Trump Takes Tech Chiefs to Beijing Summit Amid AI and Inflation Risks

Trump-Xi Summit, Space Data Center Debate, Inflation Up | Diet TBPN — TBPN

TBPN

The U.S. risks undermining its AI containment strategy and long-term Taiwan deterrence during high-stakes trade negotiations in Beijing while facing domestic stagflation.

AI-assisted summary of Trump-Xi Summit, Space Data Center Debate, Inflation Up | Diet TBPN — original source. Check the original for full context.

Source published:

Detailed notes

## Background Diet TBPN published this episode on YouTube on May 14, 2026. According to the page description, the 30-minute show delivers the best highlights from the daily TBPN tech talk show hosted by John Coogan and Jordi Hays, which covers major technology and industry news.

## TL;DR The episode breaks down the high-stakes Beijing summit between Donald Trump and Xi Jinping, spotlighting the prominent delegation of American corporate titans, semiconductor trade tensions, and shifting diplomatic postures around Taiwan. It examines Anthropic's selective rollout of Claude Mythos to Japanese banks alongside broader U.S. AI policy contradictions and China's rapid progress in open-source AI models. Additionally, the hosts evaluate futuristic compute strategies like space- and ocean-based data centers before concluding with a look at escalating inflation data and looming stagflation risks driven by energy shocks.

## Atmosphere & style The show blends sharp, quick-witted banter with high-level geopolitical and technical analysis, making complex economic and policy topics feel approachable and conversational. John Coogan and Jordi Hays trade playful jokes about executive fashion choices, airplane paint, and wild infrastructure concepts before effortlessly pivoting to serious commentary on trade wars and macroeconomic stress. The resulting vibe is energetic, candid, and plugged-in—feeling like an informal yet deeply insider debrief between tech-savvy peers.

## Key takeaways - [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) The Trump-Xi summit in Beijing features a high-profile U.S. corporate delegation including executives from Tesla, Apple, Nvidia, Boeing, and major financial institutions. - [01:10](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=70s) Nvidia CEO Jensen Huang joined the trip aboard Air Force One via Alaska, amidst confusion over initial media reporting regarding his invitation. - [04:45](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=285s) Executives like Jensen Huang and Palmer Luckey utilize consistent wardrobes to build distinct brand identities, though it presents long-term media retention risks. - [09:35](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=575s) The core trade negotiations center on securing U.S. access to rare earth minerals in exchange for predictable tariff policies and semiconductor export terms. - [12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) Key frontier AI lab leaders from OpenAI, Anthropic, and Google/DeepMind are notable omissions from the U.S. summit delegation. - [12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) Xi Jinping is seeking to shift the U.S. diplomatic stance on Taiwan toward explicit opposition to independence and veto power over arms sales. - [15:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=944s) Trump's personalized diplomacy creates geopolitical uncertainty for regional allies, particularly regarding Taiwan, China's ties to rogue states, and human rights issues like Jimmy Lai. - [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Anthropic is extending Claude Mythos access to major Japanese banks to strengthen national infrastructure against AI risks under Prime Minister Takaichi's directives. - [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) U.S. AI policy exhibits internal tension between granting restricted access to advanced models like Mythos for security while simultaneously allowing top semiconductor exports to China. - [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) China's open-source AI capabilities lag U.S. frontier models by an estimated six to twelve months, operating under concentrated funding models like DeepSeek. - [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) Google and SpaceX are discussing plans to partner on launching orbital data center satellites ahead of SpaceX's planned public listing. - [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) Tech companies are exploring alternative infrastructure environments, including deep-sea and space-based data centers, to bypass local energy constraints and regulatory friction. - [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) U.S. economic data shows accelerating inflation driven by energy shocks and Strait of Hormuz disruptions, with PPI reaching 6% and CPI reaching 3.8%. - [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) The macro environment reflects stagflation risks, where non-AI real GDP growth has slowed to 0.1% alongside elevated inflation expectations.

## Detailed walkthrough ### [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) Trump-Xi Jinping Beijing Summit and the Corporate "A-Team"

The discussion opens on the major news dominating the front page of the Wall Street Journal: the high-stakes summit between Donald Trump and Xi Jinping in Beijing. Both leaders arrived with massive delegations, turning geopolitics into a true "team sport." While political figures were part of the entourage, the hosts emphasize that the American business executives in attendance represent the real story. The delegation features a powerhouse lineup of corporate titans, including Elon Musk (Tesla), Tim Cook (Apple), Jensen Huang (Nvidia), Kelly Ortberg (Boeing), David Solomon (Goldman Sachs), Stephen Schwarzman (Blackstone), Larry Fink (BlackRock), Jane Fraser (Citi), and Dina Powell McCormick (Meta). The hosts take a playful jab at Boeing's inclusion, noting it is a "flex" given Boeing's enduring dominance in manufacturing large commercial aircraft, whereas drone giants like DJI have been unable to scale up to build a 747.

- **Corporate Titans Present:** Elon Musk, Tim Cook, Jensen Huang, Kelly Ortberg, David Solomon, Stephen Schwarzman, Larry Fink, Jane Fraser, and Dina Powell McCormick. - **Boeing's Flex:** Kelly Ortberg's presence underscores Boeing's long-standing lead in large aircraft manufacturing, contrasting it with drone makers like DJI.

---

[01:10](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=70s) The Speculation Surrounding Jensen Huang's Arrival

The conversation pivots to the logistics and drama surrounding Nvidia CEO Jensen Huang's trip. Speculation swirled after reports suggested he boarded Air Force One during a stopover in Alaska. The hosts address conflicting media reporting: CNBC originally claimed Jensen wasn't invited, prompting a correction from Trump on Truth Social where he clarified that while he hadn't invited Jensen personally, the Nvidia chief was indeed part of the trip. The hosts debate whether Alaska was just a convenient rendezvous point or if Jensen was a late addition due to dynamic scheduling conflicts. Given that Nvidia is the largest company in America and operates at the absolute epicenter of US-China chip sanctions and trade tensions, his presence on the flight—confirmed by Elon Musk—was vital for the trade dialogue.

The hosts contrast Silicon Valley's typical stakes with the sheer scale of global geopolitics. In tech, a $2 billion acquisition or a $40 billion revenue line is considered massive, but on the world stage, leaders are bargaining over a trillion-dollar oil market, global rare earth supplies, and critical semiconductor supply chains. The hosts react to grainy arrival footage from Beijing showing Trump, Musk, and Huang walking down the red carpet past flag-waving crowds, admiring the majestic presence of the 747 Air Force One while lamenting how the broader airline industry has crammed everyday travelers into cramped 737s.

- **Air Force One Logistics:** Elon Musk confirmed that he, Jensen Huang, and Donald Trump flew together on Air Force One after picking Jensen up in Alaska. - **Truth Social Mix-Up:** Trump corrected initial CNBC reports, confirming Jensen was welcome despite not receiving a direct personal invitation initially.

---

[04:45](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=285s) Jensen's Uniform, Executive Branding, and Aging Strategy

Inspecting Jensen Huang's travel wardrobe, the hosts joke about his packing list consisting entirely of leather jackets and iconic sneakers. They discuss how adopting a strict "uniform"—much like Steve Jobs's black turtleneck or Palmer Luckey's Hawaiian shirt—compounds into a powerful personal brand and lore. The hosts note a recent milestone for Luckey's company, Anduril, which raised a massive Series H funding round at a valuation over $50–60 billion, proving the power of iconic founder branding.

However, the hosts highlight a hidden risk of wearing the exact same outfit for decades: if an executive never changes their look, a viral clip of them saying something foolish 10 years ago will look like it was filmed yesterday. For instance, an AI CEO who gave a sci-fi rant about *The Terminator* a decade ago might have evolved their views today, but uninterrupted clothing choices make old hot takes look recent. To create visual distance over time, executives either need to overhaul their wardrobe or follow the "Jeff Bezos playbook"—getting radically jacked or going bald so the public immediately recognizes an older clip. The hosts joke about "Jensen at 320 lbs lean" wearing flowing leather trench coats or robes.

- **Anduril Series H:** Palmer Luckey's defense startup reached a Series H valuation benchmark of $51–$61 billion. - **The Bezos Playbook:** Changing physical appearance or wardrobe helps executives distance themselves from outdated media clips recorded years prior.

---

[07:42](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=462s) The Qatar "Golden Palace" Air Force One and Fuel Efficiency Trivia

The hosts review video footage of the opulent "Golden Palace" 747—the luxurious Air Force One jet gifted by Qatar. Inspecting its sky-yacht interior, which features high-end aesthetics and an onboard chessboard, they joke about whether world leaders use it to play literal 4D chess.

This leads to a tangent on aviation logistics and airplane paint economics. Dark colors like black absorb heat and burn significantly more fuel, which is why most commercial airlines stick to white exteriors. They poke fun at Spirit Airlines' bright yellow fleet, lightheartedly asking if the paint job contributed to their financial troubles. The hosts speculate whether a gold or mirrored exterior could reflect sunlight and boost fuel efficiency, noting that the United States Air Force has operated silver bombers and jets in the past.

- **Qatar Gifted 747:** The luxurious "Golden Palace" jet features custom sky-yacht interiors, including a chess setup. - **Aircraft Color Physics:** Black or dark paint jobs increase thermal absorption and fuel consumption, driving the industry standard toward white or reflective finishes.

---

[09:35](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=575s) Stakes of the Summit: Rare Earths, AI Guardrails, and Supermicro

Turning to the Wall Street Journal editorial board's analysis, the hosts dissect the core geopolitical stakes of the Beijing meeting. The official US narrative focuses on achieving "stability," a term central to Trump's personal brand as a self-proclaimed "stable genius." However, the WSJ warns that personal rapport cannot easily overcome Xi Jinping's broader strategic ambitions. On trade, the best realistic outcome is maintaining the status quo—a tariff truce coupled with assurances from Beijing not to hold the global rare earth supply hostage again. In return, Xi seeks access to advanced US semiconductor exports to bolster China's AI capabilities, where Beijing currently trails the US by an estimated six months.

The editorial notes that while the US administration wants to establish AI safety guardrails, formal arms-control agreements with Beijing are unlikely to be effective. Beijing has been engaging in "industrial-scale theft of American AI models," highlighted by recent Department of Justice indictments. The hosts specifically recall the indictment involving a tech executive associated with Supermicro, who allegedly ran a scheme shipping high-end AI chips to China disguised with fake labels and hairdryer/heater classifications to evade export controls.

- **AI Gap:** Current industry estimates place China's AI capabilities roughly 6 months behind the United States. - **Supermicro Indictment:** DoJ actions target illicit diversion schemes that used fraudulent shipping labels on hardware to smuggle high-end chips into China. - **Rare Earths Trade-off:** The US aims to secure stable access to rare earth minerals in exchange for predictable tariff policies.

---

[12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) Absent AI Leaders and the Taiwan "Flytrap"

The hosts point out a striking disconnect between Silicon Valley and the Beijing summit: key leaders from the top AI frontier labs are completely absent. Sundar Pichai and Demis Hassabis (Google/DeepMind), Dario Amodei (Anthropic), and Sam Altman (OpenAI) were not included in the delegation. While executives like Elon Musk, Jensen Huang, and Meta's Dina Powell McCormick are present, the absence of frontier lab CEOs means the summit is focused on immediate semiconductor supply chains and trade mechanics rather than grand existential questions about post-AGI/ASI geopolitics, fast-takeoff scenarios, or the "China Wakes Up 2027" timeline. Addressing a viewer comment about Meta's presence—given its banned status in China—the hosts suggest Meta is maintaining a footprint due to its $2 billion acquisition of Manus, though $2 billion remains a rounding error on the global stage.

Finally, the hosts detail the diplomatic "Venus flytrap" Xi Jinping is setting regarding Taiwan. Xi is pushing the US to formally *oppose* Taiwanese independence (moving away from the historic posture of simply *not supporting* it) and seeking veto power over US arm sales to the island. While Xi will frame this as a minor wording tweak that lets Trump claim credit for securing regional peace, the WSJ warns it would dismantle decades of delicate US diplomacy that has successfully deterred conflict in the Taiwan Strait.

- **Missing AI CEOs:** Sam Altman (OpenAI), Dario Amodei (Anthropic), Demis Hassabis (DeepMind), and Sundar Pichai (Google) were absent from the delegation. - **Taiwan Pivot:** Xi Jinping is lobbying for the US to shift its posture from "not supporting" to explicitly "opposing" Taiwanese independence, alongside gaining a veto on US weapons sales. - **Meta's China Footprint:** Meta's presence is tied in part to its $2 billion Manus acquisition and ongoing interest in regional tech talent.

