所有儿童；超出当年度限额的剩余资产，则按时间梯度滚动分发给来年出生的新生儿，以及往后每一年新出生的儿童群体。
布拉德最后展望了该体系的宏大蓝图：美国目前每年约有 350 万至 370 万名新生儿出生。以现有约 7000 万名未成年人为基础，叠加每年新增的 370 万新生儿，全美有望在未来十年内建立超过 1 亿个私人投资账户。这一平台将永久开放，无论是家庭、企业雇主，还是科技新贵与慈善机构，都可以持续向全美下一代的独立账户中注资，彻底重塑整个国家的财富底层逻辑。
### [01:20:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4800s) 全民持股革命：Robinhood 打造的账户、直接慈善与复利奇迹
布拉德·格斯特纳（Brad Gerstner）描述了“特朗普账户”（Trump Accounts）在现实生活中的生动使用场景：父母或者正在打暑期工的孩子，往账户里存入 10 美元或 100 美元，然后直接在手机 App 上实时查看余额。最初，布拉德和他的儿子们亲自参与设计了这个 App 的早期原型，随后 Robinhood 首席执行官弗拉德·特内夫（Vlad Tenev）及其团队接手，将其开发成了一款极其精致优雅的正式产品。通过这种微型投资界面，每个普通美国孩子都能直接持有英伟达（Nvidia）、微软（Microsoft）和苹果（Apple）的零碎股票。大卫·萨克斯（David Sacks）深表赞同地指出，想象一下一个初中或高中生在手机上打开这个“财富页面”时的震撼心情——他们不再是局外人，而是真真切切地拥有了美国顶尖企业的股权，这种“置身其中”（in the game）的拥有感将彻底改变年轻一代对资本主义的认知。
```mermaid
flowchart TD
accTitle: Trump Accounts Capital Flow Architecture
accDescr: Capital inflow from parents, employers, and philanthropists into child accounts compounding over time.
subgraph INFLOW["资金输入渠道"]
A["家庭与个人储蓄<br>(每年上限 $5,000)"]
B["雇主免税配资<br>(每年最高 $2,500)"]
C["慈善家直接赠股<br>(如 SpaceX 股权)"]
end
subgraph PLATFORM["账户架构与管理"]
D["特朗普账户 (Trump Account)<br>Robinhood 移动端界面"]
E["底层资产配置<br>(Nvidia / Apple / 标普指数)"]
end
subgraph OUTCOME["长期复利与社会效益"]
F["15 年内为底层家庭创造<br>2万亿至4万亿美元资产"]
G["18岁后滚存入 Roth IRA<br>打破贫困代际传递"]
end
A --> D
B --> D
C --> D
D --> E
E --> F
E --> G
```
对于那些有能力每年顶格存入 5,000 美元的家庭而言，这无疑是极具吸引力的免税储蓄工具；但布拉德、迈克尔·戴尔（Michael Dell）以及格温·肖特韦尔（Gwynne Shotwell）等人最核心的关切，是那 50% 感到被体制抛弃、若无此机制存款将完全为零的底层群体。如果推演未来 15 年的复利轨迹，这项计划将直接向这些一无所有的家庭注入 2 万亿至 4 万亿美元的财富。与传统慈善机构相比，该账户引入了颠覆性的“直接赠予”模式：例如格温·肖特韦尔直接向 200 万名儿童每人捐赠价值约 150 美元的 SpaceX 股票，直接计入各账户额度。这种点对点的赠予彻底绕过了拿走 40% 预算用来支付豪华办公室和管理层高薪、充斥着寻租套利的非政府组织（NGO）工业复合体。更惊人的是复利的力量：历史数据显示，若自出生起顶格存入该账户并按过去 30 年的美股平均回报率计算，孩子到 28 岁时就将成为百万富翁；即便家庭无力顶格、每月仅存入 50 美元，也足以彻底重塑命运。正如特朗普总统对布拉德所明确要求的：现有的 529 教育储蓄计划只服务了前 10% 的高收入人群，而这项账户必须贯彻“主街议程”（Main Street agenda），通过手机上的私人产权账户重建普通人的美国梦，这是自 1935 年通过《社会保障法》（Social Security Act）以来美国社会契约迎来的最深刻变革。
- **手机端微股权持有**：孩子可在手机上直观看到自己拥有的英伟达、微软和苹果等核心科技资产的零碎份额。
- **直接穿透式慈善**：亿万富翁直接向个人账户捐赠股权（如 SpaceX 股票直接记入儿童账户），避免了传统 NGO 截留高达 40% 管理成本的低效与消耗。
- **惊人的复利数据**：若每年顶格存入并保持过去 30 年的市场基准收益，持有人在 28 岁时将累积超过 100 万美元财富。
- **雇主免税激励**：雇主可选择将员工最多 2,500 美元的薪资或补贴直接注入其子女的账户，双向抵扣应税收入，实现劳资双赢。
---
### [01:22:52](https://www.youtube.com/watch?v=PHL1j2ti420&t=4972s) 搁置党派偏见：从 K 型复苏到“全民股权国家”的救赎
杰森·卡拉卡尼斯（Jason Calacanis）在此刻将讨论提升到了更高的政治哲学层面。他直言不讳地表示，自己在节目中一直坚持实事求是、对本届政府的各种政策直言批评，但他必须严肃反驳那些仅仅因为该计划冠以“特朗普账户”之名就对其大加抨击的人。杰森透露，自己此前在扑克牌局上就曾当面建议布拉德：“干脆就叫它特朗普账户吧！”他强调，无论某些人对特朗普抱有多深的情感抵触或“特朗普妄想综合征”（Trump Derangement Syndrome, TDS），都必须将党派偏见放在一旁。美国当前正在经历严重的 K 型复苏，贫富鸿沟与社会阶层撕裂已激化到前所未有的地步，以至于大量年轻一代开始幼稚地认为社会主义甚至共产主义优于人类有史以来最伟大的体制——民主加资本主义。
杰森充满激情地谈到，年轻人天生热爱资本主义和创业创造，但美国当下正处于一个生死存亡的历史关口：在这个国家迎来建国 250 周年之际，整整两代年轻人却对“美国实验”失去了信仰。社会上弥漫着极端仇富情绪，从叫嚣着掠夺肯·格里芬（Ken Griffin）的豪宅、要求一切免费，到接连发生企业 CEO 遭枪击、住所被投掷燃烧弹的暴力事件。过去几十年里，比尔·盖茨和沃伦·巴菲特在 TED 大会上大声疾呼富豪签署“捐赠誓言”（The Giving Pledge），但那种死后才清算财富的方式远不如特朗普账户直接有效——特朗普账户才是“捐赠誓言”的终极进化版，它以制度化的手段让每一个美国人都能在成长过程中成为这个国家最精髓部分（企业家精神与商业资产）的股东，真正建立一个“股权国家”（equity nation）。
- **超越党派标签**：呼吁公众搁置对特朗普个人的偏见与政治站队，认清全民资产积累对弥合社会撕裂的根本意义。
- **遏制激进思潮**：直面两代年轻人对资本主义制度丧失信心的现实危机，用直接持股取代极左的均贫仇富心理。
- **重构捐赠誓言**：将传统富豪身故后的资金捐赠，转变为生前直接赋能下一代美国公民的普惠资本注入。
---
### [01:25:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5100s) 借鉴澳洲养老金模式：重塑社会契约与超越代际焦虑的未来
杰森进一步将这项制度与澳大利亚的退休公积金制度（Superannuation）进行了对比。澳大利亚之所以常年位列全球最幸福、最具安全感的国家之一，核心原因就在于其成熟的强制储蓄机制——法律规定必须将个人收入的 12% 至 14% 划入可自主投资的退休基金账户中。反观美国，许多家庭之所以陷入绝望、甚至不敢生育下一代，是因为他们感到整个系统对底层过于残酷；无论是单亲父母还是拿最低工资苦苦挣扎的蓝领，他们穷尽一生最大的焦虑就是无法为子女提供确定性的未来。如果父母知道自己的孩子从出生起就拥有一个不断滚存的股权底盘，这种代际焦虑将得到根本性缓解。特朗普账户的作用机制正是这种基础性的资产积累底座，甚至在长远来看足以重塑并替代现行难以为继的传统社会保障体系。
谈及此处，杰森向坐在身旁的布拉德·格斯特纳送上了由衷的赞誉（"give you your flowers"）。杰森动情地表示，身边的很多朋友都曾卷入政治，政治往往极其撕裂且充满争议，但布拉德凭借极大的谦逊与智慧，在两党极端对立的政治雷区中“穿针引线”，完成了一堂教科书级别的制度设计大师课。杰森直言，这项成就的社会意义远超布拉德一生中所有商业投资成就的总和，他将因为构建了这个比 1935 年社会保障法更加宏大、更加高效的新架构而被历史牢记。大卫·萨克斯与布拉德也探讨了该账户的动态延展性：针对网友关于“为何在 18 岁终止”的疑问，布拉德指出该账户在孩子成年后可直接无缝滚存（Rollover）为个人退休账户（IRA 或 Roth IRA），未来甚至可以允许个人继续享受每年 5,000 美元的注资额度。这种将下一代完全迁移到高效的私人资本账户、逐步日落（sunset）传统低效且面临破产的政府社保模式，将从根本上杜绝政府低效官僚的干预，彻底将财富管理的主动权交还给每一个普通的美国公民。
- **澳洲 Superannuation 启示**：借鉴成熟的法定储蓄与长期投资模式，从根本上解决年轻群体的生育恐惧与家庭代际贫困焦虑。
- **两党缝合的制度杰作**：杰森盛赞布拉德在两党极化分裂的环境下成功推动了这项跨时代立法，将其确立为造福数代人的个人历史遗产。
- **全生命周期资本滚存**：账户满 18 岁后直接导入 Roth IRA 等成熟投资工具，具备向成年阶段平滑延展的强大制度弹性。
- **市场化替代传统社保**：通过建立全民去中心化的私人资产网络，为最终和平替代运转低效、债务累累的联邦政府社会保障体系提供了现实路径。
### [01:30:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5400s) 规则设计与财商教育：如何避免年轻人“挥霍一空”
针对大卫·萨克斯（David Sacks）和杰森·卡拉卡尼斯（Jason Calacanis）关于年轻人满 18 岁是否会把钱“全拿去 YOLO（挥霍享受）”的担忧，布拉德·格斯特纳（Brad Gerstner）解释了该账户的政治妥协与制度防线。布拉德最初的构想是强制资金复利滚存至 30 岁，因为人在 30 岁时对财务的认知往往更加成熟；但在政策博弈中，反对者提出“如果你到了 18 岁既能投票又能上战场打仗，政府凭什么扣着你的钱不给”。最终达成的两党妥协设立了严格限制：年满 18 岁时，年轻人最多只能提取账户资金的 25%，且必须用于买房、创业或上大学；剩余的 75% 将自动转入个人退休账户（IRA），若提前支取将面临高额惩罚金，从机制上抑制盲目提现。
更深层的意义在于彻底激活公共教育系统中的财商教育。布拉德指出，让一个身无分文的孩子学理财毫无吸引力，但如果一个初一学生发现自己的账户里有真实滚存的 12 美元，并且拥有微量耐克（Nike）、苹果（Apple）或英伟达（Nvidia）的碎股，现实教学就瞬间鲜活起来——孩子们会主动探究“这笔钱是怎么来的？复利是如何计算的？如果我每月存入 50 美元未来会变成多少？”目前全美已有 37 个州在法律上要求推行财商教育，全美已有约 25 个州（包括俄克拉荷马州、西弗吉尼亚州、印第安纳州等）正准备直接由州财政向辖区内儿童的账户注资。各州政府正在梳理过去那些低效的儿童补贴项目，砍掉中间官僚机构，将资金以整笔拨款（Block Grant）形式直接发给孩子，因为数据证明直接赋予资产能显著提升高中毕业率、置业率与创业率。
这不仅是一项公共政策，更是布拉德深沉的个人历程。布拉德回顾道，这项计划最初的草图是他于 2020 年秋天在自家餐桌上与两个儿子共同勾勒出来的，他的儿子林肯（Lincoln）全程陪同参加了与每一位国会议员、参议员乃至总统的所有会谈。布拉德与迈克尔·戴尔及苏珊·戴尔夫妇（Michael & Susan Dell）携手推进此事，他在白宫椭圆形办公室宣布该项目上线时感慨，这是他人生中最为崇高的一份荣耀。
```mermaid
flowchart TD
accTitle: 特朗普账户资金运作与提取机制
accDescr: 展示从0岁注资、资产复合增长到18岁条件提取及转入IRA的资金流向结构
subgraph Phase1 ["0 - 18 岁：全资复合积累期"]
A["联邦种子金 / 慈善捐赠 / 雇主免税注资 / 父母储蓄"] --> B["特朗普账户 (投资标的涵盖标普/科技权重股)"]
B --> C["美股大盘长期复利滚存"]
end
subgraph Phase2 ["18 岁成人节点：分流与约束"]
C --> D{"年满 18 岁资金分配"}
D -->|最多提取 25%| E["法定指定用途：首次购房 / 创办企业 / 大学学费"]
D -->|至少保留 75%| F["自动滚入传统/罗斯 IRA 账户"]
end
subgraph Phase3 ["长期退休储备"]
F --> G["继续复利至退休年龄<br/>(提前支取设法定惩罚机制)"]
end
```
---
### [01:33:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5580s) “梦之队”打造顶级政务软件：两党共识与资本主义所有权
杰森特别指出了该项目在执行力层面的罕见奇迹：美国联邦政府居然在短短 18 个月内交付了一款体验极其顺滑、媲美硅谷一线水准的消费级软件。Airbnb 联合创始人乔·吉比亚（Joe Gebbia）在顶着科技界各种非议的情况下加入本届政府，发挥了世界级产品设计师的关键作用。播客制作人尼克（Nick Calacanis）及其妻子的注册体验极其惊艳。布拉德透露，过去一年里，由他本人、迈克尔·戴尔、Robinhood 首席执行官弗拉德·特内夫（Vlad Tenev）、乔·吉比亚以及财政部长和财政部官员卢克·佩蒂特（Luke Pettit）组成的“梦之队”几乎每天开会，目标绝不是“做一款政府历史上最好的产品”，而是“做出一款跻身全球顶流行列的消费级互联网产品”。上线后的用户评分、活跃度与转化率让硅谷任何顶级消费级软件公司都会为之眼红。
该计划还展现了华盛顿久违的跨党派共识。马里兰州民主党籍州长韦斯·摩尔（Wes Moore）公开表态，共和党和民主党在过去 40 年里一直试图摸索出类似的资产建设机制，如今这一代人终于将其变为现实。在美利坚合众国建国 250 周年之际，站在椭圆形办公室面对《独立宣言》原稿，这项法案具有强烈的象征意味。
面对当下部分激进极左思潮（如部分政客试图“点火烧掉这套制度实验”）的呼声，布拉德回应道，解决社会焦虑的正确路径绝非拥抱社会主义，而是对运行了 250 年的成熟制度加倍下注——与其让下一代滑向福利依赖，不如让他们在人生第一天就直接进入资本主义体系，成为美利坚合众国的“股权所有者”与“股东”。
- **产品打造梦之队**：由布拉德·格斯特纳、迈克尔·戴尔、弗拉德·特内夫（Robinhood CEO）、乔·吉比亚（Airbnb 联合创始人）与美财政部高管联合闭门打磨一年。
- **两党罕见一致**：民主党州长（如韦斯·摩尔）公开盛赞，认为打破了 40 年的政策僵局。
- **所有权哲学**：以真正的股权所有制击碎社会主义幻象，实现全民分享资本主义复利红利。
---
