Why a16z Launched the Machine Age Fund | Jen Kha

Jen Kha, Theo Jaffee, Sophia Dew

AI-assisted summary of Why a16z Launched the Machine Age Fund | Jen Kha — original source. Check the original for full context.

Source published: 2026-08-30

Key ideas

Andreessen Horowitz (a16z) closed its dedicated $1.1 billion Machine Age Fund in roughly two months over the summer, targeting early-stage physical AI infrastructure. (1:58)
Hardware and physical compute proposals have surged from near zero to over 20% of inbound pitch volume at a16z, reversing decades of venture capital prioritizing pure software. (0:00)
The fund focuses on early inflection points, writing $25 million to $35 million checks at Seed and Series A stages across custom silicon, networking, cooling, and data center architecture. (5:00)
A critical domestic physical bottleneck looms as future semiconductor generations transition from AC to DC power, while under 2% of U.S. electricians are trained to handle large-scale DC installations. (15:00)

Detailed notes

## Background Published on August 30, 2026, by hosts Theo Jaffee and Sophia Dew alongside Jen Kha, this episode features Andreessen Horowitz (a16z) Managing Partner and Head of Global Partnerships Jen Kha discussing the firm's new $1.1 billion Machine Age Fund. The show notes outline why the physical computing stack—including chips, networking, memory, cooling, and data centers—is becoming investable again as surging AI demand strains legacy infrastructure.

## TL;DR In this discussion, Jen Kha explains the strategic rationale behind a16z carving out a dedicated $1.1 billion vehicle to back early-stage physical AI infrastructure, custom silicon, and data center innovations away from software-fund competition. Driven by soaring compute demands from AI agents and an aggressive sovereign adoption race across countries like South Korea and El Salvador, physical bottlenecks have made hardware pitches surge to over 20% of inbound deal flow. The fund targets Seed and Series A checks of $25 million to $35 million, focusing on experienced systems veterans from incumbents like Arista, Intel, and VMware to solve impending hurdles such as the transition to direct current (DC) power.

## Atmosphere & style The conversation carries an energetic, collegial, and insider-tech tone, marked by rapid-fire bantering between Kha and the hosts over venture clichés, fundraising blitzes, and the sheer momentum of modern AI infrastructure. Listening feels like sitting in on an executive venture debrief where candid humor—comparing semiconductor CEOs to pop stars—blends seamlessly with high-stakes geopolitical analysis. The dialogue delivers a distinct sense of historical return, capturing the palpable excitement of seasoned systems builders trading enterprise SaaS abstraction for the tangible silicon roots that originally built Silicon Valley.

## Bulletin - `01:58` Andreessen Horowitz (a16z) closed its dedicated $1.1 billion Machine Age Fund in roughly two months over the summer, targeting early-stage physical AI infrastructure. - `00:00` Hardware and physical compute proposals have surged from near zero to over 20% of inbound pitch volume at a16z, reversing decades of venture capital prioritizing pure software. - `05:00` The fund focuses on early inflection points, writing $25 million to $35 million checks at Seed and Series A stages across custom silicon, networking, cooling, and data center architecture. - `15:00` A critical domestic physical bottleneck looms as future semiconductor generations transition from AC to DC power, while under 2% of U.S. electricians are trained to handle large-scale DC installations.

## Key takeaways - `01:58` Legacy infrastructure designed over the past 30 years for web and enterprise SaaS is inadequate for compute- and math-intensive AI workloads. - `01:58` Establishing an independent $1.1 billion hardware vehicle prevents early, capital-intensive infrastructure bets from being crowded out by high-margin software deals during internal committee reviews. - `05:00` Institutional LPs are aggressively seeking private-market exposure across the physical AI hardware supply chain rather than relying on public markets or slow-growth private equity. - `05:00` Rather than attempting to match hyperscalers' late-stage trillion-dollar capex buildouts, the Machine Age Fund aims for high early-stage equity ownership at Seed and Series A. - `10:00` AI infrastructure has evolved into a matter of national sovereignty, with foreign governments treating AI access as foundational public utilities and economic drivers. - `10:00` Sovereign nations such as South Korea and El Salvador are moving aggressively to integrate AI across civic life, contrasting with domestic U.S. political friction and local data center resistance. - `15:00` Next-generation data centers like Switch are being engineered by technologists to address environmental and operational constraints, including liquid cooling and returning power to the grid. - `15:00` The Machine Age Fund's mandate targets everything below the software application layer in modern data centers, leaving raw power generation to a16z's American Dynamism practice. - `15:00` Demand for compute continues to accelerate exponentially as autonomous agents outpace human traffic on the internet and consume five times the token volume of human users. - `20:00` Founder profiles in physical AI infrastructure skew heavily toward 20- to 30-year veteran systems engineers spinning out of incumbents like Arista, VMware, and Intel. - `20:00` Technical diligence on hardware architectures relies on an experienced, finite network of operators and partners capable of assessing deep systems engineering and hyperscaler procurement dynamics. - `23:05` The resurgence of hardware investing marks a cyclical return to the original thesis that founded venture capital and gave Silicon Valley its name.

## Detailed walkthrough ### `00:00` The Global AI Acceleration and the Physical Bottleneck

Around the world, nation-states are rushing to accelerate AI adoption, treating artificial intelligence not merely as commercial software, but as foundational national infrastructure. South Korea recently made headlines by announcing plans to provide premium AI access to every citizen as a public utility. Similarly, El Salvador has rolled out Grok across its public school system for free and begun deploying AI-assisted doctors. These international deployments highlight how foreign governments are moving aggressively to adopt the technology, creating a sharp contrast with the mounting regulatory friction, public skepticism, and local resistance facing data center construction inside the United States.

```mermaid flowchart TD accTitle: The AI Computing Stack Breakdown accDescr: Diagram illustrating where the Machine Age Fund operates below the software and applications layers. subgraph SoftwareLayers["Software Layers (Historically Investable)"] direction TB Apps["Applications Fund (B2B & B2C SaaS)"] Infra["Infra Fund (Developer Tools & Platforms)"] end subgraph PhysicalLayers["Physical Compute Stack (Machine Age Fund: $1.1B)"] direction TB Silicon["Chips, Custom Silicon & Accelerators"] Fabric["High-Performance Networking & Racks"] Power["Data Centers, Cooling, Energy & Robotics"] end SoftwareLayers -->|"Constrained by physical bottlenecks"| PhysicalLayers ```

For decades, venture capital drifted away from hardware to pursue pure software and SaaS economics. Now, artificial intelligence is reversing that dynamic. As Andreessen Horowitz (a16z) introduces its new $1.1 billion Machine Age Fund, hosts Theo Jaffee and Sophia Dew sit down with Jen Kha, Managing Partner and Head of Global Partnerships at a16z, to unpack why the physical substrates supporting compute—custom silicon, memory, networking fabrics, cooling systems, and data center real estate—have become the most urgent frontiers for startup founders. Hardware, once representing a minor fraction of the deal pitches arriving at the firm, now accounts for more than 20% of inbound activity.

- **Sovereign and global adoption examples**: South Korea offering AI as a public utility to its citizens; El Salvador integrating Grok into schools and trialing AI medical applications. - **Venture capital landscape**: Hardware has surged from an afterthought to over 20% of the pitches evaluated at a16z. - **Fund overview**: a16z’s Machine Age Fund is capitalized at $1.1 billion to finance physical AI infrastructure, custom silicon, data centers, robotics, and energy systems.

---

`01:58` Launching a $1.1 Billion Fund in the Dead of Summer

Announcing a major $1.1 billion fund on August 28th defies traditional venture timing. As Jen Kha jokes, the classic Wall Street maxim is to “sell in May and go away,” with late August normally seeing the financial world completely offline. The fact that the Machine Age Fund came together and was announced during the summer speaks to the sheer momentum and urgency of the current technological cycle. While a16z already maintains dedicated investment pools for software—such as the Apps Fund targeting B2B and B2C products, and the Infra Fund targeting developer software platforms—the Machine Age Fund targets the layer beneath both: the physical foundation of modern computing.

Kha explains that the infrastructure powering tech today was built out over the last 30 years to serve the internet and enterprise SaaS. Because AI workloads are exponentially more compute- and math-intensive, relying on legacy architecture is like operating on a "poor man's version" that cannot keep up. As Kha frames the shift:

> “Turns out AI has solved software, right? Any software need you have today, AI can actually do it. But we have to solve for all the physical constraints in order for AI to actually solve for software.”

Establishing an independent vehicle allows a16z to broadcast a clear signal to hardware founders while solving internal portfolio-construction dilemmas. Hardware startups require significant upfront capital to get off the ground. When hardware pitches are forced to compete head-to-head within a general software or infrastructure fund against low-risk software companies producing rapid cash flow, investment committees inevitably gravitate toward the safer software option. A segregated vehicle protects early hardware bets from this bias:

> “Typically what happens is if you put it like for like against another, let's just say, infra deal or apps deal, for example, there's so much in the way of investment you have to do ahead of time that if you do the like for like, you're almost always going to bias towards that sure thing with infra, what it's doing, you know, a billion, two billion in revenue, like off the bat and off the cuff.”

- **Fund alignment**: Targets physical constructs below developer software—silicon, racks, cooling, data centers, and power. - **Venture philosophy**: Echoing partner Chris Dixon's concept of “following the nerd energy,” the firm tracks what technical builders spend their nights and weekends pursuing. - **Incentive design**: Separating hardware capital prevents early capital-intensive hardware bets from being crowded out by high-margin software deals.

---

`05:00` LP Hunger, Public Market Mania, and Early-Stage Ownership

Institutional limited partners (LPs) are aggressively reallocating capital to capture the hardware wave. Jen Kha notes that a16z raised over 23% of all venture capital globally this year, closing the Machine Age Fund in roughly two months over the summer—a speed that would have been unthinkable earlier in her career. Much of this appetite stems from LP realization that value creation in AI has been heavily concentrated in private markets rather than the public sphere. Concurrently, public supply-chain winners have seen dramatic price appreciation; Kha jokes that Sanjay Mehrotra, CEO of Micron Technology, has practically "become the Taylor Swift of the industry" alongside semiconductor stalwarts like SK Hynix and Samsung.

Entrepreneurs and investors recognize that shoehorning AI workloads into legacy architectures is fundamentally inefficient. If one were to design compute infrastructure from a clean sheet of paper, it would look completely different from today's patchwork. Furthermore, with anticipated public liquidity from generation-defining companies like SpaceX, Anthropic, and OpenAI, institutional allocators are eager to ensure they have private-market exposure across the entire hardware supply chain rather than being tied to slow-growth private equity.

When Theo Jaffee points out that a $1.1 billion fund is a drop in the ocean compared to the roughly $1 trillion in capital expenditures being spent globally by hyperscalers, Kha clarifies the fund's specific niche. The fund is not built to fund late-stage hyperscaler data center buildouts, but to take early, high-conviction stakes:

> “So what we want to invest in, so we're early stage investors. So we want to come in at the seed, you know, Series A, and then kind of invest at the inflection point.”

By deploying initial checks between $25 million and $35 million at the Seed and Series A stages, the Machine Age Fund can secure substantial equity stakes in foundational companies before valuations explode. While a16z's growth vehicles can write $65 million checks into later inflection points—such as Series B high-performance networking provider NextHop—the Machine Age Fund is geared to back ambitious, ground-up hardware startups like Unconventional, founded by former Databricks executive Naveen Rao.

- **Capital concentration**: a16z accounted for more than 23% of total venture capital dollars raised globally in the year. - **Fundraising velocity**: The $1.1 billion vehicle closed in approximately two months over the summer period. - **Investment approach**: Targets $25 million to $35 million checks at Seed and Series A for maximum ownership, reserving growth funds for later expansion rounds.

`10:00` Geopolitics, Sovereign AI, and Global Partnerships Beyond Capital

Hardware startups targeting foundational redesigns—such as starting from a clean slate to build chips specifically for AI—often pursue radically simplified architectures to push performance, retreating into a metaphorical “cave” to execute their engineering milestones. Jen Kha highlights that Andreessen Horowitz (a16z) entered at the Seed stage for one such company before it raised a massive follow-on round, underscoring the imperative of writing early checks to maximize ownership before valuation inflection points occur.

Sophia Dew asks what hardware companies specifically need today beyond standard financial backing, such as market access or navigation of complex geopolitical relationships. Jen responds that hardware and AI have escalated to the national level:

> “And as a byproduct of that, we are now a bigger and bigger, technology is a bigger and bigger part of the economy. And as such, countries are viewing on boarded into the AI age.”

To illustrate this shift, Jen notes that when the President of South Korea visited the United States a month earlier, their sole stop was Silicon Valley. With NVIDIA reaching a market cap of roughly $5.5 trillion—making it larger than the entire GDP of every G7 country except the United States—Jen points out that partner Marc Andreessen often describes tech as “the dog that caught the bus.” Just as the nations that industrialized first became the most militarily, culturally, and economically dominant over the last century, a16z believes the future belongs to countries that adopt AI first across defense, public safety, healthcare, and everyday civic life.