Here is a detailed, faithful walkthrough of this section of the transcript:

[15:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=944s) Trump-Xi Summit Analysis and Geopolitical Stakes

An economic or geopolitical conflict over Taiwan would be a total catastrophe. Chinese President Xi Jinping will likely be watching closely during his upcoming summit with Donald Trump to see if the U.S. President signals any hesitation to defend Taiwan in a crisis. Because Trump's diplomacy is deeply personal and unpredictable, regional allies like Japan are watching with extreme anxiety. Observers note that skipping a stop in Tokyo prior to visiting Beijing could be seen as a strategic mistake, as a visit to Japan would signal U.S. solidarity. Furthermore, while Trump has stated he will raise the case of political prisoner Jimmy Lai—the 78-year-old Hong Kong publisher convicted on dubious charges—Xi is unlikely to release him unless he believes the request is more than a mere token gesture.

The U.S. also seeks China's cooperation on Iran, or at least an agreement for Beijing to stop actively assisting Tehran. Retired Navy Rear Admiral Mark Montgomery recently suggested that Trump should directly confront Xi on whether China is supplying intelligence to Iran. A *Wall Street Journal* editorial highlighted that the Chinese Communist Party remains the primary industrial and financial backbone for rogue states ranging from Russia to Iran and North Korea. While the first Trump administration viewed China as a broad strategic adversary across military, economic, and ideological domains, the second administration appears to be seeking a détente, with Trump acting as the "chief dove." This approach could be viable if the U.S. uses the truce to diversify its supply chain for rare earth elements and pass a $1.5 trillion defense budget to rearm. Meanwhile, commentators speculate that major deals have already been struck behind closed doors, making the high-profile summit largely ceremonial—since bringing top industrial leaders on such a trip without guaranteed progress would be a major public embarrassment.

- China continues to serve as the main industrial base and financier for Russia, Iran, and North Korea. - A proposed U.S. defense budget of $1.5 trillion would be required to rearm effectively during any diplomatic détente. - Jimmy Lai is 78 years old and remains imprisoned in Hong Kong on politically motivated charges.

---

[17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Anthropic's Model Access, U.S. AI Policy, and the "Mythos Moment" in China

In major tech policy developments, Anthropic is granting Japan's three mega-banks—MUFG Bank, Sumitomo Mitsui Banking Corp (SMBC), and Mizuho Bank—access to Claude Mythos as early as late May. Mythos is a high-powered AI model capable of discovering and exploiting software security flaws significantly faster than prior tools. Access was previously restricted to roughly 50 elite organizations worldwide, including U.S. banks and U.K. government bodies. The decision follows instructions from Japanese Prime Minister Sanae Takaichi to harden national infrastructure against advanced AI risks, and access was likely finalized during a meeting with U.S. Treasury Secretary Scott Bessent.

This decision highlights a glaring contradiction in U.S. policy. While Secretary Bessent personally vets and approves every organization granted access to Mythos due to severe national security risks, Nvidia CEO Jensen Huang is simultaneously flying on Air Force One with President Trump to Beijing to sell AI chips to China. These chips could allow China to develop its own Mythos-level capabilities. Critics like Chris Maguire argue that the administration’s AI policy is incoherent, though others suggest this strategy is part of a larger, ongoing trade negotiation.

Anthropic CEO Dario Amodei has estimated that Chinese open-source AI efforts are only about six months behind models like Mythos. This raises questions about how a "Mythos moment" will unfold in China. In the U.S., Anthropic faced intense pushback and nationalization threats from government officials (including the Department of War and Emil Michael) before a looser democratic balance was established. In contrast, China’s AI landscape features closely held companies like High-Flyer (the parent company of DeepSeek). DeepSeek's founder maintains majority voting and economic control, having invested $1 billion of his own money with minimal dilution—a stark contrast to Anthropic's heavily diluted, multi-stakeholder structure. If Chinese models catch up, the Chinese Communist Party may exert aggressive control over these private founders, potentially consolidating hardware like Nvidia Blackwell GPUs into centralized, state-run data centers.

- The three Japanese banks receiving access to Claude Mythos are MUFG Bank, Sumitomo Mitsui Banking Corp, and Mizuho Bank. - Claude Mythos was previously restricted to around 50 corporations and government organizations worldwide. - Estimates place China's open-source AI capabilities roughly six to twelve months behind frontier U.S. models. - DeepSeek's founder funded the project with $1 billion of personal capital, maintaining firm control over the entity.

---

[23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) Space and Marine Data Centers

SpaceX and Google are in talks to partner on launching orbital data centers, providing a significant boost to SpaceX as it prepares for a historic public listing. Google is negotiating rocket launch deals to deploy tiny machine racks into orbit as part of its "Project Labs" moonshot initiative, aiming to test prototype satellites by 2027. Google CEO Sundar Pichai noted that while the timeline is cautious, orbital data centers could be viewed as a standard architecture in a decade or so. Google remains an early investor in SpaceX, holding a 6.1% equity stake. Additionally, because the OpenAI Startup Fund was an early investor in Cursor's seed round, OpenAI indirectly holds financial exposure to SpaceX, creating an intertwined web where major tech competitors hold stakes in each other.

The discussion also touches on alternative locations for massive AI computing infrastructure to avoid local regulatory hurdles and "NIMBYism" (Not In My Back Yard). The hosts banter about deploying blimp-based data centers over the Pacific Ocean—noting that Alphabet co-founder Sergey Brin owns a blimp company and Amazon's Jeff Bezos has Blue Origin—or placing data centers at the bottom of the ocean. While deep-sea deployments benefit from free space and abundant ocean cooling, powering them presents a challenge: burning fossil fuels underwater would require submerged diving bells or oxygen supplies, leaving geothermal energy as the primary alternative.

- Google holds a 6.1% stake in SpaceX from an early investment round. - Google’s Project Labs aims to launch prototype data center satellites into orbit by 2027. - SpaceX’s upcoming IPO this summer is anticipated to be the largest public listing of all time.

---

[26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) Accelerating Inflation, Energy Shocks, and Stagflation Risks

Fresh economic data reveals a worrying spike in inflation, with consumer prices accelerating to 3.8% and April producer prices jumping 6% year-over-year (far exceeding the 4.8% consensus estimate). A major catalyst is the closure of the Strait of Hormuz, which has created severe energy market shocks and supply chain bottlenecks. Beyond energy, the crisis is impacting raw materials like aluminum due to Middle Eastern smelter disruptions, leading to shortages of everyday goods like canned Diet Coke.

Surging energy prices have driven inflation expectations to multi-year highs. The break-even rate—the yield gap between standard U.S. Treasuries and Treasury Inflation-Protected Securities (TIPS)—has reached its highest point since October 2022, signaling that investors expect annual inflation to average 2.7% over the next five years. This setup points toward a dangerous stagflation scenario: while AI capital expenditures and data center investments mask overall economic weakness, the "real economy" is growing at a sluggish 0.1% while burdened by nearly 3% inflation. As Federal Reserve Chair Jerome Powell steps down after navigating previous economic turbulence, incoming leadership like Kevin Warsh will face an extraordinarily difficult macroeconomic environment.

- April Producer Price Index (PPI) surged to 6.0% year-over-year versus a 4.8% estimate. - Consumer Price Index (CPI) inflation accelerated to 3.8%. - The bond market break-even rate indicates expected average inflation of 2.7% over the next five years. - Non-AI real GDP growth has slowed to just 0.1%, raising concerns over looming stagflation.

## Notable quotes > [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) A flex given Boeing's enduring dominance in manufacturing large commercial aircraft, whereas drone giants like DJI have been unable to scale up to build a 747.

> [01:10](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=70s) Silicon Valley's typical stakes with the sheer scale of global geopolitics. In tech, a $2 billion acquisition or a $40 billion revenue line is considered massive, but on the world stage, leaders are bargaining over a trillion-dollar oil market, global rare earth supplies, and critical semiconductor supply chains.

> [04:45](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=285s) If an executive never changes their look, a viral clip of them saying something foolish 10 years ago will look like it was filmed yesterday.

> [04:45](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=285s) The Jeff Bezos playbook—getting radically jacked or going bald so the public immediately recognizes an older clip.

> [07:42](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=462s) Dark colors like black absorb heat and burn significantly more fuel, which is why most commercial airlines stick to white exteriors.

> [09:35](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=575s) Industrial-scale theft of American AI models

> [12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) China Wakes Up 2027

> [12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) Venus flytrap

> [15:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=944s) Chief dove

> [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Claude Mythos as early as late May. Mythos is a high-powered AI model capable of discovering and exploiting software security flaws significantly faster than prior tools.

> [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Mythos moment

> [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) Project Labs

> [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) Real economy is growing at a sluggish 0.1% while burdened by nearly 3% inflation.

## Mentioned Alphabet, Amazon, Anduril, Anthropic, Apple, BlackRock, Blackstone, Blue Origin, Boeing, Beijing, China, Citi, Claude Mythos, Cursor, DJI, Dario Amodei, David Solomon, Demis Hassabis, Department of Justice, Department of War, Diet Coke, Dina Powell McCormick, Donald Trump, Elon Musk, Emil Michael, Federal Reserve, Goldman Sachs, Google, Google DeepMind, High-Flyer, DeepSeek, Iran, Japan, Jane Fraser, Jeff Bezos, Jensen Huang, Jerome Powell, Jimmy Lai, John Coogan, Jordi Hays, Kelly Ortberg, Kevin Warsh, Larry Fink, MUFG Bank, Mizuho Bank, Manus, Mark Cuban, Mark Montgomery, Mark Zuckerberg, Meta, North Korea, Nvidia, Nvidia Blackwell, OpenAI, OpenAI Startup Fund, Palmer Luckey, Qatar, Russia, SMBC, Sam Altman, Sanae Takaichi, Satya Nadella, Scott Bessent, Sergey Brin, Silicon Valley, SpaceX, Spirit Airlines, Stephen Schwarzman, Sundar Pichai, Supermicro, TBPN, Taiwan, Tesla, The New York Times, Tim Cook, United States Air Force, Wall Street Journal, Xi Jinping

[Source](https://www.youtube.com/watch?v=3_-mzZ21RD4)

中文简报

特朗普携科技巨头赴北京峰会 面临AI与通胀风险 面对国内滞胀压力,美国在北京的高风险贸易谈判中可能危及其AI遏制战略以及对台长期威慑力。 本期节目深入剖析了唐纳德·特朗普与习近平在北京举行的高风险峰会,重点探讨了阵容显赫的美国企业巨头代表团、半导体贸易紧张局势以及围绕台湾的外交姿态转变。节目审视了Anthropic向日本各银行定向推出Claude Mythos的情况,以及更广泛的美国AI政策矛盾和中国在开源AI模型上的快速进展。此外,主持人还评估了太空和海洋数据中心等未来算力战略,最后探讨了能源冲击推动下不断攀升的通胀数据以及迫在眉睫的滞胀风险。 特朗普与习近平在北京举行高级别峰会,马斯克、库克、黄仁勋等美方商业巨头随行,核心经贸谈判聚焦于确保美国稀土矿产供应与半导体出口及关税稳定。 (0:02) Sam Altman、Dario Amodei 及 Sundar Pichai 等前沿 AI 实验室高管罕见缺席本次北京峰会代表团,使讨论焦点更偏向既有芯片供应链而非通用人工智能(AGI)的地缘政治影响。 (12:30) 遵循日本首相高市早苗强化国家关键基础设施安全防御的指令,Anthropic 决定向三菱日联、三井住友和瑞穗三大日本银行开放 Claude Mythos 访问权限。 (17:47) 受霍尔木兹海峡关闭引发的能源冲击影响,美国 4 月 PPI 同比跃升 6.0%、CPI 升至 3.8%,在非 AI 领域实际 GDP 仅增长 0.1% 的背景下滞胀风险加剧。 (26:44)