### [01:36:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5760s) 补全人生的“复利雪坡”：从特朗普账户到全民财富代际传承
杰森强调，科技资本与创造者必须展示出真正的同理心，给底层民众一条重新相信“美国梦”的坚实阶梯；与其抱怨 7 美元的联邦最低时薪和生活恐惧，不如用实实在在的资产积累制度取代极左派的激进幻想。布拉德顺势提出，特朗普账户为未来联邦政府获取的重大经济利益提供了一个完美的“主权财富分配承载底座”——无论是政府推动 AI 公司发展换取的股权、英特尔（Intel）芯片法案的股权分成，还是未来针对 TikTok 等海外实体的许可抽成，都可以直接注资到所有普通公民的个人账户中，享受一生的复利增长。
此外，该账户为打破社会保障制度（Social Security）的沉疴提供了全新范式。目前的社会保障税直接从工资单中扣除 12.4%，但资金宛如流入政府的“黑洞”：
1. 纳税人对账户资产没有法律产权，无法自主控制；
2. 即使向 3000 人提问，也没人知道自己确切缴纳并沉淀了多少钱；
3. 一旦缴纳人去世，这笔资金无法作为遗产顺位继承给子孙。
布拉德提出，无需大动干戈修改社保基本盘，只需将这一账户模式横向推广为面向 20 至 40 岁年轻人的“补充型 IRA”，赋予他们真正的私有产权、可继承权与投资参与权。
```
传统 IRA 与特朗普账户（Trump Accounts）机制对比：
┌────────────────────────┬───────────────────────────┬───────────────────────────┐
│ 比较维度 │ 传统 IRA / Roth IRA │ 特朗普账户 (Trump Account) │
├────────────────────────┼───────────────────────────┼───────────────────────────┤
│ 准入门槛 │ 必须有 W-2 等法定劳动收入 │ 出生即自动开设，零门槛 │
│ 起步年龄 │ 通常在 22~25 岁就业后 │ 0 岁出生第一天 │
│ 雇主免税注资支持 │ 无（需走 401k 渠道） │ 雇主可每年免税注入 $2,500 │
│ 第三方与慈善注入 │ 极度受限 │ 支持父母、慈善基金及州政府│
│ 产权与代际传承 │ 私有产权，可继承 │ 明确个人私有产权，完全继承│
│ 复利雪坡完整度 │ 截断了人生前 25 年的复利 │ 完整覆盖 0~25 岁黄金复利期│
└────────────────────────┴───────────────────────────┴───────────────────────────┘
```
对话最后探讨了这一产品对中产家庭规划的深远重构。注册会计师（CPA）群体普遍将其视为中产家庭财富规划的“终极神作”。在现行税法下，一个人在参加工作（通常是 22 岁以上）之前根本无法开立享受税收优惠的 IRA 账户，相当于美国体制生生砍掉了人生最初、也是最容易实现裂变增长的三分之一复利时间。沃伦·巴菲特（Warren Buffett）曾说“人生就像滚雪球，重要的是找到很湿的雪和很长的坡”，而特朗普账户恰恰补齐了过去被截断的黄金 25 年，不仅允许父母、慈善机构甚至雇主每年免税存入最高 2500 美元，更让每个美国孩子在出生时就站在了雪坡的顶端。在一番关于白宫敲钟的回忆与经典口头禅（"Let your winners ride"、"Queen of Quinoa"、"Rain Man David Sacks"）的互怼打趣中，全期节目在一片欢声笑语中正式收官。
## Notable quotes
> [02:58](https://www.youtube.com/watch?v=PHL1j2ti420&t=178s) 底层大模型的能力已经遭遇平台期、出现渐近线效应，如今必须灌入指数级暴涨的 Token 数量，才能榨出微弱的技术迭代。 — Chamath Palihapitiya
> [20:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1200s) 机器正在制造机器 (The machine is building the machine)。 — Brad Gerstner
> [27:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=1620s) 某种资源的利用效率越高、成本越便宜，人类反而会成倍甚至数量级式地消耗更多这种资源。 — Brad Gerstner
> [42:47](https://www.youtube.com/watch?v=PHL1j2ti420&t=2567s) 企业界对于‘模型可替代性’或‘无头化架构’有着强烈渴望，但目前没有任何人能真正将上下文状态和对话记忆从特定的模型底层架构中完全抽象剥离出来。 — David Sacks
> [45:10](https://www.youtube.com/watch?v=PHL1j2ti420&t=2710s) 智能根本没有收敛，随着超级智能具备自我递归能力，领先者的优势不仅不会缩小，反而会无限放大。 — Brad Gerstner
> [54:29](https://www.youtube.com/watch?v=PHL1j2ti420&t=3269s) 科技博弈中的经典周期：落后时开源，赶超后闭源。 — David Sacks
> [01:10:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=4200s) 医治更多社会主义的解药，就是更多的资本主义。 — Brad Gerstner
> [01:30:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5400s) 如果一个人在 18 岁时既能投票又能上战场打仗，政府凭什么扣着你的钱不给。 — Brad Gerstner
> [01:13:50](https://www.youtube.com/watch?v=PHL1j2ti420&t=4430s) 我们这简直是在批量制造‘裙带宝贝’，未来全美到处都是坐拥信托基金的富二代了！但这绝对比联邦社会保障靠谱得多。 — Chamath Palihapitiya
> [01:36:00](https://www.youtube.com/watch?v=PHL1j2ti420&t=5760s) 人生就像滚雪球，重要的是找到很湿的雪和很长的坡，而特朗普账户恰恰补齐了过去被截断的黄金 25 年。 — Brad Gerstner
## Mentioned
Jason Calacanis, David Friedberg, Brad Gerstner, Chamath Palihapitiya, David Sacks, Marc Benioff, Jensen Huang, Brad Smith, Anthony Tan, Tom Brown, Gavin Baker, Elon Musk, Gwynne Shotwell, Bret Johnsen, Nikesh Arora, Praveen, Dara Khosrowshahi, Mark Zuckerberg, Andy Fang, Alex Karp, Ali Ghodsi, Anton, Mati, Sam Altman, Dario Amodei, Bernie Sanders, Ro Khanna, Cory Booker, Gavin Newsom, Wes Moore, John Fetterman, Zohran Mamdani, Vlad Tenev, Michael Dell, Susan Dell, Ken Griffin, Bill Gates, Warren Buffett, Joe Gebbia, Luke Pettit, OpenAI, Anthropic, SpaceX, Meta, Salesforce, NVIDIA, Microsoft, Apple, Uber, Snowflake, Databricks, Coinbase, DoorDash, Decagon, ElevenLabs, Lovable, Robinhood, Micron, T. Rowe Price, Fidelity, Altimeter Capital, Groq, Cerebras, Baseten, Fireworks, Nous Research, OpenRouter, Bittensor (TAO), ByteDance, Alibaba (Qwen), Zhipu AI (GLM), Moonshot AI (Kimi), Polymarket, ChatGPT, Claude, GPT-6, Sora, Spark 1.1, Hermes Agent, Neo Terra (NEOTRA), Falcon, Humain, Invest America Act (Trump Accounts)
## Video description
(0:00) Bestie intros: Brad Gerstner fills in for Friedberg!
(2:58) OpenAI vs Anthropic IPOs: Why it matters who goes first, what they learned from the SpaceX IPO, the unlimited TAM of intelligence
(27:39) The open source decision, Meta's new model, Zuck's price war, AI duopoly
(54:29) CCP considering putting export controls on Chinese models, is open source ending in China?
(1:03:09) Trump Accounts launch, getting young Americans bought back into capitalism
Apply for Summit 2026:
https://allin.com/events
Follow Brad:
https://x.com/altcap
Follow the besties:
https://x.com/chamath
https://x.com/Jason
https://x.com/DavidSacks
https://x.com/friedberg
Follow on X:
https://x.com/theallinpod
Follow on Instagram:
https://www.instagram.com/theallinpod
Follow on TikTok:
https://www.tiktok.com/@theallinpod
Follow on LinkedIn:
https://www.linkedin.com/company/allinpod
Intro Music Credit:
https://rb.gy/tppkzl
https://x.com/yung_spielburg
Intro Video Credit:
https://x.com/TheZachEffect
Referenced in the show:
https://polymarket.com/event/ipos-before-2027
https://x.com/thejessezhang/status/2074154325933424861
https://x.com/praveenTweets/status/2074605343439810922
https://x.com/nikesharora/status/2074802778074124434
https://x.com/nikesharora/status/2074814752174522857
https://x.com/brian_armstrong/status/2070670644577280109
https://x.com/andyfang/status/2074252174226493584
https://x.com/nikesharora/status/2074630732019036574
https://x.com/finkd/status/2075218444056707458
https://x.com/alighodsi/status/2074996561306955958
https://blog.nicolasmeridjen.com/en/blog/2026-04-03-alibaba-qwen-closed-source-end-of-open-weight
https://www.chinatalk.media/p/chinas-ai-companies-are-going-closed
https://www.reuters.com/world/beijing-is-looking-curbing-overseas-access-chinas-top-ai-models-sources-say-2026-07-07
https://x.com/KurtSupeCPA/thread/2074817292550984010
#allin #tech #news
[Source](https://www.youtube.com/watch?v=PHL1j2ti420)
Source quotes
[2:58] Chamath Palihapitiya: 底层大模型的能力已经遭遇平台期、出现渐近线效应，如今必须灌入指数级暴涨的 Token 数量，才能榨出微弱的技术迭代。
[20:00] Brad Gerstner: 机器正在制造机器 (The machine is building the machine)。
[27:00] Brad Gerstner: 某种资源的利用效率越高、成本越便宜，人类反而会成倍甚至数量级式地消耗更多这种资源。
[42:47] David Sacks: 企业界对于‘模型可替代性’或‘无头化架构’有着强烈渴望，但目前没有任何人能真正将上下文状态和对话记忆从特定的模型底层架构中完全抽象剥离出来。
[45:10] Brad Gerstner: 智能根本没有收敛，随着超级智能具备自我递归能力，领先者的优势不仅不会缩小，反而会无限放大。
[54:29] David Sacks: 科技博弈中的经典周期：落后时开源，赶超后闭源。
[70:00] Brad Gerstner: 医治更多社会主义的解药，就是更多的资本主义。
[90:00] Brad Gerstner: 如果一个人在 18 岁时既能投票又能上战场打仗，政府凭什么扣着你的钱不给。
[73:50] Chamath Palihapitiya: 我们这简直是在批量制造‘裙带宝贝’，未来全美到处都是坐拥信托基金的富二代了！但这绝对比联邦社会保障靠谱得多。
[96:00] Brad Gerstner: 人生就像滚雪球，重要的是找到很湿的雪和很长的坡，而特朗普账户恰恰补齐了过去被截断的黄金 25 年。
Source quotes
[13:02] Chamath Palihapitiya: Right now, our token costs are doubling every 45 days... our upside is essentially flat... we've effectively already asymptoted. I suspect that if you can get out now, you should get out now before all of that starts to seep into the water table.