```mermaid flowchart TD accTitle: AI Adoption Driving Global Partnership Model accDescr: Diagram showing how a16z leverages LP relationships to drive AI infrastructure adoption and direct investment globally.

subgraph Macro["Macro Shift"] A["Industrialization Past"] --> B["AI-First Nations Future"] C["Tech Scale: NVIDIA > G7 GDPs"] --> B end

subgraph Strategy["a16z Global Partnership Model"] D["Global / Sovereign LPs"] E["Adopt US Technologies"] F["Direct Startup Investment"] G["Local Infrastructure Acceleration"]

D --> E D --> F E --> G F --> G end

B --> Strategy ```

- When sovereign entities and global Limited Partners (LPs) partner with a16z, the objective is to help America and its allies accelerate adoption of leading U.S. technologies while offering those partners direct co-investment opportunities. - Theo Jaffee mentions that South Korea recently announced plans to distribute premium AI to every citizen as a public utility. - Jen jokingly refers to South Korea as an “extreme form of democracy,” recounting how roughly a third of South Koreans received margin calls earlier in the summer, while also praising the country as a cultural powerhouse of the future (producing BLACKPINK, *Squid Game*, and K-pop). - Other nations are aggressively integrating AI through state-driven initiatives: El Salvador under President Nayib Bukele deployed Grok for free across public schools and is piloting AI doctors, while city-states like Singapore and the UAE leverage compact populations to rapidly subsidize and deploy AI services.

---

`15:00` Data Center Backlash, DC Power, and the Machine Age Thesis

While sovereign nations race to roll out compute, the domestic environment in the U.S. faces growing public resistance and political friction surrounding data centers, alongside misinformed anxieties about surveillance. Theo asks whether this backlash poses a existential risk to the fund or if compute builds will simply migrate to more welcoming states. Jen acknowledges that while positive public sentiment would make things easier, popular narratives rarely reflect the operational reality of modern, tech-native infrastructure:

> “And where we are investing is the next generation of those data centers like a switch that have optimized for the future on not only some of the politically more sensitive things, but even things like, for example, the next set of chips are going to be DC powered versus AC powered.”

Modern facilities—such as those operated by AWS, Meta, or a16z portfolio company Switch—are engineered by technologists rather than traditional real estate developers. Switch, for example, is architected to return power to the grid, relies on minimal water consumption, and is designed to handle the complex fluidics and liquid cooling required for high-density AI clusters.

Jen highlights a looming, underappreciated physical bottleneck: future semiconductor generations will transition to direct current (DC) power instead of alternating current (AC). Yet, less than 2% of electricians in the United States are currently trained to handle DC power at scale, which is exceptionally volatile and dangerous to install.

Turning to the broader scope of the Machine Age Fund, Sophia notes that the mandate encompasses robotics and home hardware. Jen clarifies the fund's exact boundaries:

> “The supply side is the thesis of this fund where we want to invest into everything that is guided by computer science on the infrastructure side.”

The investment thesis targets everything sitting below the software application layer inside the modern data center: accelerators, CPUs, custom silicon, memory, storage, liquid cooling, data center robotics, and system-level software. Highly regulated infrastructure, such as raw power generation, remains under the purview of a16z's American Dynamism practice.

- On the demand side, AI agents consume five times the token volume of humans, and autonomous agents have now outpaced human traffic on the internet. - With global AI penetration still below 5% of potential usage, agentic workflows and personal products (like Grok or Instinct) will push compute demand parabolic. - Addressing an earlier quip that the team might deploy the entire Machine Age Fund in six weeks, Jen clarifies for listening LPs that the remark was a joke, though it reflects the unprecedented groundswell of veteran system engineers leaving incumbent tech firms to fix severe infrastructure bottlenecks.

`20:00` The Surge of Hardware Founders and the Anatomy of Diligence

The venture landscape is undergoing a dramatic realignment in deal flow. Where hardware once represented virtually zero percent of incoming startup pitches, it now accounts for over twenty percent of the pipeline focused on the physical computing world. When asked how evaluating these massive, capital-heavy physical infrastructure deals compares to assessing traditional software venture bets, Jen Kha explains that the founder profile itself represents a striking throwback to an earlier era of technology. While there will always be ambitious students fresh out of university attempting hardware, the reality of building at this tier demands established customer networks, hyperscaler ties, and deep systems engineering mastery that only decades of industry experience can provide.