Reading summary

The high-stakes Beijing summit between Donald Trump and Xi Jinping reflects the heavy entanglement of U.S. corporate giants in geopolitical bargaining, even as Washington faces growing internal contradictions in AI export controls and worsening macroeconomic stagflation. Corporate Titans and Geopolitics at Beijing Summit (0:02) The Beijing summit between Donald Trump and Xi Jinping took on the character of an American corporate showcase, featuring an executive delegation that included Elon Musk (Tesla), Tim Cook (Apple), Jensen Huang (Nvidia), Kelly Ortberg (Boeing), David Solomon (Goldman Sachs), Stephen Schwarzman (Blackstone), Larry Fink (BlackRock), Jane Fraser (Citi), and Dina Powell McCormick (Meta). Commentators interpreted Ortberg's presence as an intentional demonstration of Boeing's entrenched lead in large commercial aircraft manufacturing over regional drone and aviation competitors. Logistical attention centered on Nvidia CEO Jensen Huang, who boarded Air Force One during an Alaska refueling stop alongside Trump and Musk following contradictory reports from CNBC and a clarification on Truth Social. With Nvidia positioned at the epicenter of U.S.-China semiconductor sanctions, Huang's presence elevated discussions from standard tech acquisitions to high-stakes negotiations over critical microchip supply chains and global resource flows. Bargaining Over Taiwan, Rare Earths, and Defense (9:35) According to the Wall Street Journal editorial board, the realistic ceiling for the summit is maintaining the status quo: negotiating a tariff truce and securing reliable access to Chinese rare earths in exchange for expanded semiconductor exports. China's AI ecosystem is estimated to lag the U.S. by roughly six months, and Beijing continues seeking high-end chips despite Department of Justice indictments targeting illicit diversion schemes, such as mislabeling hardware as heaters by individuals linked to Supermicro. Meanwhile, top frontier AI laboratory leaders—including Sam Altman (OpenAI), Dario Amodei (Anthropic), Demis Hassabis (DeepMind), and Sundar Pichai (Google)—were absent, leaving talks focused on hardware supply chains rather than long-term artificial general intelligence governance. Xi Jinping is also lobbying Washington to shift its formal posture from 'not supporting' to explicitly 'opposing' Taiwanese independence, alongside demanding veto power over U.S. arms sales. Analysts warn this concession would dismantle decades of regional deterrence. Commentators noted that while an administration détente could theoretically buy time for the U.S. to diversify critical mineral supplies and fund a proposed $1.5 trillion defense budget, China remains the core industrial and financial backer of Russia, Iran, and North Korea. Contradictions in Frontier AI Export Policy (17:47) U.S. technology strategy showed visible internal divergence as Anthropic granted Japan's three mega-banks—MUFG Bank, Sumitomo Mitsui Banking Corp (SMBC), and Mizuho Bank—access to its advanced Claude Mythos model following directives from Japanese Prime Minister Sanae Takaichi. Because Mythos can identify and exploit software vulnerabilities far faster than prior tools, U.S. Treasury Secretary Scott Bessent has personally vetted authorized institutions due to national security risks. Critics pointed out the contradiction between Washington heavily restricting allied access to Mythos while simultaneously permitting Nvidia to negotiate advanced chip sales in Beijing that could assist China in reaching parity. Anthropic CEO Dario Amodei estimates Chinese open-source models trail leading U.S. systems by only six to twelve months. Unlike Anthropic's multi-stakeholder governance, Chinese competitors like DeepSeek rely on concentrated founder control—backed by $1 billion in personal capital—raising concerns that the Chinese state could swiftly consolidate private hardware, including Nvidia Blackwell chips, into centralized facilities. Orbital Computing and Intertwined Tech Investments (23:21) To bypass terrestrial power constraints and local regulatory friction, tech companies are exploring alternative data center environments. SpaceX and Google are negotiating launch agreements to test prototype satellite server racks in orbit by 2027 under Google's Project Labs. Although Google CEO Sundar Pichai indicated that orbital infrastructure remains roughly a decade away from becoming a standardized architecture, the talks bolster SpaceX ahead of an anticipated initial public offering. The venture also illustrates the intertwined financial relationships among major tech rivals: Google maintains an early 6.1% equity stake in SpaceX, while OpenAI holds indirect financial exposure to SpaceX via the OpenAI Startup Fund's early investment in Cursor. Energy Shocks and Looming Stagflation Threats (26:44) Macroeconomic data points to accelerating stagflation risks in the United States, exacerbated by the closure of the Strait of Hormuz. The resulting energy shock and supply chain snarls have spilled into basic commodities, causing disruptions in aluminum production and packaging shortages. April Producer Price Index inflation surged to 6.0% year-over-year against a 4.8% consensus forecast, while Consumer Price Index inflation rose to 3.8%. Bond markets responded by pushing the five-year Treasury-TIPS break-even rate to 2.7%, its highest level since October 2022. While heavy capital spending on AI infrastructure continues to buoy headline economic figures, real GDP growth across the non-AI economy has decelerated to 0.1%, presenting an extraordinarily difficult landscape for incoming Federal Reserve leadership following Jerome Powell's planned departure. While the Beijing summit attempts to establish a transactional balance on tariffs and critical minerals, inconsistent technology security policies and mounting stagflationary pressures from Middle Eastern energy disruptions leave both strategic détente and domestic economic stability highly vulnerable.

精炼阅读摘要

本期节目聚焦特朗普与习近平的北京峰会及随行的美国商业巨头代表团,探讨了中美在半导体与稀土供应链博弈中的复杂考量,并深入分析了美国人工智能政策的内部矛盾、太空算力基建布局以及地缘能源冲击带来的滞胀风险。 中美峰会博弈:商业领袖随行与地缘红线 (0:02) 特朗普与习近平在北京举行高规格峰会,美方代表团汇聚了Elon Musk、Tim Cook、Jensen Huang、Kelly Ortberg等顶级商业领袖。讨论指出,相较于科技行业数十亿至数百亿美元的交易规模,本次峰会触及的是万亿美元能源市场、全球稀土供应链以及尖端半导体管控等核心利益。华尔街日报社论分析认为,美方的务实目标是维持关税休战并确保中国不切断稀土供应,而中方则寻求获取先进芯片以缩小在AI领域约6个月的差距。值得注意的是,Sam Altman、Dario Amodei等AI前沿实验室负责人悉数缺席,表明峰会议程更多集中在当下的半导体与贸易机制,而非通用人工智能的未来治理。 在地缘政治层面,讨论指出习近平正推动美国将其在台湾问题上的表述从“不支持”独立转变为明确“反对”,并寻求对美对台军售拥有否决权。此外,特朗普虽表态将提及被监禁的78岁香港出版商黎智英(Jimmy Lai),但观察人士认为中方不太可能轻易让步。社论提醒,尽管第二届特朗普政府表现出寻求缓和的姿态,但中国依然是俄罗斯、伊朗和朝鲜的重要工业与金融支柱,若要实现长期制衡,美国可能需要依托高达1.5万亿美元的国防预算重建军备并在外交谈判中保持威慑。 美国AI政策矛盾与中美模型能力竞赛 (17:47) 节目深入剖析了美国AI安全监管与出口政策之间的脱节现象。一方面,Anthropic正计划向日本三菱日联金融集团(MUFG)、三井住友银行(SMBC)和瑞穗银行(Mizuho)开放高权限安全模型Claude Mythos,该模型因具备快速发现并利用软件漏洞的能力,此前全球仅有约50家精英机构获准使用,且每项申请均需经美国财政部长Scott Bessent亲自审查。但另一方面,英伟达CEO Jensen Huang却随同特朗普专机访华以促成芯片销售,评论人士批评这种一方面严控模型输出、另一方面却可能助推对手获得先进算力硬件的做法存在逻辑不连贯。 针对两国AI竞赛格局,Anthropic CEO Dario Amodei此前评估中国开源AI能力仅落后美国前沿模型约6至12个月。然而两国的组织架构截然不同:美国前沿实验室历经多轮股权稀释与监管施压;而像DeepSeek(幻方量化旗下)这类中国团队,其创始人自掏10亿美元个人资金并持有绝对控制权。讨论推测,随着技术差距缩小,中国官方未来可能对民营AI实体施加更集中的战略控制,甚至将Nvidia Blackwell等高端硬件整合成国家级统一算力中心。 算力空间转移:轨道与深海数据中心的前景 (23:21) 为规避陆地监管限制与电力瓶颈(即“邻避效应”),科技巨头正在探索非常规的数据中心基础设施方案。Google正与SpaceX展开谈判,计划在即将启动的“Project Labs”登月项目中,通过火箭将小型机架送入太空轨道,并计划于2027年发射原型测试卫星。Google CEO Sundar Pichai认为,尽管时间表较为保守,但轨道数据中心在十年后有望成为标准架构之一。值得一提的是,Google持有SpaceX 6.1%的早期股权,而OpenAI Startup Fund因早期投资Cursor也间接对SpaceX保有财务敞口,显现出竞争对手之间错综复杂的资本纽带。 除近地轨道外,节目还探讨了海洋作为替代场景的可行性。深海部署具备天然且无限的海水冷却优势与空间容量,但水下能源供给构成重大挑战:在深水燃烧化石燃料必须依赖水下潜水钟或额外氧气管道,这使得地热能成为目前深海部署少数可行的替代动力来源之一。 能源冲击与宏观滞胀:实体经济增长逼近停滞 (26:44) 最新公布的宏观数据显示美国通胀正全面反弹。受霍尔木兹海峡关闭引发的能源冲击与供应链中断影响,美国消费者物价指数(CPI)加速攀升至3.8%,而4月生产者物价指数(PPI)同比暴涨6.0%,大幅超出市场预期的4.8%。中东地区冶炼厂的中断不仅推高了能源成本,还直接波及铝等基础原材料的供给,导致包括罐装健怡可乐在内的日常消费品出现短缺。 债券市场上,美债与通胀保值债券(TIPS)之间的损益平衡通胀率已攀升至2022年10月以来的最高点,表明投资者预期未来5年的年化通胀率将维持在2.7%左右。讨论指出,虽然庞大的AI资本支出与数据中心建设在表面上支撑了整体GDP,但扣除AI相关投入后的“真实经济”实际增长率仅为0.1%,美国经济正滑向典型的高通胀、低增长“滞胀”区间。在美联储主席Jerome Powell即将离任之际,新任领导人如Kevin Warsh将面临极其严峻的宏观政策困局。 本期节目勾勒出全球地缘政治与宏观经济交织的复杂图景:中美高层互动试图在关税与核心技术供应链上寻求临时妥协,但AI芯片与安全政策的割裂、前沿基建高昂的资源代价,以及中东局势催生的滞胀风险,均使得未来的科技竞赛与宏观经济前景充满高度不确定性。