[25:05] When you have unlimited tokens as an exercise... it really does change your thinking when costs go down.
[35:00] If you're talking about replacing a software engineer for two hours, that may take two million expensive tokens. And the consequence of using something that's 95% as good is really high... if the task breaks early, or it breaks in the middle, or it breaks at the end, there's a huge cost to that.
The absolute best thing that could happen for America in terms of winning the AI race against China is if China somehow sprouted their own doomer community... We need to get their P-Doom up.
[70:00] Brad Gerstner: The antidote to more socialism is more capitalism.
[73:09] David Sacks: We're gonna have a lot of rich kids with trust funds
[76:51] Brad Gerstner: Imagine opening up that account... and seeing, oh man, I'm in the game. I have ownership
Original transcript
[0:00] all right everybody welcome back number one podcast in the world it's july all in episode 280 freeberg is on a little vacay we'll leave it at that and uh yeah bestie brad is here how you doing brad i'm doing great i'm doing a vacay in uh maybe idaho or somewhere jay cal you know who knows who knows who knows who knows it could be anywhere it could be anywhere i mean there's lots of things he He could be in plenty of places. And it looks like you are somewhere in the Northeast. I'll leave it at that. You having a little vacay for yourself this week? I'm in my flag room, very patriotic room here. Very nice. You know, where I work on the East Coast in the summertime. Yes. And spent some time in D.C. this week. And it's been a great week. It's been a great week celebrating America 250. Great. And you're going to be out of there by the second week of August, yeah? So I have it August 10th through the 30th. I'm good? Exactly. Exactly. oh absolutely you don't know the half of it man i am on a summer bender i'm like oh where's your where your vacation by the way where are you where are you right i am in paris i did uh about eight interviews for uh at the raise conference they be coming out uh in the all feed And of course Cackling from the factory Look at him You're working in the factory. Chamath Palihapitiya. It's going to be a hot software summer for Chamath. How's your hot software summer going? It's good. Selling enterprise software is hard, but it's good. Chamath's like, man, I was such a dick to all my CEOs. Hey, the SaaS period. And now you know. I went to Geneva. I went to Geneva. He works out of Europe, sees all his European customers, and he had a dinner in Geneva, which I joined. And then me, Jensen, Brad Smith, Anthony Tan from Grab, and a bunch of other folks were put on this UN Commission for AI that Mark is the co-chairman of. When you see Mark Benioff in Action Man, this guy is a master. He is the impresario of impresarios. You see how he's built such a ginormous business. It's impressive. It's impressive. What is a commission by the United Nations for AI? What is their calling? Open source. No, but no, I mean, it's like the United Nations actually do anything. Did they actually do anything It so funny Yeah I mean Anthropic was there too One of the co Tom co founders Tom Brown because his name Yeah that good Was the anthropic guy running around saying it the end of the world No no no, he was very Tom's awesome. Tom's awesome. In fairness to him, he wears it on his sleeve, which is like, hey, we really believe we're doing the right thing. Right. And just trust us. And I think the future is open source for all these countries. well we're going to get into that that's on the docket for sure but let's start with the ipo update you know there's a trillion dollar ipo rush to the exits and you know this was a big topic of discussion brad at the liquidity summit last month and we've never seen trillion dollar ipo we had one this year already spacex trading right about where it went public so it was priced i guess to perfection and theoretically i'm going to see two more uh brad has the inside information so i'll I'll try to get it out of open AI and anthropic are slated to go out. Let's just go quickly over what happened with spacex. It ran up to $200 a share. It's been down a bit. It's at $150 sure. As I said, that's right at the IPO price. So it trading at that two trillion market cap currently seven largest company in the world and anthropic confidentially filed on June 1st I don know why they call this confidential filing when it immediately comes out but I guess the information is confidential. Polymarket says 65% chance Anthropic's IPO will happen this year on light volume, 360K. And two weeks ago, Gavin Baker, another bestie, said he thinks they're going to end 2026 with over a hundred billion dollars in revenue and very profitable he said a couple of us guests on the program that he thinks it would trade at three trillion right now if it went public chamath you made a great call on the pod you said hey good idea for elon to get out first what are the chances here chamath that these other two get out this year or maybe in like you know say nine months in the in the first quarter of next year we'll start there well i think that these are all great businesses. I think the question is, what is the market clearing price? And I think that's more of a function of how much appetite the markets have to absorb new issues and at what scale. That's number one. And I think that's mostly determined by price. So I think Anthropic and OpenAI are probably in two different places. The last time we heard from
[5:00] OpenAI, their cash burn was still quite high just because of the diffuse nature of their business and more reliance on consumer than enterprise. I think Brad mentioned it in one of the pods that Anthropic may actually be accidentally profitable. I think he said something like that. Yeah. Let me tell you something really interesting. I sat down with my CTO today and I said, how are we doing on token spend? And he said the most incredible thing. He said, right now, our token costs are doubling every 45 days. Okay. And I was like, ugh. And he said, yeah. And I said, well, what is the downstream productivity? And he said, maybe 5% max. Okay. And I said, okay, so my costs are doubling every 45 days. My upside is essentially flat. And he said, basically. And I said, well, explain why that is. And he said, honestly, what we're finding out is that you need to use a lot more tokens to get to this next iteration of improvement because we've effectively already asymptoted. And I said, so what should we do? And he said honestly we have to figure this out And so we going to take a step back and try to figure out what to do I don know how many other companies will actually go through this reckoning now But the point is everybody in the next three or four years will for sure go through it So I suspect that if you can get out now, you should get out now before all of that starts to seep into the water table. Because I think that's probably what allows you to get out at a huge price and raise a huge amount of money. All right, Brad, you are well invested and well known for being invested in these two next IPOs. So you probably have some good insights since you've talked to them on a regular basis. Chances they get out in the next six to nine months, both of them, you'd say 100% chance unless there's some outside event, you know, blockade of Taiwan, some black swan event that we're not anticipating. What do you think the chances are they're public when we're sitting here and I'm skiing in Hokkaido? Yeah, I think it's very high. But let me first say, you know, the SpaceX IPO where we were also investors and we also bought in the IPO. I mean, it was textbook. It was a hugely successful IPO. They raised billion at trillion Okay so it went out below where we are today It up 25 You know and let call it on 35 billion of forward revenue So if you think about that revenue multiples trading at 2 trillion on roughly 35 billion of forward revenue, it's an incredible achievement. I think it was textbook, I think Anthropic and OpenAI were watching very closely because frankly, we had not had an IPO of that size. And to Elon's credit and to the team's credit, Brett and Gwen, And they really pioneered some really smart and interesting things as part of that IPO. So, you know, you heard from Gavin, Anthropics rumored to be, you know, trending over 100 billion in revenue compared to the 35, right? If they exit the year at 100, that means their gap revenue next year could be well over 100. So based on the SpaceX success, I think it would be a blockbuster IPO. and I think SpaceX has shown them the way on things like the total raise, pricing, liquidity, inclusion into the indexes, how to do the lockup. I think they've gone to school. It was a staged release in terms of getting out of the lockup. It has to hit certain milestones, and some of those are time. Early inclusion in the index, raised billion The early inclusion in the index let me have you unpack that for a second because people said hey maybe this feels unfair that they should be forced to buy it what your take on that is that just like haters gonna hate or is there something to that I think there was legitimate concern right this is what I'm gonna be a concern yeah the legitimate concern is that a company that had not been through the process of being vetted post-ipo there's a lot of volatility you've seen that chart jason sure that the the peak to trough drawdown in the six months post ipo is 50 we've seen a pretty big drawdown here from the peak to trough as well so you don't want to jam it into an index at the peak and then have a 30 drawdown on top of people which often happens in ipos because people get excited it runs ahead of itself but they didn't do that here there was fear that that was going to happen so both the the exchanges and the indexes, they looked at this and they made some modifications because the other side of the argument is it's so damn big and important that it needs to be part of the index. Right. Right. And so the reason the rules had previously existed is because most companies coming public were younger, earlier, less tested, less revenues, less profitable. All the things weren't as important in the overall scheme of things. So I think that they pioneered some
[10:00] really smart things. It's worked well, it's traded well. And so I think that that provides a bit of a blueprint for anthropic but just as in terms of the enthusiasm is altimeter is a fidelity is a T row an enthusiastic buyer of anthropic based upon the things we know today around profitability and model improvement and revenue growth, etc. Yes, everybody would be pig piling in everybody would be trying to get into the top of the book. And you know, the last I heard you know again rumored that they would like to get out this year on open ai everybody knows that anthropic kind of passed open ai on a revenue trajectory but i will tell you open ai has kind of got its swagger and mojo back it's coming out you know just today with a whole new set of models we know gpt6 you know there's a lot of talk of that coming out within the next 30 days a whole new generation of models i think their revenue has really ticked back up the most recent kind of rumors i see on twitter is around 70 billion dollars texas year so just as a reminder 70 billion may not be over a hundred billion that rumored and anthropic but it still twice uh you know where where the revenue of spacex is at so can they get out at over a trillion on that type of revenue growth being one of the two frontier premier labs i think the answer to that is yes um i'm not sure there's a huge race between the two of them to get out first i think they'll both go out when it's when it's time i think open ai has a little bit more complexity just associated with the corporate restructuring that they have to go through etc so i would be surprised if they they go out before anthropic but the fact of the matter is i don't know but today as i sit here today altimeter would be a buyer at scale and at size in both of those ipos at three trillion are you a buyer or are you uh hey you know it's obviously going to trade up and down and i there's no rush because you i think were the one who said on the pod or might have been at liquidity live and when i asked you point blank hey should retail get involved in spacex what's your thoughts right and you're like hey listen it's a four percent flow five percent float it's going to trade up and down but you know a year from now it might be trading at the same basic price it's going to be priced not to perfection which it seems to have been but you i think your position was it's going to be priced reasonably there'll be plenty of time to get in you don't have to like you know panic about getting your shares yeah i once a company valued at over a trillion dollars like the get rich quick schemes are over right like that you and i share a deep passion jason we got to get retail investors we got to get the citizens of the united states in on these value creating opportunities earlier right the accredited investor laws are insane that we have in this country and keeps people from participating in these things but it is what it is right so they're coming public at over a trillion dollars i still think there's a lot of meat on the bone on spacex on anthropic on open ai but you're not going to have things that are i don't expect that they're going to be priced in a way where you're going to get a 50 to 100 durable bounce out of the ipos if so that would mean they were probably mispriced right into the ipo but i do think that these things can be compounders they're going to compound at the rate they compound revenue and i think all of these companies are going to compound revenue at well over 30 for the next many years and 30 a year for just so people understand that this is high growth in public markets on very large revenue numbers already you know growing 30 when you have 100 million revenues one thing growing 30 you know when you got 10 billion or 100 billion you know this is this becomes a different uh task so let talk a little bit about these two companies chamath and what the public going to perceive them as chat gpt seemed to be the public brand the consumer brand for you know large language models it's the ai for you know people who are doing their homework or mom and dad are trying to fix the dishwasher whatever and then claude took the lane of hey we're gonna be the one for corporate and it did seem like open ai got very distracted with with sora and you know disney relationship we're going to make a puck with johnny i've everything consumer then they realize oh wow the revenue seems to be an enterprise first is that going to wind up being the the big mistake when we look at it they kind of gave the google position the high growth position to claude and anthropic and they took the yahoo position or do you think they'll catch up on the enterprise or maybe they should just go back to trying to be the consumer version How are these going to be positioned a year from now? How's the public going to look at them? The problem with enterprise revenue is at some point, the person that's spending it has to see an ROI. I asked Fable 5 High. Anthropics new model. Yeah. Anthropics new model. I first asked it, what is the lift of the S&P 500 earnings per share growth since