Because the technical and commercial barriers are so high, the talent driving this new wave consists largely of veterans spinning out of industry incumbents. Kha points to NextHop, founded by alumni of Arista Networks who identified structural bottlenecks in the current networking paradigm and left to build an entirely new architectural solution. These seasoned operators spent twenty to thirty years inside the previous cycle of enterprise infrastructure and are now returning to entrepreneurship to architect the next computing epoch. As Kha observes: “Oftentimes, people who have the benefit of experience in prior errors of the buildout that had been effectively kind of incubating for the last 30 plus years, 20, 30 plus years, are now saying, I'm going to go build the company around this.”

```mermaid flowchart TD accTitle: Deep Systems Experience Applied to AI Infrastructure accDescr: Incumbent data center leadership feeding talent into the Machine Age thesis. subgraph Incumbents["Data Center & Networking Incumbents"] AR["Arista Networks"] VM["VMware"] INT["Intel"] end

subgraph Veterans["Experienced Systems Operators"] NH["NextHop Founders"] MC["Martin Casado<br/>(Nicera / VMware)"] RR["Raghu Raghuram<br/>(VMware)"] GA["Guido Appenzeller<br/>(Intel CTO)"] end

subgraph A16Z_MAF["Machine Age Fund Diligence & Execution"] EVAL["Technical Diligence & Hyperscaler Access"] PHYS["Physical AI Stack Rebuild:<br/>Chips, Networking, Power & Cooling"] end

AR --> NH VM --> MC VM --> RR INT --> GA NH --> PHYS MC --> EVAL RR --> EVAL GA --> EVAL EVAL --> PHYS ```

This specialized operational reality mirrors the investment team evaluating the deals at Andreessen Horowitz. Kha highlights the partners spearheading these diligence cycles: Martin Casado, who founded Nicera and sold it to VMware; Raghu Raghuram, the former CEO of VMware who originally acquired Nicera and served as Casado’s boss; and Guido Appenzeller, the former Chief Technology Officer of Intel who ran their technical data center operations. All of them spent the past decade navigating the software boom, but their roots are grounded in the physical constraints of data centers. Because the pool of technical leaders capable of solving problems of this scale and navigating hyperscaler procurement is finite, diligence relies heavily on long-standing relationships and rigorous peer-level technical scrutiny. As Kha emphasizes: “And so from a diligence perspective, it's obviously a very almost finite world of the folks who could actually go after this massive problem and also build a customer base and sell into and actually build companies of this scale and size.”

- The proportion of hardware and physical infrastructure pitches reviewed by the firm expanded from nearly nothing to over 20% of current pitch volume. - Founder profiles in this cohort heavily skew toward 20- to 30-year systems veterans spinning out of incumbents such as Arista rather than young university graduates. - Key partners leading the technical diligence include Martin Casado (founder of Nicera), Raghu Raghuram (former VMware CEO), and Guido Appenzeller (former CTO of Intel).

---

`23:05` Why What’s Old Is New Again: Silicon Valley’s Origins

Wrapping up the conversation, the hosts congratulate Kha on the launch of the Machine Age Fund, expressing excitement over how a16z aims to reshape the venture landscape by refocusing on the tangible, physical layers of computation. The hosts summarize the thematic shift with the phrase, “What's old is new again,” reflecting on how computing innovation has swung like a pendulum back to the metal, silicon, and wires that define raw capability.

Kha enthusiastically agrees, pointing out that this wave is a direct reconnection to the foundational thesis of venture capital itself. The early history of technology investing was defined not by cloud applications or software interfaces, but by the physical silicon fabric of Santa Clara Valley. As Kha notes regarding the historical circle closing: “Yes, this is a, we're almost thrown back to where venture capital actually started and why Silicon Valley is actually called Silicon Valley.” The episode concludes with mutual thanks before running through standard regulatory disclosures noting that the presentation is for educational purposes and not investment advice.

## Notable quotes > `01:58` Any software need you have today, AI can actually do it. But we have to solve for all the physical constraints in order for AI to actually solve for software. — Jen Kha

## Mentioned Jen Kha, Theo Jaffee, Sophia Dew, Marc Andreessen, Chris Dixon, Martin Casado, Raghu Raghuram, Guido Appenzeller, Sanjay Mehrotra, Naveen Rao, Nayib Bukele, Andreessen Horowitz, Machine Age Fund, NextHop, Unconventional, Switch, NVIDIA, Micron Technology, SK Hynix, Samsung, Arista Networks, VMware, Intel, Nicira, AWS, Meta, SpaceX, Anthropic, OpenAI, Databricks, Grok, Instinct

## Video description a16z Managing Partner and Head of Global Partnerships Jen Kha joins MTS hosts Theo Jaffee and Sophia Dew to discuss a16z's Machine Age Fund and the investment thesis behind rebuilding the physical infrastructure that powers AI. Jen explains why chips, networking, memory, cooling, data centers, and other parts of the physical computing stack are becoming investable again after decades in which software captured much of the industry's attention. As AI demand pushes existing infrastructure to its limits, she explains why a16z created a dedicated fund and why hardware founders are increasingly rethinking the stack from first principles. They also discuss the global race to adopt AI, what hardware startups need beyond capital, the backlash against data centers in the U.S., and why experienced systems builders are returning to entrepreneurship as a new generation of infrastructure gets built.

[Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/68d49c98-05bc-4eec-a525-c6fdb554f130/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=68d49c98-05bc-4eec-a525-c6fdb554f130&feed=JGE3yC0V)

中文简报

a16z为何设立“机器时代基金” | Jen Kha 在本次探讨中，Jen Kha阐明了a16z专门设立11亿美元独立基金的战略考量：避开软件基金的内部竞争，专项支持早期物理AI基础设施、定制芯片及数据中心创新。在AI智能体激增的算力需求，以及韩国、萨尔瓦多等国激烈的国家级采用竞争推动下，物理瓶颈使硬件提案在进项项目流中的占比骤增至20%以上。该基金瞄准种子轮和A轮阶段，单笔投资额为2500万至3500万美元，重点支持来自Arista、英特尔和VMware等老牌机构的资深系统专家，以攻克向直流电（DC）转型等迫在眉睫的障碍。 Andreessen Horowitz（a16z）在夏季历时约两个月完成了规模达11亿美元的“机器时代基金”（Machine Age Fund）募集，旨在投资早期物理AI基础设施。 (1:58) 在a16z收到的融资提案中，硬件和物理算力方案的占比从接近零激增至20%以上，扭转了数十年来风投界独尊纯软件的局面。 (0:00) 该基金聚焦早期拐点，在种子轮和A轮阶段开出2500万至3500万美元的投资支票，涵盖定制芯片、网络、散热以及数据中心架构。 (5:00) 随着未来几代半导体从交流电（AC）向直流电（DC）转型，一项关键的本土物理瓶颈正在显现：目前接受过大规模直流电安装培训的美国电工不足2%。 (15:00)

Reading summary

Andreessen Horowitz (a16z) has launched its $1.1 billion Machine Age Fund to back physical AI infrastructure, custom silicon, and data center technologies, driven by surging physical compute bottlenecks and sovereign demand for artificial intelligence. Capital Segregation for Capital-Intensive Hardware (1:58) For decades, venture capital favored the high margins and rapid cash flows of software and SaaS models. According to Jen Kha, Managing Partner and Head of Global Partnerships at a16z, inbound hardware pitches have recently climbed from a negligible baseline to more than 20% of the firm's deal pipeline. However, pitting capital-intensive hardware startups directly against low-risk software businesses within general funds causes investment committees to bias toward software with immediate revenue. Fund Positioning Across the Supply Chain (5:00) Capitalized at $1.1 billion and closed in roughly two months, the Machine Age Fund focuses on the physical stack beneath developer software, including accelerators, custom silicon, memory, networking, liquid cooling, and data center robotics. Kha clarifies that the vehicle is designed for early-stage Seed and Series A investments with initial check sizes of $25 million to $35 million, allowing the firm to secure significant equity stakes before valuations spike. Follow-on growth capital is left to later-stage vehicles, such as a $65 million Series B check written into networking provider NextHop, while raw power generation remains managed under a16z's American Dynamism practice. Sovereign AI Adoption and Domestic Bottlenecks (10:00) Global governments increasingly treat compute infrastructure as vital national capability. Examples cited include South Korea planning to distribute premium AI to citizens as a public utility, and El Salvador integrating Grok across