Detailed notes — zh

## Background 本内容由科技对谈节目 TBPN 于 2026 年 5 月 22 日发布在 YouTube 平台。页面简介显示,这是由 John Coogan 和 Jordi Hays 主持的精简版播客《Diet TBPN》,旨在用 30 分钟提炼当天直播节目的精华内容,聚焦科技、商业与地缘政治的核心动向。 ## TL;DR 本期节目首先深入剖析了特朗普率领马斯克、黄仁勋等全明星商界代表团访华的“特习会”,探讨了稀土博弈、高端芯片走私与美中前沿 AI 竞争的复杂现实。随后,节目梳理了 Anthropic 向日本三大银行开放 Claude Mythos 权限背后美方对华政策的矛盾,以及 SpaceX 联手 Google 布局轨道数据中心的太空算力构想。最后,主持人们聚焦霍尔木兹海峡危机引发的铝供应链断裂、通胀与 PPI 飙升,警示除 AI 算力投资繁荣外实体经济仅微增 0.1% 的严峻滞胀隐患。 ## Atmosphere & style 节目整体节奏明快轻盈、幽默辛辣,两位主持人默契十足,在硅谷极客黑话、网络梗图与严肃宏观地缘分析之间无缝切换。他们既能把空军一号抵京画面调侃为《沙丘》电影般的末世史诗,脑洞大开地畅想“平流层飞艇算力机房”和黄仁勋的肌肉转型,又能迅速收敛嬉笑,一针见血地指出美国在 AI 出口管制上的自相矛盾与华尔街滞胀阴霾。收听体验兼具前沿科技沙龙的机敏跳脱与财经硬核复盘的现实刺痛感。 ## Bulletin - [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) 特朗普携马斯克、黄仁勋、库克等商业代表团抵京举行特习会,聚焦供应链稳定与关税休战,但前沿大模型实验室核心负责人集体缺席。 - [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Anthropic 计划向三菱日联、三井住友和瑞穗等日本三大银行开放 Claude Mythos 模型访问权限,这是东亚机构首次获准调用该顶级网络安全与代码漏洞攻防模型。 - [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) SpaceX 与 Google 正深入谈判合作发射入轨数据中心,Google 计划通过 Project Labs 于 2027 年部署原型微型服务器卫星,推进太空轨道算力常态化。 - [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) 霍尔木兹海峡航运受阻导致全球电解铝断供并传导至消费端,美国 4 月 PPI 同比飙升至 6.0%,剔除 AI 资本支出后的实体经济增速仅 0.1%,市场滞胀风险急剧升温。 ## Key takeaways - [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) 特朗普北京峰会的商界随行团涵盖特斯拉、英伟达、苹果、波音及华尔街头部机构负责人,显示出经贸与技术筹码在元首外交博弈中的核心地位。 - [01:10](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=70s) 英伟达 CEO 黄仁勋与马斯克同乘空军一号飞抵北京,在全球芯片出口管制与供应链脱钩的风口浪尖下,半导体成为百亿甚至万亿级地缘谈判的核心焦点。 - [04:42](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=282s) 科技高管长期维持固定服饰符号有助于构筑个人品牌,但缺乏外形变动易造成历史旧言论与当前立场的认知混淆,引发类似贝索斯式形象转型的探讨。 - [08:58](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=538s) 美方以“寻求稳定”定调峰会目标,跨国企业界核心诉求是平息关税与贸易剧震,但中方可能以稀土供应和民机农产品采购为筹码,换取高阶 AI 芯片禁运的松绑。 - [12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) OpenAI、Anthropic 与 Google DeepMind 等核心前沿 AI 掌门人集体缺席北京峰会,凸显出硅谷极客圈对超级智能的远景探讨与大国现实技术外交之间的脱节。 - [14:20](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=860s) 北京在台湾问题上寻求推动美方从“不支持台独”微调为“正式反对台独”并争取对台军售否决权,此类微妙措辞若被打破可能根本性动摇数十年的台海脆弱均势。 - [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Anthropic 首次将具有极高漏洞攻防能力的 Claude Mythos 模型授权给日本三大金融巨头,体现出美日同盟在前沿战略 AI 网络防御领域的深度协同。 - [18:53](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1133s) 白宫一方面严苛审查顶尖 AI 模型的境内外使用资质,另一方面又放行头部芯片企业赴华谈判,凸显出美国对华高科技战略中的双轨矛盾与现实利益妥协。 - [18:53](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1133s) DeepSeek 背后的幻方量化由创始人以自有资金高度集权控制,与 Anthropic 股权高度稀释并受公开市场监管的模式形成反差,在中美两种不同制度下催生了截然不同的政企博弈机制。 - [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) Google 计划在 2027 年发射轨道服务器原型机,并探讨与 SpaceX 展开轨道算力合作,试图通过太空部署规避地面数据中心的监管、土地与邻避效应瓶颈。 - [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) 美伊冲突引发霍尔木兹海峡断航危机,不仅促使 4 月 PPI 年率飙升至 6.0%,还导致电解铝等金属包装材料紧缺,通胀压力迅速蔓延至终端消费品。 - [28:18](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1698s) 美债盈亏平衡通胀率升至 2022 年底以来最高水平;若剔除 AI 数据中心拉动的资本支出,实体经济实际增速仅 0.1%,经济正显现出低增长叠加高通胀的典型滞胀征兆。 ## Detailed walkthrough ### [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) 北京“特习会”与全明星商业代表团 本周最重磅、占据各大媒体头条的新闻,当属《华尔街日报》头版报道的特朗普与中方领导人在北京举行的首脑峰会。两位领导人正在这场地缘政治博弈中互争胜负。除了少部分政界随行人员外,真正引人注目的是特朗普带来的庞大美国商界领袖代表团,阵容堪称全明星级别。 代表团囊括了美国科技、制造业与华尔街的顶级领军人物:特斯拉的埃隆·马斯克(Elon Musk)、苹果的蒂姆·库克(Tim Cook)、英伟达的黄仁勋(Jensen Huang)、波音首席执行官凯利·奥特伯格(Kelly Ortberg)。主持人调侃波音此行颇有“秀肌肉”的意味,毕竟波音长期掌控大型客机制造领域,而像“大疆(DJI)至今也没能造出一架波音747”。此外,金融巨头亦全员出动,包括高盛的大卫·所罗门(David Solomon)、黑石集团的苏世民(Stephen Schwarzman)、贝莱德的拉里·芬克(Larry Fink)、花旗集团的简·弗雷泽(Jane Fraser),以及代表 Meta 出席的迪娜·鲍威尔·麦考密克(Dina Powell McCormick)。 ```mermaid flowchart TD accTitle: 特朗普访华商业代表团构成 accDescr: 展示随同特朗普前往北京的美国主要商业领袖及其所属行业分布 subgraph Tech["科技与硬核制造"] T1["Elon Musk<br/>(Tesla)"] T2["Tim Cook<br/>(Apple)"] T3["Jensen Huang<br/>(Nvidia)"] T4["Kelly Ortberg<br/>(Boeing)"] end subgraph Finance["华尔街金融巨头"] F1["David Solomon<br/>(Goldman Sachs)"] F2["Stephen Schwarzman<br/>(Blackstone)"] F3["Larry Fink<br/>(BlackRock)"] F4["Jane Fraser<br/>(Citi)"] end subgraph SocialTech["社交与前沿巨头"] M1["Dina Powell McCormick<br/>(Meta)"] end Tech --- Delegation["特朗普北京峰会<br/>商业代表团"] Finance --- Delegation SocialTech --- Delegation ``` - **科技与工业领袖**:埃隆·马斯克(特斯拉)、蒂姆·库克(苹果)、黄仁勋(英伟达)、凯利·奥特伯格(波音)。 - **华尔街金融巨头**:大卫·所罗门(高盛)、苏世民(黑石)、拉里·芬克(贝莱德)、简·弗雷泽(花旗)。 - **社交网络巨擘**:迪娜·鲍威尔·麦考密克(Meta)。 ### [01:10](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=70s) 黄仁勋的空军一号之行与地缘政治棋局 关于英伟达首席执行官黄仁勋是否自始至终都在出访名单上,外界引发了诸多猜测。此前有消息称他是在阿拉斯加中途登机的,引发了人们对他究竟是临时加入还是仅因行程便利的讨论。CNBC 曾报道黄仁勋未获邀请,随后特朗普在 Truth Social 上反驳了这一报道,澄清是假新闻,但他本人并未亲自去单独邀约。作为全球市值最高公司的掌舵人,黄仁勋的行程极其复杂多变,随时都可能发生调整。 在地缘政治博弈中,英伟达的处境尤为关键。尽管苹果在中国业务庞大,但在中美芯片制裁与出口管制的风口浪尖上,英伟达承担了最密集的交锋与博弈焦点。与此同时,硅谷平时的交易规模往往在几十亿美元收购案或四百亿美元营收级别,然而一旦上升到大国峰会层面,桌上的筹码直接演变成“万亿美元级别的石油市场、全球稀土供应链”以及地缘政治冲突。最终,马斯克证实他、黄仁勋与特朗普同乘一架空军一号专机飞抵北京,成为当晚备受瞩目的焦点。 - **阿拉斯加登机传闻**:外界传言黄仁勋在阿拉斯加中途登上空军一号,但跨国日程调配向来充满变数。 - **Truth Social 辟谣**:特朗普发文驳斥 CNBC 关于黄仁勋未受邀的报道,直称其为假新闻。 - **博弈维度的悬殊**:硅谷视角下百亿级的业务,在地缘政治层面上会被万亿美元的能源和稀缺战略矿产议题所放大。 ### [03:24](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=204s) 空军一号降落北京与波音747的情怀 现场画面展示了空军一号夜间降落北京的景象,特朗普与马斯克、黄仁勋相继走下飞机,其他商业领袖则各自乘机抵达。红毯两侧有挥舞旗帜的欢迎队伍。现场画面质感颗粒感极强、光线昏暗,主持人笑称其镜头氛围颇具科幻大片《沙丘》(Dune)中厄拉科斯(Arrakis)星球的末世史诗感,甚至打趣像置身于某种神秘地带。 空军一号威严沉稳的外观引发了两位主持人对波音747的怀旧感慨。随着现代航空业经济模型的演变,航空公司越来越倾向于精简成本,导致乘客在长途跨洋之外很难再坐上空间开阔、极具气场的波音747,而是几乎全被“塞进了波音737狭窄的机舱里”。针对空军一号的保密性,退役机型虽然可以在总统图书馆或史密森尼学会等博物馆供公众参观,但现役机型的具体内部改装与安全规格仍然是高度机密。 - **同机抵京**:特朗普、马斯克与黄仁勋搭乘空军一号同机抵达,其余商界大佬分批乘坐其他专机。 - **航空工业变迁**:民航业出于燃油与运营成本考量全面倒向双发窄体机(如波音737),四发巨无霸波音747逐渐退出一线日常运营。 - **空军一号内景**:旧款机型已在多家总统图书馆公开展出,但现役专机规格处于严格保密状态。 ### [04:42](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=282s) 科技领袖的“制服”哲学与贝索斯式进化 聊到黄仁勋的出差行李清单,主持人调侃他箱子里大概只装了满满一箱标志性的黑色皮夹克,搭配经典球鞋。从史蒂夫·乔布斯(Steve Jobs)的三宅一生黑色高领衫、帕尔默·拉奇(Palmer Luckey)的夏威夷衬衫,再到黄仁勋从皮衣延伸到特定裤型和运动鞋的“全套制服”,科技大佬建立个人标志性符号(Lore)不仅能降低认知成本,更能构筑强大的个人品牌。顺带提及拉奇创办的国防科技独角兽 Anduril 当天完成新一轮融资,字母都快用完了——估值冲上了510亿至610亿美元的 H 轮。 然而,“十年如一日穿同一套衣服”也有显著的负面风险:如果一个人容貌保养得当且衣着完全不变,他在十年前发表的言论片段在今天被翻出来疯传时,公众会误以为是昨天的最新表态。这对于观念随着认知提升而演进的 AI 行业 CEO 来说尤为棘手(例如早年读完科幻小说担忧“终结者降临”,十年后深入研究系统有了更务实的观点,却因外貌完全没变而百口莫辩)。为了拉开时间跨度,打破制服诅咒的最佳解法就是外形剧变——正如杰夫·贝索斯(Jeff Bezos)当年从瘦弱程序员直接剃成光头并练成“筋肉猛男”。主持人打趣称,如果黄仁勋未来也能转型练成“320磅纯干货肌肉硬汉”,视觉效果必定极其炸裂。 - **标志性装扮案例**:乔布斯(黑色高领毛衣)、帕尔默·拉奇(夏威夷衫)、黄仁勋(黑色皮衣配球鞋)。 - **Anduril 融资进展**:拉奇的 Anduril 斩获全新 H 轮融资,估值已高达510亿至610亿美元区间。 - **制服陷阱与贝索斯策略**:完全不变的造型容易导致数年前的旧视频与当下言论混淆;改变造型(如贝索斯的光头、健身增肌)能天然拉开不同时期观点的距离。 ### [07:38](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=458s) 卡塔尔赠送的“黄金宫殿”与飞机涂装热力学 话题随后转向网络上曝光的一段被称为“黄金宫殿”的空军一号专机内部视频。该飞机据称由卡塔尔赠予,机舱内部极度奢华,但在外观上依旧保持了普通波音747的涂装风格。