[15:00] 2024 from AI and they answered oh it's 50 percent so then I looked through it and I said well no you're including the money that Nvidia makes for something for chips so I said okay I asked a different question which is then what was the EPS growth of the S&P 493 and the answer was nine percent and I said okay well that's different and I said unpack that and the overwhelming majority of that was from pricing power sitting on top of inflation and then the other three percent was from buybacks and so the answer as far as all publicly available data was that the actual roi was somewhere between zero and two percent so i don't know i mean i think that enterprise looks really good. The problem is that very smart investors like Brad and Gavin and others at some point will start asking companies, what's your ROI? What's the actual EPS lift? And if the answer is well I don really know or I not sure And you know you don necessarily have the pricing power to continue to raise prices Enterprise is probably a little bit more brittle because there are fewer buyers and they more demanding. Consumer on the other hand, then all of a sudden becomes an incredible safe harbor because you have 10s of millions of buyers. And having those two orders of magnitude more buyers at a much smaller price point in inoculates you from the vicissitudes of an ROI discussion. So it all really depends on what the actual ROI is of this money being spent. I think that we're in the phase of just being astonished, as Brad said, about the scale of the revenue growth. But at some point, you'd have to be an idiot not to ask, well, who is paying you this? And can they sustain paying it to you? I just don't know what the answer to that question is and at some point it may not be now at some point people will have to answer that question and interestingly million dollars a year on tokens and that million dollars a year is doubling and tripling and quadrupling at some point you're going to have to show an roi that above the risk rate of return otherwise you going to have some angry investors on your hands And our discussion here for the last couple of weeks on the pod has centered around that And the industry has responded on the place where all the CTOs, CEOs, and capital allocators hang out, which is X.com, formerly known as Twitter. Here's Praveen, the CTO of Uber. And so when you ask, like, how are they getting the ROI out of this? people are now bringing that conversation front and center and they're explaining it on x and he talked remember uber was also the one that ran through all their tokens in the first quarter so then on the other side of the business which is legal operations marketing customer support hr and procurement which he lists here he says in this you know today 99 of our engineers use ai tools okay great right it's everybody's you know doing vibe coding and has coding assistance more than 70 percent of pull requests are attributed to local or cloud agents our engineers have built 2500 agentic skills so how are we bringing agentic ai beyond engineering and what they've decided to do is essentially he talks about these agentic pods and this to me seems directionally how this should be done which is you find engineers and you uh as we as we uh talked about forward deployed engineers fancy way of saying put an engineer put them into departments and have them work with the department heads who understand systems thinking how their process is done And he says he making basically long and short of it is they making massive, massive progress on the operational side of the business. So Brad, you're pretty familiar with Uber and have been a long supporter of that. This is a company that knows how to deploy technology pretty well and they're an operations machine run by an operations machine dara they should report the eps gains attributable to ai yeah well i mean and this first step seems like they're really being thoughtful about this first hey this token spend got out of control with the developers we're going to need to pause this and look at it and then second here's how it's going to lower costs and create more efficiency so brad let's talk about that side of it not just token maxing with the developers hitting the slot machine of like okay let's see if this pull request and let's see if this produces the right code or not to these departments in a more strategic way. It's not just the person who works in HR, you know, using cloud code or perplexity or whatever and trying to vibe code something. This is, hey, we're sending engineering in to work with your top systems architect and we're going to, you know, try to find that ROI, yeah? Yes, you know, first I would say to
[20:00] Ath is right. The only question is on what time frame. There's no doubt that there's a lot of money being spent today that is in the experimental bucket, right? Where I think there probably isn't direct ROI, Chamath, to your point. But I think we're so early, nobody cares. I think we're so early in terms of enterprise adoption. Remember, the total addressable market here is every single, small, medium, large company on the planet. it. And so we've never seen revenue growth like this because we've never seen a TAM like this. And if you look at the distribution of revenues across these businesses, it's not like it's concentrated with four or five customers. There are millions of customers independently, economically making the decision that is rational for them every day that it makes sense, like Praveen at Uber. And of course, they're trying to find things on both right now, mostly the cost side, cost takeouts to justify the investments that they're making in tokens. But I think we're on the verge of breakthroughs in intelligence that's going to dramatically change the revenue side of the equation for a lot of these businesses, breakthroughs in life sciences breakthroughs in product innovation et cetera where they could not divorce themselves from this even if they wanted to For example Jensen Huang has talked many times that all of his design work all of his design work now at NVIDIA is using AI to design the next generation chip The machine is building the machine. So you can't get rid of that even if you wanted to. And tiny intelligence advantages at the frontier where he sits are required. Like there's no way I don't think that Jensen is going to use anything but the best models that he can to build out those capabilities. So I just think that we're not going to see that in the next few years. You're going to see it under the hood, of course. But that occurred at Snowflake. There was tons of optimization that occurred at Snowflake. But the revenue continued unabated. The revenue growth continued unabated because they further penetrated use cases, further penetrated the enterprise. So let me be provocative here. If these guys end the year over 100 billion, I think that they're on a revenue trajectory that they could three to five X again next year. We've never seen anything like this. Never. You're saying 100 to 300, 100 to 400. And Jason like you and I have talked about this Our minds were blown if a company could go from 100 million to 300 million We talking from 100 billion to 300 billion 200 billion of incremental revenue is incomprehensible in the history of Silicon Valley Okay. And just the fact that we're even in the history of the world. Yeah, in the history of the world. So the fact that we're even talking anywhere close to this tells us something different is going on here. I think the thing that's different is that intelligence is the largest TAM we've ever seen in the history of the world. These guys are penetrating it. So, yes, the super sophisticated companies, the 8090 and Chamath, are helping optimize their token spend that are early adopters. 100% that's occurring, but it's not really changing the trajectory that the frontier labs are on. yeah and one of the interesting things about this technology that's really unique we talk about intelligence on demand when you would make a piece of software or you had some technological innovation it would typically accrue to i don't know one group of people in an organization you know maybe two groups of people right excel comes out okay yeah the accounting department's having a field day with it but it's not really affecting human resources or marketing okay yeah maybe it trickles down eventually every single person in every single organization is playing with these tools So if everybody playing with it everybody trying to apply it all at the same time it kind of like you know you got a thousand person organization people are spending 200 a month okay yeah they double it every you know x number of months okay yeah now they're spending four hundred dollars a month per person okay they're spending five thousand dollars well if the average salary is 100 150k at this organization it's only an incremental three four five percent on top of their salary so the way i look at it is did it make that person three four five times more effective at their job and i think the answer is yes so that's why there's so much token maxing going on and it's also a bottom-up type product you can just get into this product for 20 bucks a month and you know no cio or cto is like oh no you can't spend 20 bucks a month on your corporate card for this technology so when When a bottom up technology hits everybody at the same time, that's what would explain this revenue ramp that we're all having a hard time adjusting to. It applies to every single person. Like who is it impacted by the technology is my question to you, Chamath. In what organization you're working with, with 80, 90, is there a department that says, yeah, the intelligence on demand, not for us. We don't need it. Well, it's less about being dismissive that way. It's more that regulators and other people won't necessarily.
[25:00] be allowed to use it the way you want. Okay, so finance, HIPAA, yeah, there's HR data, you're not allowed to put that to work just yet. What I'm finding is, once you start using this and getting some gains, it's very addictive. And we were sitting here, Brad, I don't know, maybe in January, and I got that open claw bug. And then, you know, I started playing with this Hermes agent, which is not a French company, by the way, they just use, you know, French names, it's nouveau research or whatever it is i started playing with that it's a very peculiar piece of software but it's a very open piece of software so i went to open router i got my own keys i've been playing with glm then i talked a little bit about bit sensor on the program known as tau dollar sign tao it's a it's a crypto project somebody who is creating a subnet that is putting glm 5.2 and other models available at really cheap prices so i all of a sudden experienced because they gave me an API key, having my token cost go down 95%. And when you have unlimited tokens as an exercise, which is going to come to everybody, eventually everybody going to learn how to drop the price by 95 And it going to happen as well because people like Grok with inference this is all inference right This is what people are using They using inference to do this Well, inference is being impacted like three or four different ways. The software is getting better, open source at the same time. You're going to have distributed networks like Tau, and you're going to have better chipsets from Grok and Cerebris, etc. All that's happening at the same time. once i got down to 95 cheaper i started setting my agents instead of doing daily runs to doing hourly runs then i took my agents from doing one task and i broke them up into three agents and have them doing three different things on the hour and when you start doing hourly tasks and then you wake up in the morning and like 14 jobs have been done you're like wait a second this is completely different as one example i have it has all the all-in episodes all the this week in startups episodes, and we set these cron jobs to go find what the new trends are in technology. I have a trend spotting agent running every hour, informing me of the top three or four trends. And I just give it words really does change your thinking when costs go down, what do you think tokens are going to cost Brad and, you know, next year? Yeah we seen we seen 90 reductions in the price of tokens for each of the last two and a half years We talked a lot about Jevons Paradox which I think you referencing here which is you going to use a hell of a lot more when it happens I think the central debate right now in AI is the one that Chamath keeps pointing us back in the direction of, which is for 18 months since the deep seek moment. When the DeepSeek moment happened, the markets fell 40%, and there was a reason for that. Many started arguing that the frontier models were screwed, that open source was going to kill them, that they were closing the intelligence gap, that model routing was going to make it easier and easier to root these tasks to cheap tokens. But despite all of those arguments, and now we're 18 months into this, and I had this back and forth with Gurley a lot. I love open source. I want all the competition in the world. Let's be very clear. But despite all of those arguments, the facts in the field are just the opposite. The share of economic value, right? There's this tweet this week from Jesse Zhang that we ought to pull up here. You know, the economic value, the share of wallet is actually increasing to the Frontier Labs while the share of tokens these commodity tokens is obviously going up you know to the other guys and i had a little back and forth this week with nikesh on this kind of trying to suss out why is that the case right because what people would have thought is oh cheaper pretty damn good 90 is good enough to do all these tasks that you're talking about jason so nobody's going to use the anthropics in the open ais of the world but despite that it looks like their share of wallet I think it's not that. I think it's more that when the iPhone was a novelty, everybody would keep upgrading because you expected that the new price was worth it. And then at some point, there's a moment you can debate when it happened, where people said, you know what, I'm just going to keep the old phone because it's good enough. And I just don't see the difference. And I think that there's going to be a moment like that. Like when I use Fable 5. The problem is that it's nerfed on a bunch of things that I would normally research, you know, I was with somebody this weekend and he was telling me about some helping and I put it into Fable and it's like, it won't answer you. I'm like, okay. So I think that everybody will get to a point. They'll get to it at different times where they just say, you know what, like, it shouldn't really matter what model I'm using. If I get an answer that I think is reasonable and I can kind of go about my day separately, I think when the course
[30:00] CFO gets involved, that'll be an entirely different conversation altogether. I think that what I can tell you after this UN Commission that I joined with Benioff and Jensen and Brad Smith, there is not a single country in the world that is not trying to figure out its own sovereign AI strategy. And I don't think they believe using a closed source American model is the answer. And so, you know, I think we have to keep in mind there's trends. One is just geographic penetration of humans. use it, then do, which is an upside and an opportunity for everybody. And then the second is there is going to be the experimentation, as you said, that needs to transition to ongoing repeatable usage. And then the third is that all of that then needs to plug into the existing regulatory infrastructure that we use as societies to run the world. And I think when you put all of these things together, it's not clear to me who wins, except that you're going to have a lot of diversity of choice. Certain countries, I can tell you after this week, have no desire to subjugate themselves to any technical risk and so they willing to spend the money to have their own Now we can argue and debate whether that country has any chance but they would rather take an open source model like nvidia actually and stand up their own stack soup to nuts for their own people and their own companies inside of their own country and if the models are 99 as good or 95 as good there's going to be a claim that some countries make which is it's just good enough that's the question that's the question and then separately there are companies who will not have the earnings growth to justify this without going on some long protracted carve out of cost and most companies you know this they just don't do it they don't have the nerve to do it they're not capable of it you know you wrote that famous essay to zuck he was pressured into finally doing it absent a very few companies most people just allow the problems to compound so i just don't see a world where when you get clobbered over the head you don't look at other ways of just displacing costs and if it's like coke pepsi kind of a thing and pepsi is one one thousandth the cost of coke i don't know i just think it's a it's a risk that i think has to be managed in the perception of the market participants and the underwriters To add to that Brad just open source is very hard to implement when compared to just firing up claude and having claude already approved in your organization the number of steps it took me in order to and i'm pretty familiar with technology it took me hours to configure my new setup to get onto this bit tensor network to get open router going and to your point it does dynamically route now so i'm dynamically routing and i'm you know glm 5.2 and then if i fall back to claude but which quad am i going to fall back to and here's another piece of evidence to your point shemath there are some organizations that just aren't capable of this they don't have the the team that you know does this naturally we just talked about the cto of uber now let's talk about another cto andy fang is the cto of doordash shout out to stanley and so he as you can see here in this uh tweet a lot of people listening to all in over the last couple weeks are coming out and as i said explaining what they're doing to address this exact issue he says hey with our internal coding benchmarks we're able to confidently introduce open weight models into our ai code review without degrading code quality have the frontier model fable from anthropic to do the hardest work delegate lower level work to kimmy 2 and they are now releasing their benchmark so another group releasing their benchmarks and saying hey we know this is an issue The CTO has been charged to your point Chamath again CFO says hey make sure this is profitable We get the ROI They put that on the CTO Here's another CTO from another leading tech organization that knows how to implement this. Yeah. The really interesting thing we have to forecast right now is what happens in an earnings miss. And I think what happens in a moment where, for whatever reason, maybe there's just an externality, that there are a series of earnings misses. Where are people going to look? And I just think that people find it very difficult to lay off other people. I think it's much, much easier to cut other costs. And I think that the more successful these companies get in a very quick amount of time without really proving the ROI, I just think the bigger the risk is. It's complicated. on the field when you're working with these enterprises and just trying to explain it to them is I think that they're they're getting smarter quickly is what I would say. I would just say I I think that Chamath Stacks absolutely right about the sovereign stacks that are going to get built around the world. This is not either or we are going to have open source and we are going to have frontier intelligence. The preponderance of the tokens today are already shifting.