its public education system while trialing AI medical applications. Kha notes that tech's economic footprint has expanded to the point where NVIDIA's market cap exceeds the GDP of most G7 nations, prompting sovereign limited partners to seek access to U.S. technology pipelines. Operational Realities and Systems Engineering Experience (15:00) Building and deploying modern data center infrastructure faces tangible physical constraints. Future chip architectures are moving toward direct current (DC) rather than alternating current (AC) power, yet Kha notes that fewer than 2% of U.S. electricians are currently certified for DC environments. To mitigate political resistance and environmental strain, facilities like portfolio company Switch are being engineered by technologists to return power to the grid and limit water use through high-density liquid cooling. Veteran Operator Inflows and Historical Parallels (20:00) Unlike traditional consumer software startups led by recent graduates, the current hardware wave is driven largely by industry veterans with two to three decades of systems engineering experience at firms like Arista Networks, VMware, and Intel. Due to high technical barriers and enterprise procurement cycles, diligence at the fund relies heavily on seasoned partners such as Martin Casado, Raghu Raghuram, and Guido Appenzeller. Kha observes that by returning to physical silicon, networking, and power, the venture industry is circling back to its origins in Santa Clara Valley. While sovereign demand and agentic AI workloads continue to surge, the scaling of the Machine Age thesis remains tethered to real-world friction, including specialized workforce shortages like certified DC electricians, local data center resistance, and high upfront capital requirements.

精炼阅读摘要

针对人工智能算力受制于过去三十年传统软件基础设施物理瓶颈的现状，Andreessen Horowitz（a16z）合伙人 Jen Kha 指出，该机构设立了规模达 11 亿美元的“机器时代基金”（Machine Age Fund），旨在突破传统软件投资对重资产的偏好偏见，通过早期重仓底层硬件与物理计算架构来重构现代技术堆栈。突破物理瓶颈与隔离基金机制 (1:58) Jen Kha 表示，过去三十年建立的计算基础设施主要是为互联网和企业级 SaaS 设计的，而 AI 具备极高的计算密集度，沿用旧架构如同在“穷人版”系统上运行，无法承载未来需求。Kha 认为当前 AI 已经能够解决软件层面的需求，但前提是必须解决底层的物理限制。为此，a16z 于 8 月 28 日推出了规模达 11 亿美元的机器时代基金，并在大约两个月内完成募集；该机构在今年募集了全球超过 23% 的风险投资资金。设立独立的专门基金是为了解决内部投资组合构建的机制偏见。在通用软件或基础设施基金中，重资产、研发周期长的硬件项目若与现金流迅速、高毛利的软件项目同台竞争，投资委员会往往会偏向确定性更高的软件标的。通过将硬件资本单独隔离，a16z 得以对芯片、机架、液冷系统和物理计算节点等资本密集型底层创新提供专属支持。目前，硬件在其收到的商业计划书中占比已从接近于零飙升至 20% 以上。早期股权布局与物理基础设施边界 (5:00) 面对超大规模云厂商（Hyperscalers）每年约 1 万亿美元的资本支出规模，Jen Kha 阐明机器时代基金的定位并非参与后期的重资产建设，而是在 Seed 轮与 Series A 轮开出 2500 万至 3500 万美元的早期支票，在估值拐点到来之前确立可观的股权比例；后续扩张轮次（如 NextHop 的 6500 万美元 B 轮融资）则由 a16z 旗下成长期基金跟进。这种布局也响应了机构有限合伙人（LP）对硬件供应链未上市公司敞口的需求，尤其是在美光科技、SK 海力士和三星等上市硬件供应链巨头估值大幅飙升的背景下。该基金的投资边界严格限定在现代数据中心软件应用层之下的物理硬件与系统软件，包括定制硅片、网络互联结构、液冷设备、数据中心机器人及存储技术，而电网建设等受强监管的基础能源生成领域则归属于其“美国活力”（American Dynamism）专项。在工程实施层面，物理约束尤为严峻：下一代芯片正从交流电（AC）向直流电（DC）过渡，但目前全美仅有不到 2% 的电工接受过大规模处理高波动性直流电的专业培训。此外，由于自主智能体（Agent）消耗的 Token 量是人类的五倍且流量已超越人类，加之全球 AI 渗透率仍不足 5%，底层计算基础设施的升级已迫在眉睫。主权AI浪潮与地缘资本合作模式 (10:00) 随着算力演变为核心国家基础设施，硬件与人工智能已上升至国家竞争层面。Jen Kha 指出，NVIDIA 约 5.5 万亿美元的市值已超过除美国外任何 G7 成员国的 GDP，科技投资的发展正如同合伙人 Marc Andreessen 所形容的“狗追赶上了公共汽车”。历史上率先完成工业化的国家在军事、文化与经济上建立了主导地位，而 a16z 认为未来将属于率先在国防、医疗、公共安全与日常生活中全面采用 AI 的国家。这一趋势促使多个主权实体激进推动国家级 AI 布局。例如韩国宣布将优质 AI 作为公共事业向全体国民开放，其总统访美期间唯一一站即为硅谷；萨尔瓦多在公立学校免费推广 Grok 并试点 AI 辅助医生；新加坡与阿联酋也借助人口体量集中优势快速推进技术落地。面对美国本土数据中心建设面临的公众质疑、环保阻力与监管摩擦，a16z 通过与全球及主权 LP 建立合作，帮助盟国加速采纳领先的美国技术并提供直接跟投机会，同时投资 Switch 等具备电网回馈、极低耗水及高效液冷能力的技术原生型下一代数据中心运营商。资深老兵创业潮与重返硅谷本源 (20:00) 与消费互联网时代大学毕业生轻资产创业的模式不同，底层硬件与物理计算架构的门槛极高，涉及超大规模客户采购流程与极为复杂的系统工程。Jen Kha 强调，当前涌现的硬件创业者普遍是拥有 20 至 30 年深厚行业积淀的资深老兵，例如由 Arista Networks 前骨干创立的高性能网络企业 NextHop，以及由前 Databricks 高管 Naveen Rao 创办的从零设计专用 AI 芯片的公司 Unconventional。这些行业老兵在经历数十年的企业基础设施周期后，选择脱离成熟巨头来解决当前网络与算力架构中的结构性瓶颈。这种极度垂直的技术门槛决定了该领域的专业人才圈层相对有限，也深刻影响了 a16z 的尽职调查机制。该基金的尽调由具备深厚物理数据中心与系统背景的合伙人主导，包括 Nicera 创始人兼前 VMware 高管 Martin Casado、前 VMware 首席执行官 Raghu Raghuram，以及前英特尔首席技术官 Guido Appenzeller。节目总结认为，这种向金属、硅片与线路等物理基底的回归，印证了“循环往复，旧即是新”（What's old is new again）的历史规律，使风险投资重新回到了“硅谷之名源于硅硬件”的最初起点。机器时代基金的设立标志着风险投资正从轻资产软件溢价向重资产底层物理基础设施战略回摆。然而，面对下一代芯片直流电改造中专业电工不足 2% 的人才缺口、美国国内数据中心建设的社会阻力，以及云厂商万亿级资本开支的压迫，初创企业能否在漫长的研发周期中跨越物理交付与商业化断层，仍是该投资模式面临的长期挑战。

Detailed notes — zh

## Background 本期节目由 Jen Kha、Theo Jaffee 与 Sophia Dew 于 2026 年 8 月 30 日在播客及网络平台发布。节目页面介绍显示，a16z 管理合伙人兼全球合作伙伴负责人 Jen Kha 共同探讨了 a16z 推出规模达 11 亿美元的“机器时代基金”（Machine Age Fund）的投资逻辑，深入剖析了在 AI 需求将现有设施推向极限的背景下，芯片、网络、内存、冷却及数据中心等物理计算栈重获资本青睐的深层原因。 ## TL;DR 节目围绕 a16z 设立 11 亿美元 Machine Age Fund 的背景展开，指出 AI 的爆发已将算力瓶颈从软件层推向底层的物理基础设施与电气架构。Jen Kha 分析了为何传统的软件投资基金无法兼顾重资产硬件，以及早期重仓芯片、定制硅片与新型数据中心的必要性。对话还探讨了全球主权国家在 AI 普及上的激进竞争、美国本土面临的数据中心政治阻力与直流电改造断层，并强调深耕行业数十年的资深系统架构师正在掀起硬核基建的创业回流潮。 ## Atmosphere & style 对谈整体展现出极高的专业度与宏观视野，节奏紧凑而充满行业洞察。主持人与嘉宾的互动默契融洽，在严谨的技术架构剖析与生动的行业轶事（如将内存巨头高管戏称为科技界泰勒·斯威夫特）之间切换自如。整场交流洋溢着一种重温硅谷初心、见证历史周期轮回的兴奋感与宿命感，深刻传递出科技界从比特重新杀回原子世界的雄心与张力。 ## Bulletin - `00:00` a16z 宣布设立规模达 11 亿美元的“机器时代基金”（Machine Age Fund），专注于芯片、定制硅片、网络、机架、液冷和数据中心等承载 AI 计算的物理基础设施。 - `20:00` 硬件初创项目在 a16z 接收到的融资计划书中占比已突破 20%，投资逻辑正在从“软件吞噬世界”逆转为由物理层极限主导的算力基础设施重构。 - `15:00` 新一代芯片要求数据中心全面从交流电（AC）向直流电（DC）转型，但全美仅有不足 2% 的电工接受过高压直流电操作培训，构成了关键的人力和安全瓶颈。 - `17:00` 当前全球 AI 渗透率仍不足 5%，但互联网上 AI Agent 产生的交互量已超越人类，且 Agent 单次会话消耗的 Token 数量高达人类操作的 5 倍，算力供需缺口将呈现抛物线式暴增。 ## Key takeaways - `00:00` 全球算力竞速已上升为主权竞争，韩国将高级 AI 纳为全民公共基础设施，萨尔瓦多在全国公立学校和医疗中免费落地 AI。 - `02:30` 设立独立的 11 亿美元硬件基金是为了解决内部决策偏差，避免重资产、长周期的硬件项目与高确定性、高现金流的成熟软件 Infra 项目共用资金池而遭到抑制。 - `05:00` 该基金聚焦在种子轮与 A 轮阶段开出大额早期支票锁定高股权比例，而非与年度资本支出逼近 1 万亿美元的 Hyperscalers 拼后期重资产竞赛。 - `10:00` 硬件初创企业需要的远不止资本，更需要地缘政治护航与市场准入支持；a16z 正与主权国家生态合作推进底层硬科技落地与直接联合投资。 - `10:00` 英伟达约 5.5 万亿美元的单体经济体量已超越除美国外的所有 G7 国家，科技基础设施已成为未来决定国家综合竞争力的胜负手。 - `13:00` 美国本土对数据中心的社区与政治阻力可能加速算力基建供应链的外迁离岸，甚至催生太空数据中心等新型构想。 - `15:00` 现代 AI 数据中心本质上是由计算机工程师构建的技术系统，具备反哺电网与低水耗液冷特征，已彻底脱离传统商业地产的开发属性。 - `15:00` 全美受过高压直流电（DC）操作培训的专业电工不足 2%，成为数据中心向更高能效架构改造的技术与人才死穴。 - `17:00` Machine Age Fund 明确界定投资边界为“计算机科学驱动的基础设施”，电厂审批等外部受管制的能源公用事业由 American Dynamism 基金承担。 - `20:00` AI 基础设施创业呈现明显的“老兵回流”特征，来自 Arista Networks、Intel 等上一个时代深耕二三十年的资深架构师正纷纷离职创业。 - `20:00` a16z 组建了由 Martin Casado、前 VMware CEO Raghu Raghuram 及前 Intel 数据中心 CTO Guido Appenzeller 构成的全明星团队，专门评估底层硬核技术可行性。 - `22:38` 风投范式正在经历从纯代码向物理原子的回归，让“硅”重新成为硅谷的核心驱动力，开启硬核物理系统定义的新纪元。 ## Detailed walkthrough ### `00:00` 全球 AI 基础设施竞速与 11 亿美元 Machine Age Fund 的诞生在全球范围内，主权国家与创业生态正在以惊人的速度推进人工智能的底层基础设施建设。韩国刚刚宣布将高级 AI 作为公共事业向全体国民开放，萨尔瓦多在全国学校中免费部署了 Grok 并开始启用 AI 医生，这类海外加速落地的案例让美国本土的数据中心供应链与政策环境感受到了前所未有的紧迫感。几十年来，风险投资行业曾逐渐远离重资产的硬件领域，但 AI 正在以不可逆的势头将资本拉回物理世界。硬件项目在 Andreessen Horowitz（a16z）收到的创业商业计划书中的占比，已经从过去的无足轻重跃升至 20% 以上。为了抓住这一历史性转折，a16z 宣布推出规模达 11 亿美元的“机器时代基金”（Machine Age Fund），专注于芯片、定制硅片、网络架构、机架、冷却系统、数据中心物理构件乃至机器人和能源。a16z 管理合伙人兼全球合作伙伴负责人 Jen Kha 指出，华尔街传统上信奉“五月卖出离场”（Sell in May and go away）的夏季淡季逻辑，但他们却在 8 月 28 日迅速敲定并公布了这支基金，这本身就标志着行业所处的周期极其特殊。当前 AI 运行在为上一代互联网与 SaaS 时代搭建的“穷人版”物理底座上，而高强度的数学与计算负荷正在将现有系统逼向极限。 ```mermaid flowchart TD accTitle: a16z 基金矩阵与物理计算栈分工 accDescr: 展示 Apps 基金、Infra 基金与 Machine Age 基金在计算栈中各层级的分工定位。 subgraph SoftwareStack ["软件层 (现有基金覆盖)"] direction TB Apps["Apps Fund (面向 B2B / B2C 应用)"] Infra["Infra Fund (面向开发者工具与软件基础架构)"] end subgraph PhysicalStack ["物理层 (Machine Age Fund 专注领域)"] direction TB Compute["芯片与定制硅片 (Custom Silicon)"] Network["网络互联与机架 (Networking & Racks)"] Facility["冷却、电力与数据中心 (Cooling & Data Centers)"] end SoftwareStack -->|"依赖支撑"| PhysicalStack ``` - 韩国将高等级 AI 纳入类似公共公用事业（Public Utility）的公民福利体系。 - 萨尔瓦多在全国教育系统中免费引入 Grok，并在医疗场景中开始试水 AI 诊断。 - a16z 内部硬件类项目的 Pitch 占比从曾经的边缘状态攀升至超过 20%。 - Machine Age Fund 资金规模为 11 亿美元，专门针对软件栈之下的物理基础设施瓶颈。 --- ### `02:30` 追随“书呆子能量”：为什么需要设立独立的硬件投资基金在谈及为何不直接通过现有的基础设施基金（Infra Fund）或成长基金进行投资时，Jen Kha 阐述了底层投资哲学与机构投资组合构建（Portfolio Construction）的双重考量。a16z 始终贯彻合伙人 Chris Dixon 所说的“追随书呆子能量”（following the nerd energy），即极客们在夜晚和周末捣鼓的技术，往往就是普通大众未来的生活日常。尽管 Marc Andreessen 曾留下“软件正在吞噬世界”的著名论断，但如今 AI 已经有能力解决各种纯软件需求，真正制约发展的反而成了底层的物理极限。从基金运作与激励机制来看，设立专门的独立投资载体至关重要。硬件初创企业起步阶段往往需要巨额前期资本投入，如果将一个早期的重资产硬件项目与一个年收入轻松破十亿甚至二十亿美元、确定性极高的软件基础设施项目放在同一个决策池中对比，投资委员会几乎总会不可避免地偏向后者。独立基金向全球硬件创业者发出了极具穿透力的信号，确保机构有专项资金池在早期开出 7500 万美元级别的大额支票，从而锁定核心技术公司的早期股权。 > “马克曾有一句名言叫‘软件正在吞噬世界’。事实证明，AI 已经解决了软件问题，对吧？如今你拥有的任何软件需求，AI 实际上都能搞定。但我们必须解决所有的物理限制，AI 才能真正去解决软件层的问题。” —— Jen Kha - 投资逻辑演进：软件时代的核心是代码吞噬业务，AI 时代的核心是物理基础承载代码生成。 - 内部决策偏差风险：如果硬件与软件共用资金池，缺乏前期收入的硬件项目会被高现金流的成熟 Infra 压制。 - 独立基金职能：向外部创始人释放专注信号，并在内部为重资产、长周期项目建立专属估值与持有模型。 --- ### `05:00` 机构投资者的热情、万亿级资本支出与早期基金定位今年 a16z 募集了全行业超过 23% 的风险投资资金，这一数字折射出有限合伙人（LP）对私有 AI 市场的极度饥渴。过去 AI 领域的大量价值沉淀在未上市的私有企业中，而公开市场上相关的标的已经历了极其迅猛的估值重塑。Jen Kha 戏称，美光科技（Micron）首席执行官 Sanjay Mehrotra 在行业内简直成了“科技界的泰勒·斯威夫特”（Taylor Swift），包括 SK海力士与三星在内的数据中心供应链巨头均受到狂热追捧。