关于该专机最终是投入现役飞行还是送入总统图书馆陈列,两人展开了一番讨论。 由此引申到航空空气动力学与热力学的现实问题:机身涂装颜色对飞机的燃油经济性存在巨大影响。如果为了炫酷而将飞机漆成黑色,机体表面会吸收大量热辐射,导致空调负荷暴增、大幅消耗更多燃油,这也是绝大多数商业客机坚持通体喷涂白色的根本原因。主持人调侃精神航空(Spirit Airlines)选用了全黄色机身不知是否间接加速了其运营困境;两人还一本正经地脑洞大开,探讨是否可以用黄金涂层或将飞机包裹成“全反光镜面机身”来反射阳光、极致提升燃油效率,并列举了军方现役的部分银色高反射率轰炸机与战机作为对照。 - **专机“黄金宫殿”**:源自卡塔尔赠送的豪华改装波音747,内饰奢华,引发用途猜想。 - **机身涂装热力学**:深色机身吸热严重并显著增加燃油消耗,商业客机绝大多数采用白色涂装以最大化反射热辐射。 - **反光机身脑洞**:从镜面包装机身降低吸热,联想到美军历史及现役的各类无涂装银色/抛光金属战机与轰炸机。 ### [08:58](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=538s) AI 假想图与特习会前瞻:追求“稳定”还是各怀心思? 节目在一阵轻松的调侃中开场。主持人看着美国空军一号的画面,忍不住感叹这架飞机简直就是一艘“空中游艇”(sky yacht)。两人打趣机上是否配有国际象棋棋盘,甚至调侃整架飞机其实就是为了让领导人下“四维象棋”(4D chess)而量身打造的。话题很快转向了网络上热传的一组由 AI 生成的图像:画面中中国领导人、唐纳德·特朗普(Donald Trump)、埃隆·马斯克(Elon Musk)和蒂姆·库克(Tim Cook)在中国某处坐在一起喝着啤酒、抽着香烟,随后画面一转,英伟达首席执行官黄仁勋(Jensen Huang)也神采奕奕地从后方入场,“看起来状态相当不错”。 笑谈过后,主持人将焦点转向严肃的国际地缘政治,援引《华尔街日报》(The Wall Street Journal)社论对即将登场的“特习会”(Trump-Xi Summit)深层利害关系的深度剖析。特朗普周四在北京与中国领导人会晤,美方在会前释放的基调被包装为“寻求稳定”(search for stability)。但社论警告称,只要特朗普别一厢情愿地认为私人交情能够化解对方根深蒂固的反美战略意图,这种追求稳定的想法才算切合实际。本次峰会议题涵盖贸易、技术以及伊朗问题;在贸易层面,最理想的结局或许只是维持现状——达成关税休战协议,并确保中方承诺不再将全球稀土供应作为人质威胁。毕竟,无论市场还是商界领袖都不想要混乱,只要关税方案能长期保持平稳,大家总能找到应对办法。 - 美方对外释放的峰会基调定位为“寻求稳定”,希望为双边经贸和关税建立确定性。 - 《华尔街日报》指出,特朗普在会前被包装为寻求稳定的形象,但警告不应对个人交情过度乐观。 - 市场与跨国企业界的核心诉求在于避免混乱,普遍希望避免破坏性的供应链动荡。 ### [10:34](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=634s) 稀土勒索、芯片禁运与 AI 军备竞赛的现实断层 主持人模仿特朗普的招牌口吻调侃道:“听着伙计们,我要求的只是一点点稳定。我就想要稳定与冷静,不想要任何冲突。”毕竟特朗普过去经常自诩为“稳定的天才”(stable genius),稳定早就成了他的个人标签之一。然而社论指出,中国当前的经济疲软是促成合作的动力,中方可能会抛出购买更多美国农产品或波音客机的承诺来安抚特朗普。但这套承诺早已上演过多次,美国农民在此前贸易摩擦中失去的中国市场份额至今未能完全夺回。更关键的博弈在于技术领域——中方可能以稀土供应作为筹码,换取美国放宽对华高阶人工智能芯片的出口限制。 在人工智能领域,北京方面将其视作与美国竞争的决定性战场;业内普遍估计中国在 AI 层面仅落后美国大约六个月。尽管美方政府希望就“AI 护栏”(AI guardrails)与北京展开对话并保持沟通热线,但各方对达成真正的“AI 军控协议”并不抱太大希望。美国最根本的威慑手段依然是在基础模型与算力储备上维持绝对统治力。特朗普政府此前已就北京针对美国前沿大模型的“工业级规模盗窃”(industrial-scale theft)发出严厉警告;美国司法部也指控了包括超微电脑(Supermicro)相关高管在内的团伙,涉嫌通过张贴虚假货运标签(例如伪装成吹风机或加热器等普通家电)走私高端 AI 芯片。 ```mermaid flowchart TD accTitle: US-China Summit Strategic Bargaining Dynamics accDescr: A flowchart depicting the strategic trade-offs and tensions between the US and China during the summit. subgraph ChinaSide["中方筹码与诉求"] C1["稀土供应链出口管治"] --> Deal{"峰会博弈天平"} C2["承诺采购农产品与客机"] --> Deal C3["诉求:放宽先进制程 AI 芯片禁运"] --> Deal end subgraph USSide["美方防线与威慑"] U1["算力与基础大模型霸权"] --> Deal U2["打击工业级 AI 模型与芯片走私"] --> Deal U3["诉求:全球供应链稳定性与关税休战"] --> Deal end ``` ### [12:30](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=750s) 缺席的硅谷 AI 巨头与 Meta 的微妙处境 令人瞩目的是,在这场涉及地缘与技术重构的元首峰会中,美国一线前沿 AI 实验室的关键掌门人集体缺席。虽然代表硬件算力的马斯克和黄仁勋赫然在列,Meta 方面也有迪娜·鲍威尔·麦考密克(Dina Powell McCormick)出席,但 Google DeepMind 的戴米斯·哈萨比斯(Demis Hassabis)与桑达尔·皮查伊(Sundar Pichai)、Anthropic 的达里奥·阿莫代伊(Dario Amodei)以及 OpenAI 的萨姆·奥尔特曼(Sam Altman)均未现身。主持人分析,这些 AI 领袖与联邦政府及特朗普政府之间各有错综复杂的关系,缺席并不令人意外,但这凸显出硅谷极客圈与全球顶级地缘外交之间的深刻脱节。 硅谷眼下最关切的是所谓的“超级智能”(Superintelligence)、“AI 2027 场景”或者“奇点爆发”后的世界秩序;而北京峰会真正讨论的,依然是半导体供应链配额、出口管制细则等现实框架,根本无法解答“后 AGI/ASI 时代”美中关系何去何从。针对聊天室观众提问“Meta 旗下产品在中国早已被封禁多年,为何还会派代表前往”,主持人指出,这不仅关乎展现友好姿态与商业存在,更牵涉到 Meta 对 Manus 等具中资背景、迁往新加坡出海团队的收购风波。对于两国经济体而言,中国顶尖技术与团队能否出海落户新加坡后再被美资收购、北京将如何收紧高端人才与核心技术出口,都是触动整个行业的敏感神经,哪怕峰会通稿提及只言片语都会掀起行业巨浪。 - 核心前沿 AI 实验室(OpenAI、Anthropic、Google DeepMind)负责人集体缺席高层随行团。 - 算力巨头(NVIDIA、Tesla/xAI)代表在场,突显出地缘政策优先聚焦于硬件制造与供应链,而非长远的超级智能安全。 - Meta 虽遭本土应用封禁,但涉及新加坡出海企业收购案(如 Manus),引发两国关于科技人才流动与技术出口管制的关切。 ### [14:20](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=860s) 台湾议题的“捕蝇草陷阱”与大国外交博弈 《华尔街日报》将中方在本次峰会中为特朗普设置的策略比作台湾议题上的“捕蝇草陷阱”(Venus flytrap)。北京正竭力寻求对美台军售的否决权,并力推美方立场发生微妙却致命的质变——从现行的“不支持台独”(not supporting Taiwanese independence)转变为由官方定调的“正式反对台独”(formally oppose Taiwanese independence)。北京试图向特朗普吹风,声称这一字眼修改对美国无足轻重,同时迎合特朗普追求“在又一动荡地区促成和平”的虚荣心。然而,这种微调将彻底颠覆数十年来维持台海脆弱和平的精密外交框架。 台湾海峡的局势牵动着周边盟友的神经。虽然特朗普可能并不真正关心台湾的民主范式,但他极度反感在自己的任期内爆发直接危机,因为那将是一场“经济与地缘政治的灭顶灾难”。日本等亚太盟友正高度焦虑地密切注视,社论批评特朗普此行直接飞往北京而未提前在东京停靠展示美日团结是一大失策。此外,涉及被关押的 78 岁香港出版商黎智英(Jimmy Lai)的人权交涉、要求北京停止对伊朗提供情报及金融支持等议题,也都被置于谈判桌前。社论强调,第一届特朗普政府将中国定义为全方位战略对手,而第二届政府内部特朗普却扮演起“首席鸽派”,寻求某种形式的缓和。坊间甚至普遍推测,众多核心经贸协议早在幕后悄然敲定,这场峰会更像是一场不能空手而归的隆重仪式。 - 中方力推美方在台湾问题上的措辞从“不支持”转向“明确反对”,并寻求干预美台军售。 - 特朗普的外交风格高度依赖个人直觉与领导人关系,盟友日本因其行程未先访日而产生战略焦虑。 - 伊朗的情报协助与地区代理人网络、黎智英等案均被纳入地缘博弈的交换清单。 ### [17:47](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1067s) Anthropic 扩展 Mythos 权限:日本三大银行获核心网络安全 AI 在峰会波澜之外,另一则重磅技术动态聚焦于美国领先 AI 实验室的前沿部署。AI 初创公司 Anthropic 计划在五月底之前,正式向日本三大金融巨头——三菱日联银行(MUFG Bank)、三井住友银行(SMBC)和瑞穗银行(Mizuho Bank)开放其最新核心模型“Claude Mythos”的使用权限。这一消息极有可能是由美国财政部长斯科特·贝森特(Scott Bessent)周二在日本举行的闭门会议中亲自向日方通报的。 这也是东亚机构首次获得 Mythos 的访问权限。Mythos 作为一款在代码漏洞挖掘与自动化利用方面性能极强的顶级模型,此前全球仅有约 50 家机构获批使用,主要局限于美国本土顶尖金融企业、核心科技公司以及英国政府等极具战略互信的机构。此前,日本首相高市早苗已向内阁下达明确指示,要求全面强化排查日本关键基础设施中的网络安全漏洞,积极应对类似 Mythos 这种前沿尖端 AI 所带来的全新威胁格局。此次对日本金融支柱机构的授权,标志着美日同盟正将最核心的 AI 网络防御与攻防技术深度整合。 - 获得 Claude Mythos 权限的日本三大银行:三菱日联银行(MUFG)、三井住友银行(SMBC)与瑞穗银行(Mizuho)。 - 权限此前受到极严苛限制,全球仅约 50 家顶级机构(美企、美资大行及英国政府机构)拥有准入资格。 - 此次技术准入是在高市早苗政府全力提升国家级网络安全与关键基础设施防御的背景下达成,凸显美日在前沿战略 AI 上的协同。 ### [18:53](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1133s) 白宫对华 AI 政策的矛盾与中国版“Mythos 时刻”的推想 节目一开始聚焦于美国政府在人工智能政策上令人困惑的双轨策略。主持人引述了 Chris Maguire 的尖锐评论:一方面,美国财政部长斯科特·贝森特(Scott Bessent)正在亲自对获准使用美国最先进 AI 模型的每一家公司进行极其严苛的安全审查,理由是该模型若被滥用,将对国家安全构成毁灭性威胁;但另一方面,英伟达(NVIDIA)CEO 黄仁勋(Jensen Huang)当晚却与特朗普总统一同乘坐空军一号飞往北京,计划向中国出售高性能 AI 芯片——而这些芯片正是中国加速研发自身“Mythos”级顶尖模型所不可或缺的基础设施。主持人直言,政府的 AI 政策目前充斥着自相矛盾与逻辑断裂,“你根本无法在逻辑上同时为这两种截然相反的做法辩护”。不过,他们也承认背后的水很深,这可能属于更高层级地缘政治贸易谈判的一环。 在此背景下,主播们展开了一个更深层的讨论:若真如 Anthropic 首席执行官达里奥·阿莫代伊(Dario Amodei)所言,开源社区和中国在顶尖模型上仅落后 Mythos 大约六个月到一年,那么当中国的“Mythos 时刻”真正来临时,局面会演变成什么样?在美国,围绕前沿模型展开的是一场典型的制度博弈:国防部门官员、埃米尔·迈克尔(Emil Michael)与达里奥激烈交锋,白宫甚至一度威胁要祭出“关键供应链指定”(Supply Chain Designation)实行准国有化管制,但最终在民主制度与市场制衡下妥协,Anthropic 如今不仅照常在各大超大规模云厂商上运营,还将技术授权给日本。 然而,中国的权力结构与公司治理完全不同。深度求索(DeepSeek)背后的母公司幻方量化(High-Flyer)处于极其独特的处境:其创始人个人投入了数以十亿美元计的自有资金,几乎没有经历股权稀释,拥有压倒性的绝对投票权与经济控制权。这与 Anthropic 拥有多达 15 位联合创始人、股权高度分散、多轮稀释并在设立 SPV 后最终必须迈向上市(从而受到 SEC 全面监管)的商业模式截然相反。一个由私人高度集权控制的中国硬核 AI 团队,迎面撞上掌控力极强的国家权力,必然会触发前所未有的剧烈碰撞。 ```mermaid flowchart TD accTitle: 美中前沿 AI 治理与股权结构对比 accDescr: 展示 Anthropic 与 DeepSeek 在股权集中度与政府监管博弈方式上的根本差异 subgraph US["美国模式:Anthropic"] U1["创始人股权稀释 / 15位联合创始人"] --> U2["资本结构多元化 / 迈向 IPO 受 SEC 监管"] U2 --> U3["与国防部拉锯但保留市场自由度"] end subgraph CN["中国模式:DeepSeek / 幻方量化"] C1["创始人自有巨资支持 / 极低稀释"] --> C2["单一创始人掌控绝大多数投票权与经济权"] C2 --> C3["面临政府强力介入与全国算力统筹整合压力"] end ``` - **开源与前沿差距**:尽管最新基准测试图表显示开源模型与顶尖闭源前沿模型的绝对差距有略微拉大的趋势,但业内共识认为中国追赶的速度极快,六个月至一年的窗口期并不遥远。 - **算力集中的现实困境**:主持人抛出一个关键难题——若中国必须将所有流入境内的英伟达 Blackwell 架构芯片强行集中到单一巨型数据中心并组网训练,才能炼出 Mythos 级别的通用大模型,那么当局还能否在科技产业中维持表面上的“鼓励多元竞争”? --- ### [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) 太空与飞艇数据中心:SpaceX、Google 的轨道算力博弈与深海脑洞 话题随后转向彭博社关于太空计算的一则重磅独家:SpaceX 与 Google 正处于深入谈判阶段,计划合作发射入轨数据中心。这笔重磅交易预计将为 SpaceX 在今年夏天进行其号称“史上规模最大”的首次公开募股(IPO)之前注入一剂强心针。作为搜索与云巨头的 Google,正在通过一个名为“Project Labs”的前沿登月计划,研发搭载微型服务器机架的原型卫星,并计划在 2027 年发射升空。Google 首席执行官桑达尔·皮查伊(Sundar Pichai)对此给出了谨慎而坚定的预期:“毫无疑问,大约十年之后,我们将把轨道部署视为建设数据中心的一种更常态化方式。” 演播室的气氛随即被各种异想天开的科技狂想点燃。主播们打趣道,既然要把数据中心送上天,为什么没人讨论“飞艇数据中心”?Google 联合创始人谢尔盖·布林(Sergey Brin)旗下本就拥有一家飞艇公司(LTA Research),而杰夫·贝索斯(Jeff Bezos)拥有蓝色起源(Blue Origin)与 AWS,逻辑上早已闭环。“我们应该在平流层飞艇上塞满 GPU,然后开启流水线大规模量产!在太平洋上空飘浮着一圈飞艇服务器,既免去了地面的‘邻避效应’(NIMBYism),又没有土地与监管羁绊。”紧接着话题又拐到了深海数据中心——既然在海面上漂着,为什么不直接沉入海底打造亚特兰蒂斯式机房?虽然能利用海水冷却且免受地表监管,但海底能源供给却是个致命伤:“难道你要带潜水钟去海底烧天然气和石油?那你还得自带氧气瓶!” 此外,这起潜在合作背后盘根错节的硅谷资本网络也令主播们忍俊不禁。Google 早期就是 SpaceX 的投资方,目前仍持有约 6.1% 的股份;与此同时,OpenAI 创业基金(OpenAI Startup Fund)又是明星 AI 编程工具 Cursor 的早期投资者,随着交叉持股与多方并购,OpenAI 间接持有了一定程度的 SpaceX 风险敞口。在马斯克与 OpenAI 打得不可开交的官司大背景下,这种“OpenAI 间接投资 SpaceX IPO”的魔幻现实被戏称为“硅谷每个人都拥有其他所有人公司的股份”。 - **前沿太空算力时间表**:Google 计划在 2027 年发射首批微型轨道机架测试原型,但商业化规模部署被皮查伊定在 10 年之后的 2030 年代中期。 - **资本联动网络**:Google 牢牢坐拥 SpaceX 6.1% 的股权,无论两家在太空数据中心上最终是盟友还是竞品,资本上的绑定都已坚不可摧。 --- ### [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) 霍尔木兹海峡危机与通胀抬头:健怡可乐短缺与华尔街焦虑 节目的尾声切入了一记沉重的宏观经济重锤:《华尔街日报》最新头条显示,上月美国通胀再度加速,CPI 同比跳升至 3.8%。金融分析名嘴乔·韦森塔尔(Joe Weisenthal)在社交媒体上连呼不妙——美国 4 月生产者价格指数(PPI)同比飙升 6.0%,远高于市场预期的 4.8%。这标志着霍尔木兹海峡(Strait of Hormuz)因美伊冲突而关闭的恶果正在全面外溢并传导至终端供应链。 主播们用日常消费品打了一个生动却严峻的比方:大家喝的健怡可乐(Diet Coke)虽然咖啡因和焦糖香精不产自中东,但罐装可乐必须的铝材原料却严重依赖中东的大型电解铝厂。霍尔木兹海峡一旦断航,全球铝供应链瞬间断裂,健怡可乐正在各大超市以肉眼可见的速度缺货断供。“你必须囤积居奇了,你家里堆了四五个托盘的货吗?”这种调侃的背后正是商品价格恶性上涨的缩影。 华尔街目前的焦虑情绪已达到数年来的峰值。此前,美股市场在很大程度上对美伊冲突及伴随而来的能源冲击表现得漫不经心、视若无睹,但飙升的能源价格和连续爆表的通胀数据终于打破了市场幻想。投资者开始大举买入抗通胀金融工具,通胀预期的急剧回升让整个金融体系重新笼罩在加息与滞胀的阴霾之中。 - **核心经济指标异动**:美国 4 月 PPI 年率录得 6.0%(预期 4.8%),CPI 上涨至 3.8%,通胀预期创下多年新高。 - **供应链外溢效应**:霍尔木兹海峡封锁直接引发中东电解铝断供,金属包装短缺已传导至包括可口可乐在内的快消巨头。 ### [28:18](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1698s) 盈亏平衡通胀率攀升与滞胀阴影下的实体经济 市场目前正同时抛售普通美国国债以及用于对冲通胀风险的抗通胀债券(TIPS,Treasury Inflation-Protected Securities)。根据美联储的数据,这两种债券收益率之间的利差——即所谓的“盈亏平衡通胀率”(break-even rate)——近期已经飙升至 2022 年 10 月以来的最高水平。这一指标的变化表明,投资者普遍预期未来 5 年的年化通胀率将维持在平均 2.7% 左右。 尽管这并非意味着世界末日,但潜藏的巨大风险在于当前繁荣的极度不均衡:GDP 增长数据几乎完全被 AI 资本支出(CAPEX)和大型数据中心的建设所拉动,而红利并未均摊至整体商业环境。除去由技术投资吹起的虚火,实体经济实际仅以 0.1% 的极低速度在微幅增长,却同时承受着未来 5 年高达 2.7% 的预期通胀压力。这种“低增长加上高通胀”的局面正是典型的滞胀(stagflation)配方,对任何经济体而言都将是一段极其痛苦的煎熬,必须引起金融市场、实体经济界以及政界决策者的高度警惕。 - **盈亏平衡通胀率(Break-even Rate)**:普通美债与 TIPS 收益率之差创下自 2022 年 10 月以来的新高。 - **通胀预期**:美联储数据显示市场押注未来 5 年平均年通胀率将在 2.7% 左右徘徊。 - **增长失衡与滞胀风险**:全美 GDP 增长过度集中在 AI 资本支出及数据中心基础设施,剔除该部分后实体经济增速仅为 0.1%,形成了“滞胀(Stagnation + Inflation)”隐患。 ### [29:20](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1760s) 鲍威尔历史评价、接班人挑战与节目收尾 回顾美联储主席杰罗姆·鲍威尔(Jerome Powell)的任期,虽然历史最终将如何定论尚无定数,但他客观上曾一度成功“把飞机平稳降落”(landed the plane),实现了难能可贵的软着陆。然而局势瞬息万变,中东战火骤起,外部宏观环境再度恶化。面对接踵而至的滞胀压力与复杂的地缘政治风波,无论谁来接棒美联储最高帅印——比如被外界屡屡提及的凯文·沃什(Kevin Warsh)——其所要面临的挑战和任务难度恐怕都丝毫不逊于前任。 最后,两位主持人向听众致意,呼吁大家在 Apple Podcasts 和 Spotify 上为节目留下五星好评,并前往官网订阅简报。节目预告将在太平洋时间次日上午 11 点整准时开播,并在欢快的互道再见与对粉丝的告白声中结束了本期录制。 - **货币政策交接**:鲍威尔曾成功完成抗通胀的软着陆,但地缘政治冲突(如中东局势)与再通胀风险正使后续操盘变得极其棘手。 - **下任掌门预期**:以凯文·沃什(Kevin Warsh)为代表的潜在继任者将接手同样严峻甚至更艰难的经济考题。 - **互动与播出信息**:节目呼吁观众在 Apple Podcasts 和 Spotify 给予五星好评并订阅官网通讯(tbvn.com / tbpn.com),下期节目将于太平洋时间工作日上午 11:00 准时直播。 ## Notable quotes > [00:02](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=2s) 大疆(DJI)至今也没能造出一架波音747 — 主持人 > [10:34](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=634s) 听着伙计们,我要求的只是一点点稳定。我就想要稳定与冷静,不想要任何冲突。 — 主持人(模仿特朗普口吻) > [08:58](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=538s) 寻求稳定 > [10:34](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=634s) 工业级规模盗窃 > [14:20](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=860s) 正式反对台独 > [14:20](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=860s) 经济与地缘政治的灭顶灾难 > [18:53](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1133s) 你根本无法在逻辑上同时为这两种截然相反的做法辩护 — 主持人 > [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) 毫无疑问,大约十年之后,我们将把轨道部署视为建设数据中心的一种更常态化方式。 — Sundar Pichai > [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) 我们应该在平流层飞艇上塞满 GPU,然后开启流水线大规模量产!在太平洋上空飘浮着一圈飞艇服务器,既免去了地面的‘邻避效应’(NIMBYism),又没有土地与监管羁绊。 — 主持人 > [23:21](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1401s) 硅谷每个人都拥有其他所有人公司的股份 — 主持人 > [26:44](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1604s) 你必须囤积居奇了,你家里堆了四五个托盘的货吗? — 主持人 > [29:20](https://www.youtube.com/watch?v=3_-mzZ21RD4&t=1760s) 把飞机平稳降落 — 主持人 ## Mentioned TBPN, John Coogan, Jordi Hays, 唐纳德·特朗普 (Donald Trump), 埃隆·马斯克 (Elon Musk), 蒂姆·库克 (Tim Cook), 黄仁勋 (Jensen Huang), 凯利·奥特伯格 (Kelly Ortberg), 大卫·所罗门 (David Solomon), 苏世民 (Stephen Schwarzman), 拉里·芬克 (Larry Fink), 简·弗雷泽 (Jane Fraser), 迪娜·鲍威尔·麦考密克 (Dina Powell McCormick), 史蒂夫·乔布斯 (Steve Jobs), 帕尔默·拉奇 (Palmer Luckey), 杰夫·贝索斯 (Jeff Bezos), 桑达尔·皮查伊 (Sundar Pichai), 戴米斯·哈萨比斯 (Demis Hassabis), 达里奥·阿莫代伊 (Dario Amodei), 萨姆·奥尔特曼 (Sam Altman), 黎智英 (Jimmy Lai), 高市早苗, 斯科特·贝森特 (Scott Bessent), 埃米尔·迈克尔 (Emil Michael), 谢尔盖·布林 (Sergey Brin), 乔·韦森塔尔 (Joe Weisenthal), 杰罗姆·鲍威尔 (Jerome Powell), 凯文·沃什 (Kevin Warsh), 特斯拉 (Tesla), 苹果 (Apple), 英伟达 (NVIDIA), 波音 (Boeing), 波音747, 波音737, 高盛 (Goldman Sachs), 黑石集团 (Blackstone), 贝莱德 (BlackRock), 花旗集团 (Citi), Meta, Anduril, 大疆 (DJI), SpaceX, Google, Google DeepMind, Anthropic, OpenAI, 深度求索 (DeepSeek), 幻方量化 (High-Flyer), 三菱日联银行 (MUFG Bank), 三井注友银行 (SMBC), 瑞穗银行 (Mizuho Bank), 超微电脑 (Supermicro), Manus, 精神航空 (Spirit Airlines), LTA Research, 蓝色起源 (Blue Origin), AWS, Cursor, Claude Mythos, Blackwell, Project Labs, 空军一号 (Air Force One), 华尔街日报 (The Wall Street Journal), Truth Social, CNBC, TIPS (抗通胀债券) ## Video description Diet TBPN delivers the best of today’s TBPN episode in 30 minutes. TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays 11–2 PT on X and YouTube, with each episode posted to podcast platforms right after. Described by The New York Times as “Silicon Valley’s newest obsession,” the show has recently featured Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella. Follow TBPN: https://TBPN.com https://x.com/tbpn https://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231 https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235 https://www.youtube.com/@TBPNLive [Source](https://www.youtube.com/watch?v=3_-mzZ21RD4)