[35:00] team poured cheaper, lower, uh, uh, lagging models out of open AI or lagging models out of Anthropic or, or the other frontier labs that are out there. Obviously we, you know, we talk about those two space X's released an incredible model, you know, in the last two days met is out with, you know, a, a, a terrific model today. So Gemini is still in the hunt. So there are lots of choice. The meta thing was really intense because I thought, okay, you know, we talked about the game theory, which was Mark should scorch the earth with open source. I think they flubbed that play. But then I think he is now said he's going to create a price for it. And so if you look at the tweet or the quote, there was a post, Jason, I don't know, Nick, if you can find it. I got it. Announcing it. He was basically like, hey, guys, I'm going to give you the same quality at like one one hundredth of the cost. Now, again, there's a lot between here and there. There's a lot of enterprise distribution that's required. And, you know, there's been a couple of misfires before. But I thought it was interesting that the vector of challenge was on cost. Yeah, here's the Mark Zuckerberg tweet. Just so I cured up for you there, Brad. And he at think D That was his old handle back when he was in college F And he done more tweets today Over this new spark announcement then he done in his history So he getting into the x conversation Quote today, we're releasing new spark 1.1 a strong agentic encoding model at a very low price It's available through our new meta model API and in meta AI so he's coming out saying hey We got the strongest agentic tool here. Please come use it he also wants to have his own essentially you know he wants to jump into not the hosting space but he wants to provide tokens as well so again i think we're gonna have a tremendous amount of selection the competition is great for america but i think if you look at the things people are doing let me give you an example the premium workload jason you talked about summarizing a document may take 20 000 cheap tokens to do of course shoot that to a lagging model or an open source model but if you're talking about replacing a software engineer for two hours that may take two million expensive tokens and the consequence of using something that's 95 is good is really high right because you have a long task and if the task breaks early or it breaks in the middle or breaks at the end there a huge cost to that so you still burn the tokens right you and back to this analogy and the time machine and you lose and the time and the compute so if an ai agent is replacing a 200 an hour consultant right take that as an example so three consulting firms they're competing they need the smartest consultant they're charging 200 bucks an hour the difference between spending three bucks on a cheap model or 15 bucks on an expensive model to replace a 200 an hour consultant it's just irrelevance that inference cost difference is irrelevant if you're getting something that's bulletproof for 15 bucks and so i think that's what we're seeing play out the best evidence for all of this is just revenue growth right we can sit here and speculate all day long as to what this company not from anthropic and opening eye but from their customers i know i'm talking about what is anthropics revenue growth compared to open ai compared to the open source models millions of independent actors are choosing every single day the open source companies are growing right but they're growing selling something that is really really cheap and there a there there room in every single market for premium products for mid products and for commodity products and i think we see a lot of this token growth people are speculating that the intelligence gap between that commodity stuff and the frontier stuff is going to collapse to the point that people won't pay for the frontier stuff there is no evidence of that on the field today it may develop over the course of the next couple years but it's not on the field today yeah and just to give people an idea we keep mentioning what sovereigns are doing to give you the specifics on that the uae very famously has their own abu dhabi technology innovation institute shipping falcon you probably have heard about that the saudis have humane and they're doing their own models that are arabic llms and then this week japan is investing six billion dollars in a consortium it's called the neo terra the neo N-E-O-T-R-A consortium, and they're doing that and skipping ahead to physical AI, i.e. robotics. Okay, joining the conversation here, the one, the only, Saxxy Poo. Sax, bringing you into the discussion, talking a little bit here about the debate that we started here on the podcast, getting ROI from tokens, where are the tokens gonna accrue to, open source versus the frontier models, bunch of CTOs chiming in on X this week
[40:00] last couple days in fact talking about how they're managing intelligent routing first to open source models then falling back to fable and the frontier models how do you think this is playing out and if you're an investor in the space how do you think about the frontier models and their growth when you have uh you know cfos coming in and saying hey justify this cost and do you have a cheaper solution and what is that cheaper solution well look i think that enterprise CTOs would like to shift their token consumption to cheaper models for the obvious reason that that would be more efficient. And they are seeing their compute costs or their token costs is skyrocketing right now. So everyone's trying to figure out how do we put the brakes on this or at least control it, make sure we're getting ROI. You also have the AI sovereignty issue that we discussed last week that Alex Karp talked about, where they're worried about giving up the secret sauce or the alpha in their business to a frontier lab that may one day be competing with them. So there's no question that enterprises would like to diversify. They would like to get off of these frontier models when they can The problem is I think in most cases they don have the technical ability to do it I mean Coinbase figured out how to do it DoorDash figured out how to do it which is to say they built a token routing system a layer of middleware that allows them to sort of send frontier tasks to frontier models and non-frontier tasks to more mundane models. But I don't think your average enterprise has the technical capability to do that. So I think this is a case of the spirit is willing, but the flesh is weak. I mean, they are willing, they would like to diversify off of these closed models, but they are unable to do it. And so this is why the share of wallet of closed models, it actually increased. I think that open source went from 19% last year to 11% this year. So open source as a share of enterprise spending is actually decreasing. Now, I don't think that means that usage is decreasing. I think usage is skyrocketing in both these categories. It also may be the case that because the whole point of using an open model is you just pay for the compute costs. You don't have to pay a lab. So it may be the case that it hard to measure that usage in terms of spend But nonetheless I mean Anyone who saying that these closed models are going to lose or are somehow losing you just not seeing it in the data Like Brad saying the revenue is skyrocketing And I think the most you can say is that enterprises that are technically capable would like to gravitate towards hybrid architectures. But at the same time, it takes technical expertise and it is just phenomenally convenient, whether you're a developer or an enterprise, just to go with the frontier labs. And that's why their revenue is skyrocketing. It is the easiest choice. It's the most refined product. Yeah, and there's one other thing here as well. And this was discussed in a really interesting blog post by the founder of Decagon, which is enabling AI-powered customer support for enterprises. and what the founder said is, look, open models are great when you know exactly what you're trying to do. Why? They're smaller, cheaper models, but you have to do post-training. You have to have the data set and you have to know exactly what you're going to use them for. But if you don't know exactly what you're going to use them for, you want the most powerful general intelligence that you can get right So what he said is that for mature use cases yeah you want to go open But for immature use cases which are all the new things people are discovering right now you just going to want to use the most capable general model that you can And then once you figure out what the workflow is and what the workload is going to be and exactly what you're trying to accomplish, then you can use a small, highly trained model. And I think you said something like... You'll optimize in the post to get the game. For customer support, your model doesn't need to know physics, for example. And so you don't need that capability, but enterprises are still trying to figure out exactly what all these workflows are going to do. So I think that's another factor, which is to say that it depends on the use case and how mature that use case is. And you really want the most powerful frontier models that you can at the discovery of all the potential for the technology. And then let me just wrap up. Then there's one other interesting that I saw was by Nikesh Arora, who also said that what he's seeing is, yeah, enterprises would like to diversify. They would like what he called model fungibility. They would love to commoditize these models, right? And just hot swap them. Yeah. Headless is the term being used, right? Yeah. Yeah. That'd be ideal for enterprises is you sort of swap out the model for the cheapest one that gets your task done. But then what do you do about memory? What do you do about content?
[45:00] context? What do you do about history? And what he said is no one's really figured out a way to abstract that stuff away from the model yet. And again, this goes to the technical challenge of creating this middleware layer that would do the most efficient token routing. It's, well, you know, it doesn't work unless you can make all of that context and memory and history fully portable to the cheaper model that you want to basically hot swap to. Which means you have to have some technical ability sacks and the people with technical ability the tip of the spear the one percent of people deploying this technology are starting to figure that out here's another proof point and some more evidence this is ali the founder of databricks i think the company you're very familiar with brad and what he realized when you start taking apart the harness and you start looking at the skills you look at the memory and all this accoutrement that you put around your tasks, he said, we find that the same model, using the same model, not using open source versus open AI or Claude, we found that for the same model, the choice of harness can significantly save costs by about 2x So they found with GLM 5 that this forms extremely well and that their tasks literally are getting cut in half using the same model but with a different harness and that rings true to me i once you built one of these agents and i was talking about one earlier i'm running every hour on the hour to find trends i asked it to please start optimizing it and when i optimized it it was like 80 less token use and And now in these apps, you can go to your analytics sacks and you can actually see your token use by hour, by job and across which models you're using. This is really sophisticated and hard for a consumer to do of the technology, but it's definitely a trend. So with the harness that he was using, is that something they built in-house? Yes, I think it's. Yeah. OK, got it. So they basically create their own application. omni gent yeah in front of these and that it can multiplex different harnesses and models for different tasks so he's not only routing to the right llm he's routing to the right harness and people don't even know what skills are people don't even know what the memory is at this point they're that's all abstracted into the cloud product or the propuxity product etc yeah there'll be a massive business there already is all these inference clouds you know the base tens of the the world the fireworks of the world every single hyperscaler in the world is going to do this They all going to provide you know tools that allow you to achieve some level of model fungibility The big question is, at the end of the day, we're going to have it's going to be very heterogeneous. But what is the mix between these two? I, again, think the TAM is so damn big here that you're going to have huge open source use cases, sovereign use cases, etc. You're going to have plenty of room for the frontier labs. Let me throw something out I'd like to get your opinion on. You know, to a certain extent, there's this implied assumption in the world that there's going to be this convergence of intelligence. Right. And if you look at the benchmarks today, seems like everybody, you know, on the benchmarks, they are converging. But yet, if you look at the revenue distribution, it's not converging at all. One of the questions I have, will the model router itself be smart enough to overcome, David, the inherent intelligence advantages of the generalized of the frontier labs? The non-consensus argument might be that intelligence is not converging at all, that super intelligence becomes fully self-recursive. you know, recursive. And as it becomes recursive, you actually extended the lead Because the smarter your model gets the more revenue you get the more compute you can buy the more you compute you can buy the better the model is that you can build So I think there a chance that over the course of the next two to three years as we take on much more complex agentic tasks that the distance between the frontier and everybody else doesn converge it actually extends We shall see. But, you know, I think there's this implicit assumption in all the arguments today that everything's converging. I'm not I'm not sure that we've really run that to ground. Another piece of evidence to put into this mix, I interviewed Anton, the CEO of Lovable. Lovable is an app or as a service that allows you to vibe code different pieces of software. They've got a really interesting take on that. They went from $100 to $600 million in revenue over the last two years. They went from $0 to $350 million in the first two years of the company. The product's been out, I think, roughly for like 30 months. Then I also spoke to the CEO of Eleven Labs, Mati. And I asked both of them point blank, are you guys, you know, your major customers of, you know, the frontier models? Yes, they're spending tens of millions of dollars with those frontier models. I asked them, hey, are you concerned about data leakage and them, you know, releasing competing products? Eleven Labs is doing voice and obviously cloud code, you know, is an obvious concern.