然而，许多工程师意识到现有的物理基础设施只是在“削足适履”地迎合 AI，从一张白纸重新设计的下一代计算架构蕴藏着更大的超额收益。面对主持人关于“全美超大规模云厂商（Hyperscalers）今年资本支出将近 1 万亿美元，11 亿美元基金在此规模下如何立足”的疑问，Jen Kha 强调了早期风险投资与晚期公募投资在组合构建上的本质区别。该基金并不是要去和科技巨头拼数据中心基建的资本支出，而是聚焦在种子轮与 A 轮阶段，通过单笔 2500 万至 3500 万美元的投资在拐点来临前锁定高比例的所有权。例如，在面对高性能 AI 原生网络公司 NextHop 时，a16z 成长基金在 B 轮领投了 6500 万美元；而 Machine Age Fund 更倾向于在极早期押注像前 Databricks 核心高管 Naveen Rao 创办的 Unconventional 这样的颠覆性底层项目。 > “在我的职业生涯之初，想在夏季的短短两个月内搞定并关闭一支基金是根本不可能的，对吧？大家都在休假。这件事情能如此迅速地完成，恰恰反映了来自私有资本池以及那些看清发展大势的 LP 们的庞大需求。” —— Jen Kha - 全球 Hyperscalers 年度算力资本支出规模预计逼近 1 万亿美元级别。 - 公开市场芯片与内存大厂（如美光、SK海力士、三星）成为供应链瓶颈的最大受益者。 - Machine Age Fund 明确将投资阶段定位于 Seed 和 Series A，通过早期大额支票追求高占股比，而非参与后期的重资产竞赛。 - 投资案例分层：NextHop 适合成长基金在 B 轮拐点进入，而 Unconventional（Naveen Rao 创立）则是 Machine Age Fund 锁定的从零到一典型底层企业。 ### `10:00` 早期芯片投资逻辑与超越资本的全球地缘合作针对芯片设计，目前的创业团队正尝试从第一性原理出发，探索如果专门为 AI 设计一款芯片究竟该怎么做。他们的目标是采用更精简、更基础的架构来大幅提升芯片性能。这类团队通常会选择“闭关深造”（going off in the cave）沉心研发。a16z 在种子轮（Seed）便果断出手，随后该团队在估值暴涨后完成了新一轮巨额融资。Jen Kha 强调，在拐点发生前尽早介入至关重要，这不仅能以较小资金换取最大化的股权占比，还能避免后续因估值高企而错失良机。谈到硬件初创企业的独特需求，Jen Kha 指出，这类企业需要的远不仅是单纯的资金支持，更渴求市场准入与地缘政治层面的伙伴关系。随着算力竞争上升至国家战略高度，科技与主权国家之间的界限正在模糊。例如韩国总统近期访美，其在全美唯一的行程就是直奔硅谷。Jen Kha 援引合伙人 Marc Andreessen 的经典比喻阐述了科技行业当前拥有的庞大体量与宏观影响力： > “Marc Andreessen calls it, technology is the dog that caught the bus.” 科技已经成长为全球经济中不可忽视的核心力量。Jen Kha 提到，市值达到 5.5 万亿美元的英伟达，其经济体量已经超越了除美国以外的所有 G7 成员国。纵观过去一个世纪，最先完成工业化的国家在军事、文化和经济上最为成功；而在未来，最先将 AI 深度融入国防、公共安全、医疗及政务系统的国家，将领跑全球竞争。基于这一趋势，a16z 在维护全球 LP 关系时的定位发生了根本性转变：不仅是帮助 LP 投资旗下的基金，更是在主权国家层面成为科技战略合作伙伴。a16z 致力于帮助美国及其盟国加速拥抱以美国为核心的前沿技术，鼓励这些国家在其国内全量部署相关技术，甚至协同 a16z 对底层硬件企业进行直接联合投资。这种模式能够以远超以往技术周期的速度，在当地推进产业落地，并让这些国家真正享受到技术红利带来的经济增长。 ```mermaid flowchart TD accTitle: 全球地缘战略与硬件投资协同架构 accDescr: 展现 a16z 如何连接资本、美国底层硬科技与主权国家生态 subgraph Fund["a16z 投资战略"] Seed["种子轮早期重仓"] --> Capital["最大化股权占比与资本效率"] Capital --> TechStack["聚焦 AI 原生硬件底层创新"] end subgraph GeoPolitics["主权国家与全球 LP"] NatSec["国家安全与经济竞争力提升"] DirectInv["协同直投重点硬件公司"] GovAdopt["全自主与政务/公共体系采纳"] end subgraph Impact["全球技术加速"] TechStack -->|"技术输出与基础设施构建"| GovAdopt Capital -->|"联合主权基金支持"| DirectInv GovAdopt --> NatSec end ``` - **英伟达对比 G7 体量**：英伟达市值达到约 5.5 万亿美元，单体经济规模已超越美国以外的所有 G7 国家。 - **韩国总统访美行程**：在美期间仅停留硅谷一地，直接对接前沿科技巨头与顶级投资人。 - **历史周期的对照**：正如 100 年前工业化决定国家地位一样，未来国家的强盛将完全取决于 AI 全民采用与落地的速度。 --- ### `13:00` 全球 AI 普及竞赛与美国的逆风环境针对国家层面的 AI 落地，Theo Jaffee 提到韩国近期宣布将向每位公民提供高级 AI 工具，将其作为公共事业基础设施（Public Utility）普及。Jen Kha 调侃称韩国展现了“最极致形态的民主”，其双刃剑效应非常鲜明——例如此前韩国约有三分之一的国民因剧烈的杠杆交易遭遇了追加保证金（Margin Call）。但同时，韩国就像是来自“未来的国家”，不仅产出了在全球具有统治地位的内容（如 Blackpink、热播剧《鱿鱼游戏》等），更在拥抱 AI 上不遗余力。这种国家层面的 AI 狂热并非个例。萨尔瓦多总统 Nayib Bukele 展现了极为激进的态度，他们已经在公立学校体系中免费部署了 Grok，并在医疗体系中积极引入 AI 医生。此外，新加坡、阿联酋等城邦国家凭借人口集中、财政充裕的特点，也在迅速推行技术补贴甚至全员免费部署。与之形成鲜明对比的是，美国国内正面临舆论与政治层面的阻力。由于部分针对监控隐私或资源掠夺的误解，美国本土对数据中心的反对情绪与政治逆风不断加剧。Jen Kha 分析指出，这种情况可能反而催生出意想不到的新机遇与离岸转移：正如埃隆·马斯克构想在太空中建造数据中心一样，美国国内对基础设施的阻力可能会促使数据中心及整个供应链大规模向海外拓展。 - **韩国 AI 公共化**：正式启动国家级项目，向全体公民普及高级 AI 服务作为基础公共设施。 - **萨尔瓦多激进落地**：学校教育系统全面免费接入 xAI 的 Grok 模型，公立医疗推行 AI 辅助问诊。 - **离岸转移趋势**：美国各州因政治阻力对数据中心审批收紧，迫使算力基建寻找离岸承接地乃至探索太空数据中心。 --- ### `15:00` 数据中心污名化背后的真相与直流电基础设施瓶颈关于数据中心引发的公众抵触，Theo Jaffee 询问这种舆论反弹对基金的影响有多大：究竟是导致建厂艰难的重大阻力，还是只要有部分州放行就能解决？Jen Kha 直言，如果全社会都正面支持数据中心，投资推进自然会轻松得多，但目前的舆论叙事严重脱离了工程实际。很多公众误把现代数据中心混同为传统的“房地产开发项目”，但事实上，以 AWS、Meta 算力中心以及 a16z 投资的组合公司 Switch 为代表的新一代数据中心，完全是由懂技术的科技工程师而非房地产商打造的。这些现代数据中心在设计之初就考虑到了公众的关切点：Switch 不仅没有挤占当地电网，反而具备反哺电网的能力；其耗水量极低，并且早早针对芯片液冷散热系统进行了管路适配。目前造成恶劣影响的只是极少数不合规项目，绝大多数现代数据中心都已经完成了针对 AI 算力形态的底层重构。更为关键的技术瓶颈在于物理层面的电气改造。下一代芯片要求从传统的交流电（AC）供电转向直流电（DC）供电，然而全美现存的绝大多数数据中心根本不具备匹配 DC 的电气架构。更棘手的是专业人才的严重断层。Jen Kha 特别警告了这一潜在的安全与人力瓶颈： > “And by the way there’s less than 2% of electricians in the U.S. that are actually trained on DC power because it’s very dangerous and you could potentially kill yourself and it’s very volatile to work with.” 整个算力供应链目前充斥着这类由于物理规范变动导致的系统性瓶颈，必须由具备计算机工程思维的人才彻底重构，这也是 Machine Age Fund 重点下注的核心领域。 - **Switch 数据中心设计突破**：具备反向向城市电网输电的架构设计，大幅削减工业用水，支持先进的流体散热接口。 - **交流向直流演进阻碍**：全美受过高压直流电（DC）操作培训的合格电工不足 2%，其高危险性与挥发性成为硬件安装与并网改造的最大短板。 - **科技基因取代地产基因**：现代数据中心的壁垒已转向芯片架构兼容、流体力学设计及电源管理，传统商业地产模式彻底失效。 --- ### `17:00` Machine Age Fund 的投资蓝图与算力供需大爆发对于 Machine Age Fund 超越数据中心之外的投资全景，Jen Kha 说明，其涵盖软件层以下的所有实体组件：涵盖专用加速器、CPU、定制化芯片（Custom Silicon）、存储器（Memory）、存储设备（Storage）、液冷模块，以及数据中心内部专用的巡检与运维机器人、底层系统软件等。这些在过去几十年软件盛行期被风投视为“被遗忘的语言”的领域，如今已全线重回投资视野。从底层逻辑来看，该基金由剧烈失衡的供需缺口所驱动。在需求端，AI Agent 消耗的 Token 数量是人类用户的 5 倍；如今互联网上 AI Agent 产生的交互量已经超越了人类总和。随着面向消费者的 Agent（如 Crock-Bot、Instinct 等）渗透到日常场景，算力需求曲线将以抛物线形态暴增，而目前全球的 AI 普及率依然不足 5%。在供给端，该基金的界限十分清晰：专注投资所有由**计算机科学驱动的基础设施**（Computer Science Guided Infrastructure）。因此，他们不会涉足像电厂开发这类高度受管制的能源领域（这部分由 a16z 的 American Dynamism 基金负责），而是聚焦数据中心内部由系统级代码与微电子驱动的物理堆栈。针对此前业内传闻“a16z 可能会在六周内把整支基金投光”的说法，Jen Kha 开玩笑澄清这是针对 LP 的夸张幽默，但背后的确反映了极其庞大的机会池。大批资深的系统工程架构师正纷纷从传统大厂走出重新创业。Jen Kha 援引了投资人 Martina 的评价，指出正是因为这批工程师深知现有基建在面对 AI 需求时是多么低效，才激发了这股重构物理计算堆栈的创业浪潮。 | 投资范畴划分 | 包含的细分领域与组件 | 核心驱动因素 / 团队边界 | |:--- |:--- |:--- | | **Machine Age Fund (核心聚焦)** | 加速器芯片、定制硅片、内存、液冷系统、高密度存储、机房运维机器人、系统级固件软件 | 完全由**计算机科学驱动**；解决数据中心内部的物理与架构瓶颈 | | **American Dynamism (协同基金)** | 发电厂、电网审批、传统公用事业、强监管政策领域 | 解决外部受管制的电力供应、主权韧性及国家安全级基建 | - **Agent 算力消耗量**：AI Agent 单次会话消耗的 Token 规模达到人类操作的 5 倍，且网内机器流量已超过人类流量。 - **AI 渗透率天花板**：当前全球个人与企业的 AI 整体使用率仍在 5% 以下，未来增长空间呈现指数级状态。 - **创业人才供给回流**：具备数十年深厚经验的硅谷底层系统开发者集体下场，推动新一轮硬核基础设施创业潮。 ### `20:00` 硬件创业潮的兴起与资深系统建设者的回归 Jen Kha 揭示了一个令人瞩目的数据转变：在 a16z 接收到的融资商业计划书中，过去几乎看不到任何硬件相关的项目，而如今与硬件及物理世界计算相关的 Pitch 占比已经突破了 20%。这一惊人的增长正是当下创业热情从纯软件向底层物理基础设施转移的最佳写照。当 Theo Jaffee 问及这类大型交易的尽职调查（DD）周期以及它与传统软件风投有何不同时，Jen Kha 直言，这个领域的创业者在某些层面上就像是一种“时代的倒流”（throwback）。与大学刚毕业就凭一个软件点子创业的年轻团队不同，构建物理计算基础设施门槛极高。它不仅要求创始人深刻理解物理世界的工程动力学并能将系统落地，更需要与 Hyperscalers（超大规模数据中心服务商）以及大型企业客户拥有极其深厚的商业信任与人脉。因此，当前涌现的这一波优秀人才，大多来自上一个基础设施建设时代，甚至是从现有巨头企业中剥离出来的老将。例如，前面提到的 Next Silicon 等公司，其创始人便是来自 Arista Networks 等网络硬件巨头的资深专家。他们在行业内深耕了二三十年，敏锐洞察到了 AI 算力架构重构的痛点，决定亲自下场打造新一代物理底座。Jen Kha 指出：“Oftentimes, people who have the benefit of experience in prior errors of the buildout that had been effectively kind of incubating for the last 30 plus years, 20, 30 plus years, are now saying, I'm going to go build the company around this.” ```mermaid flowchart LR accTitle: Hardware Diligence and Founder Dynamics accDescr: 展现资深技术专家重返硬件创业以及 a16z 专业投资团队进行尽调的协同关系 subgraph Founders["创业者画像 (资深行业老兵)"] F1["30年数据中心/硬件经验"] F2["巨头企业离职创业 (如 Arista)"] F3["深谙 Hyperscaler 需求与物理架构"] end subgraph A16Z["a16z 投资与尽调团队"] M1["Martin Casado (Nicera 创始人 / 前 VMware)"] M2["Raghu Raghuram (前 VMware CEO)"] M3["Guido Appenzeller (前 Intel 数据中心 CTO)"] end Founders -->|"技术验证与产业资源对接"| A16Z A16Z -->|"快速尽调与战略资本下注"| Fund["Machine Age Fund"] ``` 在尽职调查层面，a16z 组建了一支由数据中心和底层基础设施顶尖权威构成的全明星团队，以应对复杂硬核的技术评估： - **Martin Casado**：网络虚拟化先驱公司 Nicera 的联合创始人兼 CEO（该公司后被 VMware 巨资收购）。 - **Raghu Raghuram**：VMware 前首席执行官，不仅主导了当年对 Nicera 的收购，更是 Martin Casado 在 VMware 期间的直属上司，对数据中心级软件与硬件销售有极深理解。 - **Guido Appenzeller**：技术功底深厚的业界大牛，曾任 Intel 的首席技术官（CTO），主管 Intel 的整个数据中心业务板块。 Jen Kha 强调，有能力解决这种世界级难题、同时又能建立庞大客户群并撬动巨大商业规模的人才圈子其实非常有限：“So from a diligence perspective, it's obviously a very almost finite world of the folks who could actually go after this massive problem and also build a customer base and sell into and actually build companies of this scale and size.”正因如此，a16z 的投资判断往往建立在与这些行业顶尖老将多年相知、互信的基础之上，团队能迅速穿透技术黑盒，准确判断一个极具雄心的物理架构项目能否真正落地。 ### `22:38` 重返硅谷初心：让“硅”回到风险投资的核心访谈临近尾声，主持人 Theo Jaffee 与 Sophia Dew 对 Jen Kha 带来的分享表达了由衷的感谢，并对 a16z 设立 Machine Age Fund（机器时代基金）的前瞻性布局赞叹不已。他们感叹，整个风险投资的范式正在经历一场意义深远的回归——过去几十年里，软件彻底吞噬了世界，聚光灯几乎全在 SaaS、移动互联网和应用层软件上；而如今，AI 的算力瓶颈与电力、物理极限将所有人硬生生拉回到了物理世界。正如嘉宾和主持人在交流中所共鸣的：“What's old is new again.”（旧物成新潮）。这场从比特到原子的转向，某种意义上正是风险投资这项事业诞生之初的真实样貌。Sophia Dew 一针见血地指出：“Yes, this is a, we're almost thrown back to where venture capital actually started and why Silicon Valley is actually called Silicon Valley.”如今的科技界正在重新向“硅”（Silicon）本身、向芯片、向电路、向散热管道和发电机致敬。硅谷正重新回归它名字的本源，重新开启一个由硬核物理系统定义的新纪元。 ## Notable quotes > `02:30` 如今你拥有的任何软件需求，AI 实际上都能搞定。但我们必须解决所有的物理限制，AI 才能真正去解决软件层的问题。 — Jen Kha ## Mentioned Jen Kha, Theo Jaffee, Sophia Dew, Andreessen Horowitz, Machine Age Fund, Chris Dixon, Marc Andreessen, Sanjay Mehrotra, 美光科技, SK海力士, 三星, NextHop, Naveen Rao, Unconventional, 英伟达, Nayib Bukele, xAI, Grok, AWS, Meta, Switch, American Dynamism, Next Silicon, Arista Networks, Martin Casado, Nicera, VMware, Raghu Raghuram, Guido Appenzeller, 英特尔 ## Video description a16z Managing Partner and Head of Global Partnerships Jen Kha joins MTS hosts Theo Jaffee and Sophia Dew to discuss a16z's Machine Age Fund and the investment thesis behind rebuilding the physical infrastructure that powers AI. Jen explains why chips, networking, memory, cooling, data centers, and other parts of the physical computing stack are becoming investable again after decades in which software captured much of the industry's attention. As AI demand pushes existing infrastructure to its limits, she explains why a16z created a dedicated fund and why hardware founders are increasingly rethinking the stack from first principles. They also discuss the global race to adopt AI, what hardware startups need beyond capital, the backlash against data centers in the U.S., and why experienced systems builders are returning to entrepreneurship as a new generation of infrastructure gets built. [Source](https://mgln.ai/e/1344/afp-848985-injected.calisto.simplecastaudio.com/3f86df7b-51c6-4101-88a2-550dba782de8/episodes/68d49c98-05bc-4eec-a525-c6fdb554f130/audio/128/default.mp3?aid=rss_feed&awCollectionId=3f86df7b-51c6-4101-88a2-550dba782de8&awEpisodeId=68d49c98-05bc-4eec-a525-c6fdb554f130&feed=JGE3yC0V)

Source quotes