Source quotes

[0:02] 主持人: 大疆(DJI)至今也没能造出一架波音747 [10:34] 主持人(模仿特朗普口吻): 听着伙计们,我要求的只是一点点稳定。我就想要稳定与冷静,不想要任何冲突。 [8:58] 寻求稳定 [10:34] 工业级规模盗窃 [14:20] 正式反对台独 [14:20] 经济与地缘政治的灭顶灾难 [18:53] 主持人: 你根本无法在逻辑上同时为这两种截然相反的做法辩护 [23:21] Sundar Pichai: 毫无疑问,大约十年之后,我们将把轨道部署视为建设数据中心的一种更常态化方式。 [23:21] 主持人: 我们应该在平流层飞艇上塞满 GPU,然后开启流水线大规模量产!在太平洋上空飘浮着一圈飞艇服务器,既免去了地面的‘邻避效应’(NIMBYism),又没有土地与监管羁绊。 [23:21] 主持人: 硅谷每个人都拥有其他所有人公司的股份 [26:44] 主持人: 你必须囤积居奇了,你家里堆了四五个托盘的货吗? [29:20] 主持人: 把飞机平稳降落

Source quotes

[0:02] A flex given Boeing's enduring dominance in manufacturing large commercial aircraft, whereas drone giants like DJI have been unable to scale up to build a 747. [1:10] Silicon Valley's typical stakes with the sheer scale of global geopolitics. In tech, a $2 billion acquisition or a $40 billion revenue line is considered massive, but on the world stage, leaders are bargaining over a trillion-dollar oil market, global rare earth supplies, and critical semiconductor supply chains. [4:45] If an executive never changes their look, a viral clip of them saying something foolish 10 years ago will look like it was filmed yesterday. [4:45] The Jeff Bezos playbook—getting radically jacked or going bald so the public immediately recognizes an older clip. [7:42] Dark colors like black absorb heat and burn significantly more fuel, which is why most commercial airlines stick to white exteriors. [9:35] Industrial-scale theft of American AI models [12:30] China Wakes Up 2027 [12:30] Venus flytrap [15:44] Chief dove [17:47] Claude Mythos as early as late May. Mythos is a high-powered AI model capable of discovering and exploiting software security flaws significantly faster than prior tools. [17:47] Mythos moment [23:21] Project Labs [26:44] Real economy is growing at a sluggish 0.1% while burdened by nearly 3% inflation.