[50:00] that are lovable. Are you going to make your own models? Both of them said that they're working essentially on their own models. Those are major customers of the Frontier Labs who they want to get off of the Frontier models and they want to have their own proprietary model. The ability to create verticalized models is getting easier and easier every six months or so. So that's going to be another trend to look for is these verticalized models for voice, verticalized models for building code and we're going to see people stop using the Frontier Labs and these are major major eight and nine figure customers I think they're going to just run for the hells and only use the Frontier models. The counterpoint there is 11 labs I love Matty. Yeah. Do you really think he's going to use an inferior model? He's got to have the best voice agent in the world and if the best voice agent in in the world is given to him by using the Frontier Labs. Can he afford in a competitive marketplace to say, I'm gonna use the cheaper version, the thing that I built for myself, even though it's not as good as the other thing? If he builds something better, I totally agree with you. Which is what he believes he's doing. He believes he making a better version So that the question that I just put on the table whether or not you gonna see this convergence whether it in fact is that easy I don think it that easy but we shall see It may come down to how discrete and sort of predictive the use is So like Decagon, the customer support AI company, they said that 90 percent of their usage now is being sent to open models. But those are open models that they've had the opportunity to post train on and do a huge amount of customization based on all of their learnings and all of the data that they've gotten. So, again, it comes back to maturity of the use case. If you know exactly what you're trying to do, it's probably easier. But, you know, I don't think that's most enterprises, though. and maybe there's going to be a pattern where, you know, all the immature use cases, which is to say all the things you're figuring out, you're just going to want to use the most powerful model possible. And then once it gets really well defined, maybe you start moving some of those workloads to post-trained open models. Purpose built models. Yeah. Yeah. Purpose built. Yeah. It could be something like that. That's kind of what's happening with the people who are the tip of the spear. They're working on the routing of the jobs. They working on the harness and they want to be independent and have that AI sovereignty we talked about last week Well but just to take on Brad point for a second I actually agree with what I think you saying Brad which is the market today seems to be pushing towards duopoly or it has become a duopoly, certainly measured in terms of revenue. If you were to look at market share based on token revenue, there's only two companies making meaningful revenue. Anthropics at, what, 60-something billion ARR, open AI at 40 something billion ARR. I don't know if anybody else even registers. And it may be the case that the more tokens that Anthropic and open AI produce, I mean, we've got to remember every token that they're serving up is on behalf of a use case, right? So they themselves are learning from that and they're getting better at them providing whatever offering that is. And so who knows, like the gap may be growing. A year ago, it seemed like we had five major labs, you know, now it seems like there's a top two and then everybody else. So I mean, look, I could see AI easily becoming another tech market that becomes a duopoly, which by the way is the trend, the historical trend is like monopoly or duopoly in most tech categories for better or worse Yeah In this case though however we using revenue as the metric to determine the winner Keep in mind when you doing open source those are dark tokens Those don come up as revenue So we don know the utilization that occurring at DoorDash when they using an open source model We do know their fable and their anthropic spend right And because we see that in the anthropic revenue ramp The more they deploy these things on their own hardware using commoditized hardware using the NeoClouds you don see that it doesn come up as revenue it comes up as free the only thing you're paying for there is the hosting cost you know and that's that will come up on nvidia's balance sheet so the gains you'll see there will be cerebris neoclouds caruso cloud etc so just keep that in mind when we're having this discussion let's talk a little bit back to sovereignty here the ccp said that they might or there's a report out according to reuters reuters generally does a good job of this They dropped a couple of anonymously sourced reports about AI in China, and these were published about 15 minutes apart. The big scoop that CCP officials, Chinese Communist Party, are reportedly considering restricting overseas access to China's top models. So two Chinese regulators met with Alibaba, ByteDance, and Z.ai. They're the ones who are doing GLM 5.2 that we keep referencing. They're discussing limiting access to the top
[55:00] open and closed models outside of China. Why are they doing this? Well, they're making any theft or leaks of AI research a national security offense, and they want to control who can fund Chinese AI labs. And we saw this with Manus, which was a Chinese company, tried to go to Singapore. The CCP pulled those employees from Singapore back to China. And so here is their main concern. The quote is that they're concerned about mythos. Chinese authorities are deeply worried about the potential for mythos to exploit software vulnerabilities and that Washington might deploy a model against Chinese interest. Sacks, last week, I proposed the reverse to you in your previous position as czar of AI. Do you think the United States should be banning those models? Now we have the opposite. China is saying, potentially, according to these reports, allegedly, that they might restrict them. So explain the game on the field here. If you're going to look into what China's thinking, why would they want us to not have those open source models? And how is this chessboard developing? Well, last week I explained why it would be harmful to the U.S. to ban open models. So if you China and you want to harm the U maybe you would want to I mean it does kind of make sense because our companies are benefiting a lot from all this R that they doing Now at the end of the day I think the story is probably a little bit overstated I think there are a few Chinese models that were open source that have gone closed source, but I don't think they're all, I'd be surprised, let's put it that way, if they all went closed. So for example, The number one model in China, as I understand it, is ByteDance's model, which is already closed. That's kind of like their ChatGPT equivalent, and it's always been closed. Then you've got Alibaba's Quen, which was open and now I think is going closed. And Zipu, which has GLM 5.2, which we've talked about a couple weeks ago because it seemed to be catching up to what was then commercially available as the American frontier. at certain tasks. They, I think, are going close too after having been open. And so this is, I think, the tactic is you stay open until you catch the frontier or you get close to it. And then there's a really compelling incentive to go close because you want to capture all the value for yourself. Which by the way is exactly what Sam Altman did famously at OpenAI Not only did they go from a nonprofit to a for they went from open models to closed models So it exactly paralleling what Sam realized three years ago Yeah. I mean, in a way that was what I think Meta's original strategy was, was that Llama was going to be open, but then they actually, they sort of backed away from open a little bit, but it's just kind of an obvious strategy, right? Is that if you want to catch up, you go open because by the way, you're not going to make any meaningful revenue on closed anyway, because you're not close enough to the frontier. So why would anyone buy your product? But if you go open, you get the developer community on your side. And you get utilization. More people use it, which in AI gives you reinforcement learning. Yeah. Well, having spent some time in DC this week and talking with the White House and Treasury, etc on this topic what i can tell you is while there may be some you know debates about regulation of u.s models the one thing there's absolute agreement on is doing everything to stay ahead of china and you know and and the president all the way up to the president very interested how far are we ahead of china what are the things we need to do to stay out of china is a unifying force in washington and the idea that we were going to kind of take our frontier labs off the field off off the playing field letting Chinese open source models run free And on top of that distilling our models I will tell you GLM 5 has watermarks from mythos all over it, right? So we know they were distilling, et cetera. And I think the US government is going to take steps against distillation, which they should do. So I think that China doing this, in some ways, I don't think it hurts the United States. The United States can spin up open source models. We've got reflection spinning one up. Obviously, we've got the good work going on at NVIDIA with their open source models. The labs, I talked to a couple of the frontier labs about open source models. I said, why aren't you guys making open source models? They're like, there's not a lot of demand for it. If there was a lot of demand for it, we would make it. And so I think the US is in a good position. I think this is probably more chess playing by China than actual threats, because it would hurt them a lot more than it would hurt us. Yeah, and then to just back up your point, Sax, about when you're behind, go open, and then once you catch up, tighten things up. That's exactly what they did with Android, right? Google released Android. At a certain point, they were like, in order to use Android in the license, you have to include Google search. You got to use Google Drive. You got to use Chrome. And they started tightening it up. So it's not really an open source project at this point. Just by the way, I think.
[60:00] The absolute best thing that could happen for America in terms of winning the AI race against China is if China somehow sprouted their own doomer community. Yes. We need like a Chinese Yud over there. We got to get their P-Doom up. We got to get their P-Doom up. We need a lot more people over there freaking out about, you know, job loss or RSI or whatever. yeah that'd be the best thing that could ever happen to us is if they start cracking down on their labs in the same way that the doomers want to do over here yeah brad let me just say just one comment i mean look i agree with you that from the president on down everyone wants to win the ai race and in fact that was you know in the big ai policy speech the president gave about one year ago that was the whole thrust of the speech was declaring that we were in an ai race and america I think the big risk is more that, and this would not be at like the top level. You know, I think if the president could make every single decision, it'd be perfect. The issue is at a lower level in the bureaucracy, do people somehow do things that are counterproductive? Maybe they think it gonna help us in the race against China but they end up doing something that ham They just like ban something or without you know really truly understand all the implications of it So I think there no question that the administration wants to win the AI race the president definitely does And at the top levels, they will all make smart decisions. The question is whether at lower levels of the bureaucracy, you can get mistakes being made. And then you have the influence of Congress, whatever they want to do. Those guys, they're more responsive, I think, in a way to like the doomer community that's creating a lot of political pressure right now well and the throttle you know paradoxically to all of this might not be the software might not be the chips it might be energy yeah chamath i mean when you look at your data center projects and the other ones that are going out there if we need more tokens if people need more inference we have a gating factor in the united states which is which is energy there's a an analysis that my team put together which i think is quite staggering. If you just look at the load growth that's expected between now and 2050, we are about three entire California's worth of energy short. And that's just assuming regular consumption of devices and cars fridges televisions and computers So we have an enormous problem in the United States with respect to electrons Yeah and if you put Taiwan into the mix here where the chips are coming out of I had a really big wake call There was a Wall Street Journal article about this. The amount of LNG, which is what Taiwan runs on, is like they have two or three weeks of it. China decides to blockade Taiwan. They're going to run out of energy immediately. So this is energy both in China and Taiwan and in the United States. It's all dependent on that. We have to get nuclear running, more solar running, more batteries, more of everything. And that is obviously a regulatory challenge here in the United States. All right, let's talk about your time in D.C. Brad Gerstner went to D.C., everybody. And huge, huge congrats, Brad. You've been harping on about this, you know, accounts now called Trump accounts, the Invest America accounts. And tell us what happened in D.C. this week, because I think you finally have the number one app in the world. Trump accounts is the number one app in the world. Congratulations. And a bunch of announcements. So what the contours of the announcement And maybe you could take us behind the scenes There it is Trump accounts the official app number one in top downloads Your kids can invest in the future This has been a four-year mission in the making. Thanks to you guys, you were early supporters, backers. We talked about it on here. Founders are crazy. You guys probably looked at what I was working on and thought, you're nuts, you're wasting your time on this. And so, you know, when it got signed into law last year, that's a huge moment in a founder journey that it's like getting your first round of funding, maybe like, okay, we actually, this thing is going to happen. But on July 4th of this year, the app went live, right? So that means millions of accounts got created, the accounts got funded. And to celebrate that and to really kind of take the next step forward, you know, we designed a joint bell ringing first in history between the NYSE and NASDAQ from the Oval Office that was incredible. We had 100 CEOs there, kids there, families that were impacted. And the president really, you know, kind of laid out that this is much bigger than just a program to give a few people some accounts. This is really about making every child a capitalist. In fact, the president suggested that we're going to