[2:30] Jen Kha: 如今你拥有的任何软件需求，AI 实际上都能搞定。但我们必须解决所有的物理限制，AI 才能真正去解决软件层的问题。

Source quotes

[1:58] Jen Kha: Any software need you have today, AI can actually do it. But we have to solve for all the physical constraints in order for AI to actually solve for software.

Original transcript

[0:00] South Korea, by the way, just announced that they're giving premium AI to every citizen as sort of a public utility thing. So very, very topical. But by the way, it's not just them. It's El Salvador. They implemented Grok in their schools, for example, for free. And they're utilizing AI doctors, for example. And you see these different examples around the world where they're accelerating their AI development and adoption way faster than in the U.S. this data center supply chain bill that happened overseas because of this sentiment in the U.S. as well. For decades, venture capital moved further and further away from hardware. AI is pulling it back. GenCob, managing partner and head of global partnerships at E16Z, joins Theo Jaffe and Sophia Du on NTS to discuss E16Z's machine age fund and why the physical infrastructure underneath AI is suddenly one of the most active areas for founders. why the existing stack wasn't designed for today's AI workloads, what needs to change across chips, networking, memory, cooling, and data centers, and why hardware has gone from a 20 fraction of the pitch's A16Z sees to more than 20%. Jen also explains why this infrastructure race is increasingly global, how governments and institutional investors are thinking about AI as a national priority, and why, as she puts it, "what's old is new again." Hello, everyone, and welcome back to MTS. Andreessen Horowitz just launched a new $1.1 billion machine age fund focused on the physical infrastructure underneath AI, from chips and networking to data centers, robotics and energy. And joining us today is Jen Kau, who's a managing partner and head of global partnerships at A16Z to talk about the machine age fund and the investment thesis behind it. Jen, welcome to MTS. Hello, hello. It's good to be here. Great to have you here. I know Theo just said a big woo, but it's pretty exciting. And we wanted to talk about why this Machine H button, why now? For sure. By the way, the classic adage is sell in May and go away. This has been the most prolific summer at all. The fact that we're announcing a fund on August 28th, when typically, you know, Wall Street is dead, is like a classic sign of where we are in the cycle and the time, which is just there's so much going on. But we announced this $1.1 billion machine age fund to invest into all of the physical constructs of the world that is now so bottlenecked, everything below the software stack. So we've got our Infra Fund, which invests into products for developers. We've got our Apps Fund, which sells into, you know, business to business and business to consumer. This is below all that. All the physical parts of enabling AI from data centers to chips to custom silicon, networking, Raxxas, all the stuff in the physical world that was honestly largely an uninvestable category for the most part. for the last 30 years, because it kind of built that infrastructure out for the last era of the internet. And then of course, of SaaS, that's all now getting rebuilt, because AI is way more mathematically intensive and compute intensive. And all that infrastructure, sort of the poor man's version that we're limping along with today needs to be repurposed for the AI age going forward. So why a separate vehicle rather than doing this investment through like the main like growth infra and other funds? Yeah, so our view is... Our job is to follow the entrepreneurs. My partner, Chris Dixon, calls it following the nerd energy. What is a nerd doing on nights and weekends is probably what us normies will be doing in the future. And... Oftentimes, entrepreneurs are that early signal into that. And so the reason why we decided to establish a separate fund is, one, there's just a groundswell of opportunity. So we went from, you know, Mark had famously said software's eating the world. Turns out AI has solved software, right? Any software need you have today, AI can actually do it. But we have to solve for all the physical constraints in order for AI to actually solve for software. And so by creating a separate fund, we're putting out the back signal to the world, to entrepreneurs who are building software. to spend time with us, first and foremost. Second of all, just from an organizational perspective, it's very helpful to have a dedicated pool vehicle for these type of investments. Because if you think about When these companies typically raise capital, they will raise – quite a bit of capital to kind of get off the ground. And typically what happens is if you put it like for like against another, let's just say, infra deal or apps deal, for example, there's so much in the way of investment you have to do ahead of time that if you do the like for like, you're almost always going to bias towards that sure thing with infra, what it's doing, you know, a billion, two billion in revenue, like off the bat and off the cuff. And so by separating [5:00] it into a fund, we're kind of staking in the ground, A, our commitment to the space, and then more importantly, B, from an organizational portfolio construction perspective, that the intention of this is to capture those winners at the earliest stages, maximize ownership. And then when you have that $75 million check that you're going life for life, you have that dedicated pool of capital to really pursue after that category. Yeah. Jen, one thing I'm curious about is your specific role, because I know you lead the partner relationships and you're in charge with institutional investors change? What are the conversations like? What makes them want exposure to the AI hardware now? Yeah, it's a great question. So first of all, we raised, I think the numbers close to over 23% of all venture capital this year. Wow. Wow. That's amazing. But as a part of that, I think this is representative of a few things. One, If you look at most of the, the value accrual in AI. That's largely been on the private side, not the public side. And so people are just starving for private capital in general, because that's where all the growth is. Two, if you think about most of the data center supply stack on the public side, those stocks have been ripping. We were joking around internally on our all hands. Sanjay, the CEO, has become the Taylor Swift of the industry, right? He has made much credit to him. also like SK Hynix, like Samsung. These are all kind of companies along the data center supply chain that have gotten so much interest because we're so constrained on the supply side. And, you know, entrepreneurs are looking at it and saying, gosh, I know those are inefficient ways to build for the future with AI because it's not, again, configured appropriately to it. And if you were to rebuild from a blank sheet of paper, it would look very much differently than the kind of shoehorning the current infrastructure today. And so we kind of follow, you know, the entrepreneurs as I was alluding to earlier as a part of that. And all of that is at the earliest stages. And then the last one I would say is, you know, we've seen in the last... a couple of months. So obviously with SpaceX going public, Anthropic coming shortly, you