[65:00] going to auto create accounts for all 50 million kids or upwards of 70 million kids under the age of 18. So he called on us to get the accounts open faster for more people to have more impact to make sure no child is left behind. Right. Just slow down because a lot of people don't even know what a Trump account is. Just explain what it is and then you should contrast it to like a 529 account and some of these other things. Great. So. As you guys know. The idea was very simple, $1,000 for every child at birth that could compound for their life in a privately owned investment account. So you're born, you get a social security number, and you get an investment account. And if you do that, you start with $1,000, and somebody matches that, and you save $10 a week, that's $50,000 at age 18. And that's invested in the S&P 500? S&P 500. So when these accounts are created, all that money goes into the S&P 500. There's no cost. It's a free account for the lifetime of the recipient. And that was packaged into the Invest America Act which was passed into law a year ago as part of the reconciliation bill So that what actually occurred on july 4th of this year all those accounts were created for all of these kids that the reason the trump account app is number one in the app store because parents started hearing about this and said whoa i need to go download and get this set up for my child we had over a million and a half counts created in the first 24 hours after the launch of this we had over a billion dollars of deposits so I was contributing money into the accounts of my nieces, my nephews, my kids, friends, kids, every account app has a QR code, Jason. So somebody can just send you the QR code for your kid. You double your Apple pay on your phone, double click, and you send them 25 or 50 bucks. So that is kind of the most essential part of it. But we also had a bunch of announcements around And just to be clear, you can get access to that when you're 18, 19, 20 years old and start putting it towards school or you can roll it into your Roth IRA, I guess your retirement account. Obviously, Michael and Susan Dell were the anchors here. It over billion for each of 25 million children primarily lower and middle income kids SpaceX president Gwen Shotwell she joined the party put 350 million in her SpaceX shares and for children of lower income communities. So with this, there's a device or some way to do it. So you can target specific communities by geo or by, I guess, their net worth. No, it's just zip code and age, zip code and age. Okay. And then Micron put in 250 million, up to a thousand per employee. So that seems to be a really interesting way to do this like you can do an employee i'm sorry an employer contribution and brad brad did it for all kids in indiana i think right correct okay this is a big number here this is a big announcement brad the guy who complains when we make him buy in for 10k after 10 p.m at the poker game and who like rage quits the game when he loses six thousand dollars somehow brad dropped a hundred million dollars i mean oh my god let's get a round of applause in a golf club brad this is i mean i i've never heard of you doing any philanthropy like sometimes you show up with a bottle of wine to the game but this is a big number this is a big decision for you huh well i think this will become the largest direct philanthropic platform in the history country We told the president we think we can raise billion in the first 12 months And so the scale of the philanthropy, the nature of the philanthropy, directly to America's kids without a charitable middleman that's directing who gets what and how it's distributed. So you think about the people who now are, you know, we're $10 billion or $100 billion. How do they give that money away at scale effectively now we have a platform they can do that with it goes directly into the accounts the money can't be taken out until the kids are 18 it comes for there was a bunch of noise about how some people won't do it because it's called trump accounts and that some people said you know this is going to create this weird class divide by people who who had TDS and refused to give their kids, because I think your website or something said something like 13 million bucks by the time they're 50. And, you know, I tweeted something to be effective. That is an irresponsible amount of money to not give a kid because you don't like the fact that it's called a Trump account. It's it's it's patently insane. If you go on blue sky, which is like the open source, uh, lib TDS, um, social
[70:00] network people are like just but one you know y'all trust those trump accounts i sure as hell don't and so there's a bunch of people yeah that yeah speak to that for one second mike so i would say this you know it was the enabling legislation is the invest america act a lot of people a lot of democrats call them invest america accounts they're officially trump accounts and the facts on the ground are that parents aren't listening to that noise the parents who are signing up for this are across the income spectrum across the economic spectrum they're across the political spectrum they know and understand that their first responsibility is making sure their kids have a connection to the american dream have savings for their life but yes they are called trump accounts and i i've read some of that blowback but the president himself let me just make this case very strongly there's nobody who i've talked to about this over the last two years who cares more about every child getting an account than the president himself in fact that occupied a lot of a lot of our conversation over lunch he said how do we get more people auto in this faster i want every kid to have the shot to have this i don want anybody being left out and left behind because their parents are too busy working two jobs or because their parents may have an issue with it being called a Trump account. So the president is pushing us very hard and the treasury secretary. And he told you, Brad, he gave you instructions. He gave you an order that he wants you to auto create the accounts. This is a brilliant move. He know, we know the social security numbers of people who are under 18 he told you get to work and automatically create the accounts are you going to do what the president trump has commanded you to do brad are you going to auto create them brad are you going to disobey the president we our intention is to get all 50 to 70 million accounts created over the course of the next 90 days using all of this data but you know listen we got to work through treasury the white house social and Bernie Sanders and Ro Khanna who are going to try to stop you. Are they going to try to stop you from doing this? Are they giving you blowback because they don't want to give Trump the win, which is totally retarded, but. No, listen, I'll give credit where credit is due. You know, Cory Booker's come out and supported these and Gavin Newsom, Governor Wes Moore, you know, John Fetterman, Senator from Pennsylvania. So there are plenty of Democrats who are able to get over that hurdle But you bring up a good point And I said this on CNBC yesterday On the one hand, you have Bernie and Mondami. They want to take and tax all these corporations. They want to control all that money in Washington and decide who gets it. Right. It's a very dependent on Washington model. on the other side you have the president in this administration and frankly a lot of democrats who are more in the orthodoxy closer to the center who say no let's set up a private account for every kid in america let's fund them let's not make them dependent let's make them independent of the government to build wealth on their own financial literacy on their own more likely to graduate from high school start a business buy a home those are two very different worldviews for America and I think the antidote to more socialism is more capitalism. And as I told the president, this is more capitalism. Sachs, if this succeeds and, um, Brad does, as he's been instructed by the president, we're going to go from 50% of people owning equities in the country to as much as 70 maybe even 75 of the country having access And for the first time being part of equity nation what your thoughts on this sax look i think that a great thing and i think that this is a tremendous new philanthropic platform and that's really important especially in this time of growing anger and backlash and populism against billionaires and people questioning whether the system is rigged and whether they can be successful in america whether they will be able to be part of it this is a really important antidote to that. But I almost think that the philanthropic aspect maybe has gotten almost too much attention because people are naturally attracted to the freebies. And the part that I think hasn't gotten enough attention are all the comments I saw on CPA Twitter, where all these accounts were talking about what an unbelievable, I guess you could say, estate planning strategy this is, or basically- Tax advantaged. Like a wealth management technique, whatever you want to call it, like planning for the future. And there's never been anything like this before. They were basically saying, this is like in the top three, you know, there's certain things that you just have to do. Like if your employer offers a matching 401k, you have to do it because otherwise it's just, you're losing out on free money. And if you don't do a health savings account or you don't max out your Roth IRA, there's just certain things you have to do.
[75:00] because they're so tax advantaged, or you're getting free money, right? And in this case, you're getting both. There is the opportunity for, frankly, the free money for your kids, right? But also the tax advantage is huge. So let's just go through this, and Brad, correct me if I get any of this wrong. So you can donate up to $5,000 a year to your kid, as long as they're under 18. And it's not just you, it's any friends and family or others can contribute as well, which is new. and then they get tax-free compounding until they're 18, and your employer can contribute up to $2,500 tax-free. So at a minimum, you should go to your employer and say, sign up for this, and if you have to, take $2,500 out of my salary and make it a donation to my kid's Trump account because then that's a huge tax savings, right? Neither side has to pay tax on it. So like Brad said, this is basically like an IRA. You get tax-free compounding. Then when the kid turns 18 they can get access to it and they can do a rollover into an IRA or into a Roth IRA which is even better because when the Roth IRA matures you don pay tax on the money that gets distributed out of it Whereas with a traditional IRA all the taxes get deferred until the end The difference is that when you do a IRA to a Roth IRA conversion, you're supposed to pay taxes at that point. And I saw one really clever CPA say that, well, the best way to do this is wait till your kid is actually not a dependent anymore. Like, so maybe they're in college or they just graduated from college and they're in like the 0% tax bracket because they're not making any money and then do the conversion. And so you'll be able to convert the Trump account very cheaply into a Roth IRA. And now they're going to have two to $300,000 potentially in that account that they can then do tax-free investing for the rest of their life, or they could potentially start a company with that. Other things you can do with an IRA is you can use part of the money on a down payment for the first home purchase. Or if you get into a health emergency, you can use money for that. So there's all these things you're allowed to distribute money out of an IRA without incurring a penalty. But generally speaking, the point of an IRA is to save for retirement. And this is where I think it gets really amazing is because if you start with two to at age 18 you be at million plus by age 60 If you just let it compound I mean there ranges because There a lot of nepo babies we we creating here we gonna have a lot of rich kids with trust funds yeah so this is like this is all you have to do to make sure that your kid is protected for retirement is if you and your family and your friends and your employer can just contribute to their trump accounts yeah like that that's like i have way better than social security right i have an idea i have to go and sell some enterprise software so i have to leave but i'm really proud of you i think this is incredible you should convince open ai and anthropic to give the equity of those companies if this is going to be as big as you say 100 billion 300 billion zillion trillion put it into the uh accounts of every kid can you just explain how that works against the five thousand dollar i gotta go love you guys see you later all right good luck on the sales call yeah so always be closing jamoth always be Chamath closing, ABC. Chamath always be closing. You know, again, David and Chamath, as we build out the platform at scale, so imagine now you have 50 million accounts that are opened. I've said on CNBC, you know, I've obviously talked with Dario and Sam and Elon and others about making those donations I don like this idea of shaking down our companies taking their shares and then putting them in some government slush fund that perhaps Bernie or AOC or somebody is going to control in the future I said it got to be voluntary and number two it should go into citizen accounts, right? Privately held in citizen accounts at compound for their life. And so you asked the question, David, how does it happen given the limits that you have, you know, the $5,000 per child. That's why you have to have so, you know, 50 million accounts open, right? Because then you can take dollars in at scale. But we can also set up a pooled account, David, where it can be distributed over time. So you distribute it to all the kids subject to the limits that you have today. And then any remainder you can distribute to the three and a half million kids that are going to be born next year, or the three and a half million kids born the year after or the three and a half million kids born after that. We are on a trajectory now that we're gonna have over a hundred million of these counts set up over the next decade, okay? So we could have 70 million today and then you're gonna add 3.7 million a year. So you're gonna be at a hundred million private, individual accounts that are compounding for people's lives that anybody can donate money into. The people themselves, I think one of the things that gets
[80:00] As moms and dads or somebody working their summer job, put in 10 bucks, put in 100 bucks, put it, you know, into these accounts. They get to see it on their phone. You know, this is when we started this, my sons and I designed this app. And it's basically what we ended up with. It's a, you know, Joe. Shout out to Vlad at Robinhood helped you with it. Yeah. Vlad and Joe have implemented a more elegant version of this. But every kid owns a little bit of a video, a little bit of Microsoft, a little bit of Apple. Imagine opening up that account, David, in middle school or high school to the money page. And now you're getting excited that you're seeing, oh, man, I'm in the game. I have ownership. And while you reference what it could be for families who can contribute $5,000, obviously, Michael Dell and I and Gwen and everybody else were focused on the 50% of Americans who feel left out and left behind who would otherwise have zero. If you do the math on this, over the course of the next 15 years, you could have somewhere between two and four trillion dollars added to the accounts of families and kids who would have otherwise had zero. I just want to go ahead We talk a lot about philanthropy piece Basically the way the philanthropic aspect works is that other people philanthropists can contribute towards that per kid Right. So when Gwen Shotwell contributes 2 million shares of SpaceX to 2 million kids, each kid's getting a share of stock that's worth it. $150. $150. $150. So now that's counting against their $5,000 limit. But so that's what makes it compelling is, OK, look, every every family that can afford to do the 5000 should because it's just like so compelling from a tax and saving standpoint. But then even for families who can't, they're going to be beneficiaries of philanthropists who just want to give this type of direct giving. And it seems to me this is so much more efficient and so much better than the whole NGO industrial complex where. Absolutely, where they take 40% for their offices and their salaries. Yeah, they're just drifting. Yeah. One of the numbers I saw was kind of amazing is, again, this goes back to the power of compounding, is that if a Trump account had been maxed out and you have the standard market rate of return that we had for say the past 30 years then by age 28 that kid will be a millionaire Incredible That right That right All the numbers you hear me quote the and the it doesn assume maxing You know that just assumes people are adding $50 a month because I've been focused as Michael and others have really on the families who don't have the capacity to save today. We're getting all of them into the game. And the president directed us. He said, listen, we have 529 accounts that already helped the top 10%. that's not who we're focused on this is about the main street agenda this is about all the families that he ran for to feel left out and left behind and we're reconnecting them to the american dream through uh universal ownership they all have their own account they all have a private account on their phone um it's a game changer for the country it's the largest change to our social contract since 1935 and social security and importantly i think it couldn't come at a better time you know We have this fight for the sole of- Also, just every employer should be signed up to be an employer that can contribute. Because again, you could take that $2,500 and hopefully look, it's additive and it's not just a substitute. But even if it's just a substitute and JCal, your employer takes 2 bucks out of your salary and puts it in your kid Trump account then it reducing your taxable income So it a no brainer If you a profitable company your employees are going to love you Yeah Yeah But I think every employee is going to want to and every employer should do it because it's a tax savings for both. Right. So let me the tax savings here is huge. Yeah. So just off of the mechanics of it i just want to maybe level up here for a second yeah i have been often critical i call balls and strikes and i can tell you in detail the things that i have a problem with this administration and their actions they've done and i have done it here on the pod i want to address the people who are nagging this and specifically nagging it because it has the name trump accounts on which i told you at the poker game go up trump accounts i don't know if that was like an obvious thing or i was the person who told you to do it i'm not taking any credit here but i remember I remember that conversation where I was like, just call them Trump accounts. Like, if whatever criticism you have of Trump, however valid you may feel it is, this has nothing to do with Donald Trump and how you feel about him. Put your TDS on the side, put your valid criticisms on the side. In this country, we have a K-shaped recovery going on. We have immense tension between the haves and the have-nots. To the point at which people actually believe
[85:00] Socialism and communism is a better operating system than the best operating system humanity has ever created, which is called democracy plus capitalism, right? And kids love, love capitalism. They love building businesses. But we are in an existential moment right now. If these kids believe, young kids, and there's a couple of generations of them right now who do not believe in America anymore, while we're sitting here on the 250th anniversary of this amazing experiment known as America, this is the most American thing you can do do so put aside your tds put aside your valid criticisms and embrace this and give the flowers to brad to the people donating like michael and susan dell and gwen this is beautiful this is the most beautiful gift i've ever seen to a country and this could be something that's a unifying principle that brings us back together as a country that everybody gets to participate in capitalism. And this is the number one way to do it, which is to let kids on their smartphone, instead of saying, you know what, Mondavi's right. I should get a free bus ride. I should get a free I should get free pizza And we should take Ken Griffin and seize his you know penthouse is a payout of terror F all that You know and we have we have CEOs getting shot and their homes firebombed Well, you know what? If you're one of those CEOs, you've done incredibly well. There was something called the Giving Pledge where they pushed, you know, affluent people at the TED conference for decades, Bill Gates and everybody. Warren Buffett, everybody was pushing for this. this is like the perfect version of the giving pledge because you're not just saying i'm giving away my wealth by the time i die you're very strategically saying every single person in america gets to be part of the best part of america which is entrepreneurship and everybody will be part of the equity nation and my final point is one of the happiest countries in the world is australia if you've ever gone to australia everybody feels safe and we have a large number of people in this country who do not feel safe and the reason they don't feel safe is because they don't think their kids are safe to the point at which people do not want to have kids in this country because they feel the system is just too hard this could change that if people feel hey kids have a shot and i don't have to worry about my kids i worry about my kids and i'm affluent i can imagine being a single parent and what anxiety you must have as a single mother or father and you and you making minimum wage and you behind you know the eight ball for your entire life and now your kids are set that all people want that the only thing a parent wants is to make sure their kids have a better future that was the promise of this country and somehow it went off the rails for the last two generations this puts it back on the rails this is super annuation funds in australia in australia people are extremely happy the reason they're happy is they're forced to put 14k a year or whatever it is 12 or 14 percent i think of their income into essentially a 401k that they get to direct to a certain extent it's forced savings this does the same thing at a very basic level this could replace social security this replaces the giving pledge and i just want to say brad you know a lot of my friends got involved in politics some of them on this very program a lot of people it's very divisive you threaded the needle here it was This is a master class in balancing these two crazy parties and the divisiveness in this country. I just want to give you as your friend your flowers. This is just absolutely outstanding what you did. You have been incredibly humble in your approach This would not have happened without you Brad This is your legacy of everything you done in your life lots of success and i seen it up close and personal this is a million times x everything you've done in your whole life you'll be remembered for this this is architecture as big as social security i mean it's a new bigger it's an entirely new platform it's not because it's philanthropy but it's also retirement savings i mean right and everything in between. And this is not static. This is dynamic. We can add to this. There could be other features. I'm seeing a lot of people in the comments say, why stop at age 18? Why can't you? I mean, you have the Trump account rollover into an IRA or Roth IRA after age 18, but why can't you keep it going? And then people can keep that $5,000 contribution going. And it doesn't mean we take away retirement benefits that are owed to current social security recipients. But sunsetted. But at a certain point, you could just say that, hey, the next generation is going to be on this platform rather than the old one. And it would be a lot better. A lot more efficient than government not running it, right, Saxton? Yeah. We don't want the government running this. Let me ask you a question about this, Brad, because I do see one thing that people say, which is what if your kid turns 18?
[90:00] and then they just want to blow the money. How do you trust that they're going to put it to good use as opposed to, I don't know, YOLO and whatever? As you know, these things are always political trade-offs and balances. I wanted them to have to compound until they were 30, because I figured by 30, you were a little bit more mentally developed on financial issues. But, you know, the argument ultimately came, you're old enough to vote, you're old enough to fight a war. If by 18, we don't allow you to have control of your own money. So that's where a political, you know, consensus was built, David. But the other thing, remember, is they can only take up to 25% out to buy a home, start a business, go to college, the rest rolls into an IRA, and there are built in penalties for early withdrawal in an IRA. So they're disincentives for people to pull out. But let's be clear, we have to do a much better job in our education, our public education system, leveraging this as the platform. You know, if a kid doesn't have any money, it's hard to get excited about learning about money. But if a kid in the game and has 12 bucks in the seventh grade now you got my attention I own a little bit of Nike I own a little bit of apple i own a little bit of nvidia let talk about that how did it get there how did it compound if i added 50 bucks a month what does it turn into all of these things will now be present on every child phone in america 37 states require financial literacy every state should build this into the curriculum we're working with a lot of states on that you know we haven't talked about that we have about 25 states who are going to add money into the accounts of the kids in their states. So you've got a lot of- The states are going to do it. States, state action, Oklahoma, West Virginia, you know, Indiana, et cetera. So that's also sweeping the country where the states are looking at programs they already have where they're spending money for kids that are ineffective. And they're saying, instead of continuing to spend money on these things that aren't working why not just block grant the money directly to the kids because we got the middleman yeah if you give the kids the money more likely to graduate more likely to you know to buy a home start a business etc so i think that we're as i said in in the oval this is day one and i am fully committed to the next decade as is my partner in crime on this michael and Susan Dell I appreciate your guys comments on the legacy of this It has been the most profound work and kind of honor of my life And standing in the Oval Office with my two sons, who are really my two co-founders on this, we made the sketch of this at our kitchen table in the fall of 2020. That's where the conversation started. Lincoln's been with me in every single meeting with every congressman senator president former president etc The president shouted him out, you know again when we were there that journey as a father with my kids I mean has been my the payback has been really extraordinary The other thing that's notable here and again like you got to call balls and strikes and give credit where credit is due Joe Jebbia joined this administration a lot of people in the tech industry are like, oh, you know You joined the trump administration, whatever. Yes, okay. There's some criticism he's an incredible world-class designer i was talking to producer nick our producer here also happens to show the same last name as me he signed up for this right he's doing well but you know his wife signed up for it the software is fantastic let's pause for a second the american government has made exceptional software and this all got done in trump first 18 months immense credit for this this is like of all the you know challenges this presidency has had and ran war and other issues we made great software the american government because of joe jebbia makes kick-ass software like just also major shout out to him and you know he could be doing whatever he wants he's done incredibly well as the co-founder of airbnb and he's doing this like that that's a real patriotic thing to do. I'm just over the moon with this. I think it's fantastic. The dream team. You had like Michael Dell, myself, Vlad Tennis, Joe Gabia, the treasury secretary, Luke Pettit at the treasury talking basically every day for the last year. And our objective was not we want to build the best thing that the government's ever launched. We wanted to build one of the best consumer products, period, that's ever been launched. Mission accomplished. And there's any Silicon Valley consumer company would be thrilled with the numbers that we're seeing, the ratings that we're seeing, the engagement that we're seeing. So, you know, it's been fun doing it with that incredible team. And, you know, everybody's in this for the right mission as well. And so I agree with you, Jason. It's rare that government recruits or embraces, you know, that mission.
[95:00] And I think the tent's getting a heck of a lot bigger. It is bipartisan. It's bipartisan in support. You know, this is important. What Governor Moore from Maryland said yesterday, he said Republicans and Democrats have been trying to do something that looks like this, feels like this for 40 years. this guy got it done give him credit this is great for america great for our kids and especially on the 250th anniversary you know you're standing in the oval office looking at the original declaration of independence and you realize what people laid down for this experiment and then i i read this chatter in my feeds about mandami and others literally wanting to set this great experiment on fire yeah right wanting to burn the place down because they they think they have a better formula no the answer is evolving and doubling down on the formula that has worked for 250 years and rather than making all these kids socialists we get them all into the game of capitalism they become owners owners and shareholders in america so that's what we got accomplished and And appreciate the chance to talk about it Listen incredible Brad and um you know just amazing to watch you do it and if you if you if you really think about it the tech industry it needs to win and capitalists and creators and the makers the people who build stuff this is our chance to say here's an example of something that helps the people at the bottom right and i talked a bunch about um the minimum wage here we got seven dollar minimum wage like we need to address that as well these are the things that if we address what people are scared about what people who are who don't have what we all have here uh if we have empathy for those people and we actually care about them when we give them a path to believe in the american dream they will take that path 100 we have to give them a better path than these socialist lunatics and this is so much of a better Let's do it again. Let's do another one of these things. Let's keep growing this spirit of getting everybody in the country to have equity in these great companies. That's why the entire world is trying to replicate what we do here. Every single country I go to wants to recreate Silicon Valley. We have it. We're not stopping here. And David to your point whether it the AI companies or frankly whether it the Intel shares or whether it the TikTok fee I now have a place where all of those wins achieved by this administration can go They ought to go directly to all the citizens, the country into their accounts and compound for a lifetime. And then as far as people over the age of 18, there's certainly a lot of talk about that as well. So not again as Social Security is a sacred promise by both parties. Nobody's going to change that. But there's a huge opportunity to have a supplement here. Why shouldn't people, you know, between 20 and 30 or 20 and 40 also have a Trump account that they can begin on a supplemental basis, adding dollars that they own and control? Remember, the big difference between this and Social Security, Social Security takes 12.4% of my W-2 income and puts it into something akin to the black hole of government. I don't own it. I don't control it. If I ask a room of 3000 people how much they've contributed, they have no idea. If I die, I don't have title. It doesn't pass to my heirs, etc. So it's really not mine. But if we created a supplemental IRA, it doesn't even require new legislation. I don't think it's just expanding the age. These these are IRAs after the age of 18 Right Then people could start building supplemental wealth and participate in these gains And so there a lot of conversations going on about that as well And so we're not done. But it was a hell of a milestone on the 250th birthday of America to ring the bell in the oval and to watch all these kids get their accounts lit up. It was pretty special. What I think is really cool about the Trump accounts, I think it's an amazing philanthropic platform. But in addition to that, it's an amazing platform for middle class family planning. That's the point I'm trying to make is all the CPAs that I'm seeing talking about this are saying, this is like one of the greatest things ever. And if I could only tell my clients to do one thing, this would be the one thing. Finally, something for the middle class, right? This is what people have been asking for. Hey, let's get something for the middle class. Let's get something for the, you know, yeah. And I think the market gap, correct me if I'm wrong, but basically the the market gap that was created here is that you can't get an IRA, which is basically a tax advantage, a savings account until you have your first job, right? And get earnings, which would be for most people at age 22 plus. So for that first 22 years, your kids can't have an IRA, right? Correct. And you're effectively giving every child,
[100:00] Earth, IRA. You know, it's a Trump account that has certain rules. It's actually better than an IRA. Better than an IRA. It's better than an IRA because with an IRA, your employer can't contribute $2,500 tax-free, can they? I don't think so. And philanthropists can't contribute to it and moms and dads. But the most important thing, you know, like I say this, you know, Buffett's like the secret is to find a really small snowball on a really long hill. Okay? But the size of the hill in this country, we've cut off the first third of the hill forever. Nobody saves anything until they're 25. The easiest compounding in the world is between zero and 25. All right. Listen, this has been amazing. Yeah, because they're dependents, frankly. They're still on mom and dad's stuff. Exactly. So you pick up the first third of life. Incredible. Right? In compounding. And so it a it really remarkable Let go I mean it an amazing job All right We see you all next time Bye We let your winners ride Rain Man David Sack. I'm going all in. We open source it to the fans, and they've just gone crazy with it. Love you, Wesley. I'm the queen of Kinoa. I'm going all in. Let your winners ride. I'm going all in. Besties are gone That's my dog taking an ocean Your driveway sex Oh man My habitasher will meet me at once We should all just get a room and just have one big huge orgy Because they're all just useless It's like this sexual tension But we just need to release them out Wet your feet Wet your feet We need to get merch I'm doing all this I'm going all in