know, OpenAI just on the tails. I think people are now seeing this change. shift from what has historically been private staying longer and longer, no liquidity coming out. And now you're going to see multiple trillion dollar businesses go public. And so people will have access to that. But there's a whole swath of companies on the private side that are not yet public. And so people want exposure to that. And they don't want to, quite frankly, continue to invest into other asset classes like private equity, etc., that are built off When I started my career off, it would have been impossible to close a fund effectively in two months over the summer, right? People are off on vacation, etc. The fact that this happened so quickly was a reflection of the fact that there's so much demand for this from the private pool capital side and from LPs who recognize where the puck is going. Absolutely. One thing I'm curious about is, you know, $1.1 billion is a lot of money. But on the scale of like the global hyperscaler build out, I think like American companies are going to be spending something like a trillion dollars this year, a thousand times as much. What is the specific niche of what the Machine Age Fund is going to be doing at this scale? Yeah, so it is a... This is the subject of a lot of debate, Thea, when we were chatting internally about how we should size the fund, right? There's a partner who I will name unnamed who said, gosh, we're going to deploy this fund in like six weeks. Like, what are we even talking about here, right? So here's where I would reconcile on the portfolio construction side. So what we want to invest in, so we're early stage investors. So we want to come in at the seed, you know, Series A, and then kind of invest at the inflection point. effort to do is maximize ownership at the very earliest stages where we can put, you know, 25, 30, $35 million and have a very substantive ownership. That's very different than if you're a late stage investor and slash or a public market investor that has to come in at these later inflection points to get any ownership and also ownership that's substantive enough to drive fund returns. And so I think that we're just looking categorically two different things. So I'll give you an example. You know, we invested in a company called NextHop, which is building of AI-first high-performance networking. And this was a company that was at the Series B inflection point, but we put in, you know, $65 million at the growth stage. That's a great opportunity for our growth fund to be coming in at that inflection point where they're taking off. Ideally, though, what we're doing is actually investing into companies like Unconventional instead. So this is the Naveen Rao company who was formerly with Databricks that we've known for a very long time. from, [10:00] the design perspective of if you were to design a chip for AI. And so, What they're trying to do is target much simpler structure to enhance performance in the chips themselves. And they're kind of going off in the cave and they're going to go and do this endeavor and project. But we invested at the Seed And I, And subsequently, they raised a huge up right after that. And so it's really important to come in at the early stage before these things start to inflect, and then to be able to maximize ownership and therefore put capital work at maybe smaller amounts that we would otherwise not be able to just waiting. Yeah, Jen, one of the things that you explicitly wrote about as well is you are looking for partners that can provide more than just capital. So this includes market access or things like geopolitical relationships. What are the specific needs that you're seeing hardware companies need that maybe prior companies you guys would work with before didn't? Yeah, this is a great question because this is now coming to the point where it's at a national level. So a month ago, the president of Korea came to the U.S., And their first and only stop in the U.S. was in Silicon Valley, which is really interesting, right? If you think about the context of the fact that companies now, NVIDIA, you know, at 5.5 trillion is the same size as NVIDIA. countries in terms of their GDP. So NVIDIA is larger than all of the G7 countries except the U.S. And so that scale is just totally different in terms of our partner Mark Andreessen calls it, technology is the dog that caught the bus. And as a byproduct of that, we are now a bigger and bigger, technology is a bigger and bigger part of the economy. And as such, countries are viewing on boarded into the AI age. And particularly, you know, if you look at countries of the last 100 years that were the most militarily, culturally, economically successful, they were the ones that industrialized first. We think countries of the future are gonna be the ones that adopt AI first across defense, across public safety, across healthcare. And that infiltrates all throughout the government and then also comes bottom up from their citizens utilizing it as well. And so in that vein, to answer your question, Sophia, How we think about our relationship with our LPs, who are oftentimes global in nature, thinking about these national priorities, is to say, how can we help them, you know, utilize the latest and greatest technologies, which are oftentimes U.S.-based technologies, right? How can we help America and her allies accelerate into the future by not only investing in our funds, but potentially utilizing these technologies by adopting these technologies and then also oftentimes coming alongside of us in investing directly in some of these companies as well? doing so, you could actually accelerate the adoption process locally in way more ways than the distribution of prior technologies and prior technology cycles as well. And so, um, To your earlier question around, you know, what the intention of global partnership means is, you know, we want to be partners not just in the capital construct, but also in how a country and government is thinking about technology if we can help them accelerate that adoption. And then also, quite frankly, benefit from the economic development of it as well. South Korea, by the way, just announced that they're giving like premium AI to like every citizen as sort of a public utility thing. So they're very, very topical. Yeah, South Korea is like the most extreme form of democracy, by the way. And it cuts both ways because, you know, I don't know if you saw this, Theo, I bet you saw this. Like a third of South Koreans got margin called earlier this summer with a situation. So they take democracy in extreme forms. And also they are, we oftentimes joke, Korea is like the country of the future, right? They have the number one girl band with black money. They have like Squid Games. They had, you know, gosh, K-pop Demon Hunters, right? And so like they're like the new Hollywood of the future from a content perspective, but they're also full board into AI. So it doesn't surprise me that they're rolling this out. But by the way, it's not just them. It's El Salvador who, you know, President Bukele has been super aggressive around utilizing AI. They implemented Grok in their schools, for example, for free. And they're utilizing AI doctors, for example. And there's a lot of countries that you wouldn't expect. And sometimes there's city-state countries like Singapore or UAE, for example, where it's easier, just given the population, to adopt technologies and then also make it subsidized or free, for example, in some of these instances. But you see these different examples around the world where... they're accelerating their AI development and adoption way faster than in the US. And unfortunately, you know, in the US, we've been seeing a lot of political tail headwinds with with, you know, [15:00] data centers and, you know, in some instances around surveillance, etc., that are just falsehoods. But, you know, we potentially view that as an opportunity as well, where Elon's building data centers in space, we might also see the influx of a lot of this data center supply chain bill that happen overseas because of this sentiment in the U.S. as well. I'm interested in that aspect of it. How does public backlash against data centers affect this fund at all? I can imagine going either way where maybe it becomes much harder to build data centers and that's bad. But also I can imagine us being in a regime where if anywhere in the U.S. allows data centers, then the data centers will just get built there and then things will be basically fine. Or is it something more in between? So for the fund itself, obviously it would be a much easier process if everyone was on the camp of being positive on data centers. It's just not the reality of where the world is today. But unfortunately the narrative has gotten in the way of reality which is if you look at most data centers and particularly the modern ones So I talking about AWS Meta data centers Switch data center which is one of our portfolio companies for example They are actually modern versions of the data center that are built by tech people, not real estate people. And they're actually building for all the nuances of the sensitivities people have around data centers. So in the case of Switch, for example, They actually contribute power back to the grid. They don't take away. And they've built their infrastructure to enable that. They use very little water. And then particularly, you know, they're one of the very few data centers that are actually capable of being able to manage around fluidics in the future for chips. And so there's increasingly, you know, everyone paints data centers with a broad brushstroke, but there's a very small percentage of bad actors with data centers. Vast majority have actually configured for the new form factor. And where we are investing is the next generation of those data centers like a switch that have optimized for the future on not only some of the politically more sensitive things, but even things like, for example, the next set of chips are going to be DC powered versus AC powered. And most data centers aren actually built for that infrastructure And by the way there less than 2 of electricians in the U that are actually trained on DC power because it very dangerous and you could potentially kill yourself and it very volatile to work with And so you're just seeing all these bottlenecks in the infrastructure that need to get fixed in order for AI to proliferate. And this is where the vast majority of the focus of the fund is oriented around. But even expanding beyond just data centers, you know, you guys have a mandate specifically. It includes robotics, home AI hardware. You guys have a broader thesis. Can you share more about what your broader thesis is? Yeah, for sure. So it's everything. Think of, again, below the software stack, but you can almost even analogize it to anything in the data center we are focused on. So accelerators, CPUs, custom silicon, memory, storage, again, all things that were forgotten language, if you will, from an investable category perspective the last several decades. Liquid cooling, you know, all of that physical stuff that needs to get built out. And then related to that, you know, robotics within the data center, for example, system software, etc. And so, you know, the thesis around it is, you know, kind of at a very top level. we all know the demand for AI You know agents are using five times the amount of tokens as humans are and we just crossed over to agents before The volume of agents are now more on the internet than humans are and so that's just going to continue to go parabolic, especially as consumers start to have personal use cases like, for example, with Crock-Bot or Instinct or others. And there's less than still 5% of AI usage today. And so that's the kind of demand side. The supply side is the thesis of this fund where we want to invest into everything that is guided by computer science on the infrastructure side. So we don't want to invest into regulated industries like power, for example. That's more of our American dynamism team. But everything on the computer science guided side inside the data center is the focus of the fund. So relatedly, I'm also curious about like you said earlier, like people are saying you might deploy the entire fund in six weeks. I'm not sure like how joking that was, but like, how long does it actually take? For my LPs listening, that is joking. There's just so much opportunity, right? And in ways of which, you know, we have, again, seen this groundswell of entrepreneurs coming out the woodwork and saying, like, I know that this process is incredibly inefficient for what I see in terms of demand. Like, I want to build a company after this. And so, you know, I think Martina was quoted [20:00] We went from seeing basically nothing in terms of hardware pitches to over now 20% of our pitches are hardware kind of physical world related side. And so that's really the impetus also behind some of this groundswell of entrepreneurial enthusiasm as well. So how long does it take to identify and diligence these things, especially deals of the scale? Is it like different from more traditional like software VC deals? Yeah, so it's interesting that the founders in this category are, are sort of like a throwback in some respects. Like you're unlikely, although I'm sure there's some entreprenizing, you know, young individual coming out of university who are wanting to go after these categories. But because it requires so much in the way of relationships with the hyperscalers, with customers, really understanding the physical dynamics of how to get something like this off the ground, you are seeing a generation of talent that are... from a prior error, sometimes spinning out from, you know, some of the incumbents to start companies. As related to that, you know, the company that I was referring to earlier, Nextop, were former folks from Arista, for example, that had recognized this problem and decided to start a company to go actually after the new version and form of this. And so what you're seeing is a new generation of companies. Oftentimes, people who have the benefit of experience in prior errors of the buildout that had been effectively kind of incubating for the last 30 plus years, 20, 30 plus years, are now saying, I'm going to go build the company around this. So it looks very different than in the past. And the diligence, to your point, the diligence cycles on these, our team is... Maybe just to spend a second on it here. So Martin Casado, you know, formerly CEO and founder of a company called Nicera, which sold into VMware. Raghu Raghuram, who was a former CEO of VMware, who actually acquired Nicera and then was Martin's boss while at VMware. So both of them come from a very deep kind of selling into the data center background. And then Guido Appenizer, who is actually the former CTO of Intel, is also someone who is very, very technical, sold into and ran kind of the data center business for Intel as well. And so you're seeing folks who, you know... for maybe the last 10, 15 years, became more pulled into software, come back into the hardware side of the world, and leveraging that experience as a part of it. And so from a diligence perspective, it's obviously a very almost finite world of the folks who could actually go after this massive problem and also build a customer base and sell into and actually build companies of this scale and size. So from a diligence perspective, it's oftentimes people we know and have gotten very close to over the years and can actually do the diligence around it to really assess to whether something can get off the ground. Definitely. Jen, really wanted to thank you for coming and joining us today on MTS to share more about the Machine Age Fund. Congratulations to all of you on watching this. This is so exciting. It's going to be really fascinating to see how Venture Capital and A16Z, how you guys changed the game. So looking forward to it. What's old is new again. Yes, this is a, we're almost thrown back to where venture capital actually started and why Silicon Valley is actually called Silicon Valley. True. So excited for it. Thanks for having me on. Talk to you both soon. Absolutely. Talk soon. This information is for educational purposes only and is not a recommendation to buy, hold, or sell any investment or financial product. This podcast has been produced by a third party and may include paid promotional advertisements, other company references, and individuals unaffiliated with A16Z. Such advertisements, companies, and individuals are not endorsed by AH Capital Management LLC, A16Z, or any of its affiliates. Information is from sources deemed reliable on the date of publication, but A16Z does not guarantee its accuracy. Thank you. Thank you. you