Original transcript

[0:02] The big news this week and I'm sure what [0:04] you know everyone will be focused on is [0:06] of course on the cover of the Wall [0:08] Street Journal. That's the Trump Xi [0:11] Jinping summit that's happening in [0:13] Beijing. [0:15] Trump and Xi Jinping vie for wins at [0:17] summit. They're duking it out and Trump [0:20] brought They both brought a big crew [0:22] with them. [0:23] >> A team. [0:24] >> Yeah, Trump brought [0:25] >> Geopolitics. [0:26] >> A couple couple [0:27] >> Team sport. [0:28] >> couple political people that we won't [0:30] even care about but the business leaders [0:32] are what is particularly interesting. So [0:34] we got Elon Musk from Tesla, Tim Cook [0:37] from Apple, Jensen from Nvidia, Kelly [0:39] Ortberg from Boeing. Little flex there [0:42] cuz Boeing has been able to maintain the [0:44] lead in manufacturing large planes for a [0:46] very long time. [0:48] DJI has not been able to scale up and [0:50] make a 747. What's going on over there? [0:52] David Solomon from Goldman Sachs, [0:53] Stephen Schwarzman from Blackstone, [0:55] Larry Fink from BlackRock, Jane Fraser [0:58] from City and Dina Powell McCormick from [1:00] Meta. [1:02] So [1:03] we should read through [1:05] the stakes of the Trump Xi Jinping [1:07] summit in the Wall Street Journal from [1:10] the Wall Street Journal [1:10] >> into that, let's let's get into the more [1:12] important stuff cuz [1:14] do you think Jensen was always coming on [1:16] the trip, right? There's been some [1:17] speculation he got on [1:19] he got on the plane in Alaska. [1:21] >> Oh yeah. [1:22] >> Was that just the most convenient place [1:24] for him to get on or was he a late [1:25] addition? [1:26] >> I don't know. Why would he have been [1:27] been a late addition? I don't I don't [1:29] know. It seems like [1:30] >> Cuz there was there was some reporting [1:31] earlier this week [1:32] >> the largest company in the in America. [1:34] >> but but there have been statements made [1:37] from what I saw that he wasn't going [1:38] after. [1:39] >> Yes, well so Trump had a Truth Social [1:41] post that said CNBC incorrectly reported [1:43] that the great Jensen Huang of Nvidia [1:45] was not invited but then he he corrected [1:47] it. He said he didn't invite him [1:48] personally. [1:49] >> I don't know. These schedules move [1:50] around. It's got to be really really [1:51] difficult if you're the leader of the [1:52] biggest company in the world to move [1:55] things around and and commit to this. I [1:57] don't know. [1:58] >> what's more important for Jensen and his [2:01] objective? [2:02] >> No, of course it's critical importance, [2:03] but uh at the same time, there might [2:06] have just been other moving pieces of [2:07] like, okay, when can he go? Where will [2:09] he arrive from? Where will he get picked [2:10] up from? I don't know. These things [2:12] always like shift. I I don't have a good [2:14] thesis. I mean, other than the fact that [2:15] like that is the company that has maybe [2:17] the most tension between the trade deals [2:19] with China. Like Apple obviously does a [2:21] ton of business in China, but uh has [2:23] been much less of a focus in the chip [2:26] sanctions. And so, having Jensen there [2:28] is a little bit more of a discussion [2:30] point. What do you think? [2:31] >> I mean, I I think on China, like the [2:32] Nvidia stuff seems like pretty fine now, [2:35] but what but yeah, but Meta I think has [2:37] like some conflict, right? Cuz I mean, [2:38] Meta's yeah, maybe it's like not that [2:39] big of an acquisition on like market [2:41] cap-wise, but [2:42] >> Yeah. [2:43] >> They're blocking it. [2:44] >> of these things, you know, we always say [2:45] see in the in the context of Silicon [2:47] Valley, it's like a $2 billion deal or a [2:49] company that's doing 40 billion in [2:51] revenue. And then you get over there and [2:53] the stakes are like, you know, a [2:54] trillion-dollar oil market or like all [2:56] the rare earths in the world. And [2:58] and it's you know, geopolitical [3:00] conflict, which is obviously a much [3:02] bigger deal. And so, there might be more [3:04] reporting that comes out about Jensen's [3:07] goals there, what actually happened, how [3:10] he found his way onto uh Air Force One, [3:13] but that was confirmed by Elon Musk, [3:15] right? He said that Elon, Jensen, and [3:17] Trump were all on Air Force One together [3:19] flying over. Um I certainly want to [3:22] wouldn't want to get there. I would want [3:23] to arrive on the cool plane. Can we play [3:24] the video? Trump arrives in Beijing with [3:26] Elon Musk and Jensen in tow. I guess the [3:29] rest of the business leaders came [3:31] separately. Their arrival with uh are [3:34] those cheerleaders or something? I don't [3:36] know what [3:37] They They They seem to be waving Oh, [3:39] flags. They're waving flags. That makes [3:40] sense. Well, there they go, walking down [3:42] the red carpet. Air Force One is truly [3:45] stately. There's [3:47] It's so It's so disappointing that the [3:50] economics of the airline industry didn't [3:52] just get bigger and bigger and so you [3:54] find yourself very rarely on 747s unless [3:57] you're flying across the world because [3:59] they do have such a presence and there's [4:01] spaces inside. [4:03] But we have we have been [4:05] you know [4:06] crammed down into the 737s for pretty [4:09] much everything. [4:09] >> Where they give a proper tour of Air [4:11] Force One or is it kind of locked down? [4:13] >> I think they've done tours. I mean there [4:16] are several Air Force Ones I believe [4:17] that are on display at presidential [4:19] libraries and so you can go and tour [4:21] them. There might be one at the [4:22] Smithsonian. I'm not exactly sure but [4:25] I'm pretty sure you can go walk around [4:26] old ones. I wouldn't be surprised if [4:28] they keep the the current spec like [4:31] somewhat secretive. [4:33] There we go. [4:34] This is very grainy footage. It looks [4:36] like oh maybe it's just the middle of [4:38] the night. [4:38] >> me of Dune. [4:40] >> Dune. Arrakis. That's Taiwan though. [4:42] Anyway, let's go through Jensen's pack [4:45] list. He was having trouble thinking of [4:47] what was he going to bring. Of course he [4:50] landed on just a ton of leather jackets. [4:53] The back [4:53] >> And even the sneakers I feel like are [4:55] pretty iconic. [4:56] >> Yeah. [4:57] Like the uniform is really starting to [4:59] compound where it used to just be the [5:01] jacket. Soon we will know the exact [5:04] brand and specs of the the pants as well [5:06] as the shoes and the whole Jensen [5:09] uniform will be out in full force. [5:11] >> Yeah, the clothing industry doesn't want [5:13] you to know this but [5:15] wearing just just getting a few pieces [5:17] of clothes that you like and then [5:18] wearing them until they fall apart is [5:21] incredibly underrated. Not saying that's [5:22] what Jensen's doing but it's pretty [5:24] funny. [5:25] I I tend to wear almost the exact same [5:27] thing every day and so sometimes people [5:30] on the team will come in and be like oh [5:31] we're matching today. [5:33] >> Mhm. [5:33] >> Like every day you dress like this you [5:35] will match with me. [5:36] >> Yeah, that's right. That's right. [5:38] Um [5:39] there's obviously a personal brand [5:40] benefit to adopting a uniform. We've [5:44] seen it with Jensen, Steve Jobs with the [5:45] black turtleneck, Palmer Lucky with the [5:49] with the Hawaiian shirt. Big news from [5:51] Anduril today. New round series. They're [5:54] running out of letters. Series H? [5:55] >> H. [5:55] >> Series H. 51, 61 billion, something like [5:58] that. Uh fantastic progress. You adopt [6:01] the uniform, you become more iconic, you [6:02] become more recognizable outside of your [6:04] face, and that builds your personal [6:06] brand, your lore, and it helps you sort [6:08] of differentiate and stand out, right? [6:09] The flip side, the thing you got to [6:11] watch out for is that if you are always [6:13] dressed exactly the same and you never [6:15] switch your wardrobe, clips of you [6:18] saying something a decade ago can go [6:21] viral. And if you're not if you're aging [6:23] gracefully, it will look like you said [6:25] them just yesterday. And that can be a [6:27] big problem if, say, you're an AI CEO [6:30] who a decade ago was reading some sci-fi [6:32] and thinking like, "Man, there's [6:33] something to this Terminator thing." And [6:35] then maybe in the past decade you've [6:36] done a lot of work, you've investigated [6:38] the systems, you have a more informed [6:40] position and opinion. Well, the old [6:43] video of you still looks like you today. [6:45] And so, there might be something to [6:48] leveling up the wardrobe and changing [6:50] the wardrobe over time that actually [6:51] creates more distance between, "Oh, [6:54] well, clearly that was [6:56] >> So, this is indirect way of saying you [6:57] want Jensen to be in a leather sort of [7:00] like, you know, sort of trench coat [7:01] eventually. Like flowing robes. [7:03] >> Well, the easiest thing to do is go [7:05] bald. That's what Jeff Bezos did. If he [7:07] was saying some crazy stuff back when he [7:09] was building Amazon in the '90s and he's [7:11] changed his opinion, it will be very [7:13] evident from new clips that, "Oh, that's [7:15] an old one because it's very it's very [7:17] iconic." His older look. Now, his new [7:19] look, he looks transformed. He looks [7:21] very different. [7:22] >> Well, I I don't think Jensen has to go [7:23] bald, but he can do the the Bezos [7:25] playbook, which is just get like get [7:26] really jacked. [7:27] >> Get really jacked, yeah. [7:27] >> he gets more jacked, you'll just see, [7:29] "Oh, yeah, there's I mean, he's looking [7:30] like 320 lean right there." [7:32] >> There you go. [7:32] >> Like that must be pretty decent. [7:34] >> 320. Jensen at 320 would be electric. [7:39] Well, he's [7:39] >> we have a video of the Air Force One [7:42] Golden Palace. [7:43] >> Pull it up. [7:44] >> is the Golden Palace? [7:45] >> That's the [7:46] >> what they call the interior these days? [7:48] >> the Air Force One that was gifted by [7:50] Qatar. [7:51] >> Okay. And that one is never going to be [7:53] in use, [7:54] to the library, I think. Is that the [7:56] idea, or is it going to get flown [7:57] around? Well, let's let's watch the [7:59] >> Presumably, it's going to be used. [8:01] >> Golden Palace. Really [8:03] particular aesthetics here. [8:05] It looks like a normal 747 on the [8:06] outside. [8:08] >> I feel like isn't there a big [8:10] There's a big swing in [8:12] in [8:13] fuel efficiency based on the color of [8:14] the plane. If you paint your plane black [8:17] cuz you think it looks cool, it will [8:18] attract more heat, and you will burn a [8:20] lot more fuel. And so, most companies go [8:23] with white. [8:25] Spirit Airlines went yellow. I don't [8:27] know if that had anything to do with [8:28] their downfall, but [8:30] you would think is there a chance is [8:32] there some weird economic logic where if [8:34] you paint it gold, the gold will reflect [8:37] the sunlight and increase fuel [8:40] efficiency. [8:41] >> exterior is not gold. [8:42] >> So, you should wrap your plane in like a [8:44] mirror like [8:45] >> Yeah, maybe. [8:46] >> But yeah, I don't know if that actually [8:48] >> That'd probably reflect the most, right? [8:50] >> Mirrored plane. There are some silver [8:53] There are some silver bombers and and [8:55] jets that are in service across the [8:58] United States Air Force. I can't call [9:00] upon it right now. [9:01] >> It really is a sky yacht. [9:03] >> Chess board. You think anybody's played [9:05] chess on there? [9:06] >> 4D chess. [9:10] Yeah, the whole plane is actually used [9:11] for 4D chess, John. [9:14] >> 4D chess. [9:15] Uh AI is being put to good use [9:18] visualizing Xi Jinping, Donald Trump, [9:21] Elon Musk and Tim Cook having a beer and [9:23] I guess a cigarette as well in in China [9:25] somewhere. And then look who pulled up. [9:27] If you scroll down, Jensen coming in [9:29] from the back looking pretty good. [9:31] Looking pretty good. Maybe not 320 lean, [9:33] but he's certainly up there. Let's go [9:35] through the Wall Street Journal's report [9:38] and the what the editorial board had to [9:39] say about the stakes of this summit. [9:41] What what what's actually at stake? So, [9:43] the United States wants stability, [9:45] uh but Chinese Communist leader has [9:47] larger ambitions, says the Wall Street [9:50] Journal Journal's editorial board. [9:52] President Trump visits visits Chinese [9:54] President Xi Jinping in Beijing on [9:56] Thursday, and the pre-meeting US spin is [9:59] a search for stability, in quotes. Uh [10:02] it's a nice idea as long as Mr. Trump [10:04] doesn't think personal rapport can [10:06] overcome Mr. Xi Jinping's anti-American [10:09] purposes, says the Wall Street Journal. [10:11] The agenda ranges from trade to [10:13] technology to Iran. Uh on trade, the [10:16] best outcome may be ratifying the status [10:18] quo, a truce on tariffs with a promise [10:21] from Beijing not to hold the world's [10:23] rare earth supply hostage again. The [10:25] markets, business leaders, no one wants [10:27] chaos. Everyone can sort of work around [10:30] a tariff plan if it's going to stay in [10:33] peace for a long time. [10:34] >> Trump to be coming in being like, "Look, [10:36] guys, all I'm asking for is a little [10:38] stability here." [10:40] >> Yes. [10:41] >> I just want [10:42] stability. I want calmness. I don't I [10:46] don't want any conflict. [10:47] >> Yes. Didn't he famously call himself a [10:49] stable genius? That he he stability has [10:51] been part of his brand for a long time. [10:53] Uh at least the one that he projects. Uh [10:56] Mr. Xi Jinping, on the other on the [10:58] other hand, [10:59] uh China's weak economy is an incentive [11:01] for him to cooperate, and he he'll hope [11:04] to placate Mr. Trump with promises of [11:06] buying more farm goods. Uh we talked [11:09] about this previously. More aircraft and [11:12] uh other things. Uh but Mr. Xi Jinping [11:15] has made that promise before, and US [11:18] farmers have never regained their lost [11:19] market share in China. Rare earths [11:21] ransom offer, uh the rare earths ransom [11:25] offer may be more US advanced chip [11:28] exports to China. Xi Jinping views [11:31] artificial intelligence as a decisive [11:33] theater in Beijing's competition with [11:35] the United States, and he is trailing, [11:37] though not by much, maybe 6 months by [11:38] most people's estimates. [11:40] Uh the administration wants to talk with [11:42] Beijing about AI guardrails, and by all [11:45] means keep the phone lines open. Uh but [11:48] don't expect much from AI arms control. [11:52] Uh and the best deterrent is US [11:53] dominance on models and computing power. [11:55] Beijing will be happy to make [11:56] pronouncements about responsible [11:57] stewardship and then pursue its own [11:59] interests with little regard for norms [12:01] or laws. Beijing is in is engaging in [12:03] quote industrial-scale theft of American [12:06] AI models, the Trump administration [12:08] warned this year, and don't forget the [12:09] Justice Department's indictment this [12:11] year of a technology executive and [12:13] associates allegedly running a [12:15] sophisticated operation to divert [12:17] high-end chips to China. That was uh [12:19] >> Supermicro [12:20] >> Supermicro. Yeah, they were doing the [12:22] the the heater. What was it? The [12:24] hairdryer? [12:25] >> Yeah, they were putting the the fake [12:27] >> shipping labels on there. [12:28] >> big shipping labels. Noticeably absent [12:30] from this trip are the key leaders of [12:33] the top AI labs. Um I mean, Elon Musk is [12:36] there, yes, and Jensen is there, [12:39] obviously deeply involved in AI, as as [12:41] are many of the other folks. Dina Powell [12:43] McCormick at Meta, but you don't have [12:45] Demis from Google, you don't have Sundar [12:47] from Google, uh DeepMind. Uh you don't [12:49] have Dario from Anthropic or Sam Altman [12:51] from OpenAI. And each one of those [12:53] leaders has their own complicated [12:55] relationships with the federal [12:57] government, and the Trump [12:58] administration. Uh so, it's not shocking [13:00] that they aren't there, but uh it does [13:02] feel like there's a little bit of a [13:04] disconnect between the conversation [13:05] that's happening in Silicon Valley and [13:07] the conversation that's happening on the [13:08] global geopolitical stage at the highest [13:10] levels. And so, uh if you were looking [13:13] for answers to some of the biggest [13:16] questions uh posed about how [13:19] superintelligence will play out in the [13:20] global stage, the AI 2027 scenario, the [13:23] China wakes up scenario, uh you're going [13:26] to have to keep waiting because uh [13:29] that's not really what this this this is [13:31] about. It's about maybe a little bit of [13:34] semiconductor supply chain details and [13:37] export restrictions, but not answering [13:40] the big questions about what a US-China [13:44] relationship looks like in a post-AGI [13:46] world, a post-ASI world, a post [13:50] you know, a fast takeoff scenario. [13:52] >> OTP says, "Why is Meta there at all?" [13:54] Good question. All Meta platform [13:56] products have been banned for quite a [13:58] long time. I don't think that will [14:00] change, but they have this Manus thing. [14:02] I think it might just be to show some [14:04] friendliness, some presence there. [14:07] Again, I'd be shocked if there's any [14:10] time between Xi and Trump that gets [14:12] dedicated to Manus cuz like $2 billion [14:14] is really a rounding error when you look [14:16] at all of these other issues. [14:19] But uh [14:20] >> But it is a big question for both [14:21] economies. Not not only can American [14:24] companies acquire Chinese companies that [14:27] have relocated to Singapore and have [14:29] sort of moved out. Like how how strict [14:31] will China be around maintaining talent [14:34] and restricting exports of whole [14:37] companies and technologies? [14:39] Their even like a single sentence on it [14:41] from this talk will clearly resound [14:44] throughout the industry. And so the [14:46] Venus flytrap that Xi Jinping is setting [14:49] for Mr. Trump is on Taiwan. Xi Jinping [14:51] wants veto power over arm sales from the [14:54] United States to Taiwan and he is [14:56] pressing for the United States to [14:59] formally oppose Taiwanese independence [15:02] as opposed to the current posture of not [15:05] supporting it. Xi Jinping will argue the [15:07] tweak is of no great consequence to [15:09] Americans and stoke and stroke Mr. [15:12] Trump's ego that he can bring peace to [15:15] one more troubled region. Yet that [15:17] change would disrupt decades of US [15:19] policy that for all its delicate [15:21] diplomatic wording has held the peace. [15:24] Taiwan is not the aggressor in the [15:26] Taiwan Strait, uh Xi Jinping's fiction [15:29] that opposing inde- independence would [15:31] indulge. Mr. Trump might may not care [15:34] about Taiwan's freedom or its example [15:37] that a prosperous Chinese democracy is [15:39] possible, uh but the president doesn't [15:42] want a crisis on his watch, which would [15:44] be an economic and geopolitical [15:45] catastrophe. That's 100% accurate. [15:48] >> You would never want an economic [15:50] geopolitical catastrophe. [15:52] >> No. [15:53] >> On his watch. [15:54] >> Uh Xi Jinping will be looking to see if [15:55] Mr. Trump suggests he won't defend [15:58] Taiwan in the clutch. Uh Trump's [16:01] diplomacy is above all personal and no [16:05] one can predict what he'll do in the [16:06] room. Japan and others in the region are [16:08] watching with anxiety, a reminder that [16:09] US support for Taipei is an interest [16:12] that informs America's alliances around [16:14] the world. One mistake would be not stop [16:17] not stopping in Tokyo to advance Beijing [16:20] in the in advance of Beijing as a signal [16:22] of solidarity with Japan. Trump has said [16:24] he'll bring up the case of political [16:26] prisoner Jimmy Lai, but Xi Jinping won't [16:28] move to release the 78-year-old [16:30] publisher who was convicted of bogus [16:32] charges in Hong Kong unless he believes [16:34] Trump's request is more than a token [16:36] gesture. Uh the US wants China's help on [16:39] Iran, and it would be an improvement if [16:41] China at least stopped actively helping [16:42] the enemy. Uh retired Navy Rear Admiral [16:46] Mark Montgomery suggested last week that [16:48] Mr. Trump should ask Xi Jinping directly [16:51] if he's helping Iran with intelligence. [16:53] And the Wall Street Journal goes on to [16:55] say, "The larger context for this summit [16:57] is that the CCP continues to be the main [16:58] financier and industrial base for the [17:00] world's bad actors from Russia to Iran [17:02] to and North Korea. The first Trump [17:04] administration understood China as a [17:05] strategic adversary, military, economic, [17:07] and ideological. The second Trump admin [17:10] is searching for detente, the and Mr. [17:13] Trump is the chief dove. Uh this has [17:15] some merit if America spends the [17:17] interlude diversifying its rare earth [17:18] supply chain and passing a $1.5 trillion [17:22] defense budget to rearm, says the Wall [17:24] Street Journal editorial board. Anyway, [17:25] >> Yeah, and a a lot of people are [17:27] speculating that there's already been [17:29] significant deals that have been made [17:32] and that this whole thing is ceremonial [17:33] because if if you bring if you make [17:35] this, you know, huge trip and you bring [17:38] all these leaders of industry and then [17:41] you don't come away with some [17:43] significant progress, like it just feels [17:45] like a even bigger ouch. [17:47] >> Uh, in somewhat related news, [17:49] Anthropic has granted Mythos access to [17:51] three major Japanese banks. Japan's [17:54] three mega banks are set to gain access [17:56] to Claude Mythos, the powerful [17:58] artificial intelligence model developed [18:00] by US startup Anthropic as soon as the [18:02] end of May. The banks MUFG Bank, [18:05] Sumitomo Mitsui Banking Corp and Mizuho [18:08] Bank were likely informed of the move by [18:10] the US Treasury Secretary Scott Bessent [18:12] in a meeting in Japan on Tuesday. This [18:15] will mark the first time that a company [18:16] from the East Asian nation has been [18:18] granted access to Mythos, the AI model [18:21] which can discover and exploit software [18:22] security laws far faster than earlier [18:25] tech had been restricted to just 50 or [18:27] so corporates and organizations [18:29] worldwide, including US firms, US banks, [18:32] and UK government organizations. So, [18:33] this isn't the first international [18:35] expansion since the UK did have access, [18:37] but now Japan does. Japanese Prime [18:39] Minister Sanae Takaichi had instructed [18:42] cabinet members to ramp up efforts to [18:43] find cybersecurity weaknesses in Japan's [18:46] infrastructure and minimize risks posed [18:49] by cutting-edge AI models like Mythos. [18:51] Access to Mythos was likely mentioned in [18:53] her meeting with Bessent on Tuesday. And [18:55] so, Chris Maguire says tonight the [18:58] Secretary of the Treasury is personally [19:01] vetting and approving each company that [19:03] gets access to the most advanced US AI [19:05] model because the risks of the model [19:06] being misused to hurt US national [19:09] security are so high. Also tonight, [19:10] Jensen Wong is flying on Air Force One [19:12] with President Trump to Beijing to sell [19:14] China the AI chips it will use to [19:16] develop its own mythos level AI model as [19:18] soon as possible. The administration The [19:20] administration's AI policy remains [19:22] inconsistent and incoherent. It is [19:24] impossible to justify these two [19:25] approaches simultaneously. And [19:28] Yeah, I I'm I'm still sympathetic to the [19:30] the the export control take. At the same [19:32] time, I feel like there's a larger [19:34] negotiation that's going on, so I can [19:36] sort of see both sides. But [19:38] the interesting question that I've been [19:39] noodling on that I haven't fully gotten [19:41] to the bottom of is [19:42] you know, Dario has said that [19:44] you know, the the open-source community [19:47] China is maybe six months behind mythos. [19:50] And that seems to have fallen in line [19:51] with the progress that we're seeing from [19:53] all the American labs. And so the [19:56] interesting question is what would the [20:00] mythos moment look like in China? [20:04] Because in America, there was a showdown [20:06] with the Department of War and Dario [20:08] going back and forth and Emil Michael. [20:10] And there was threat of a supply chain [20:11] designation, which it doesn't feel like [20:13] ever really went anywhere because the [20:16] business is still cooking and it's [20:18] available on all the hyperscaler clouds [20:20] and the model's being granted to Japan. [20:23] And so it feels like the the the [20:26] administration has backed off of that. [20:28] But it was a really tense moment and it [20:29] did call into question all these all [20:31] these debates around what rules and what [20:35] authorities should rest with the private [20:37] sector versus the government. [20:39] And America handled it in sort of a [20:42] democratic way loosely. [20:45] Sort of evaluating the technology, [20:47] having the option to sort of nationalize [20:50] and there's this bigger discussion [20:51] around that. But in China, we've seen [20:54] many times where tech companies have [20:56] gotten really big and it feels like the [20:58] the CCP is applying pressure to the [21:00] CEOs. And High Flyer is [21:03] the the company behind Deep Seek is in a [21:05] very unique position where the founder [21:08] has an immense amount of control and [21:11] you know no matter what you think of the [21:12] >> billion of his own money and [21:15] a really low dilution round. [21:16] >> I'm pretty sure he's a solo founder who [21:18] like owns [21:20] probably like majority control at least [21:21] majority voting control if not majority [21:23] economic control and that's a wildly [21:25] different than Anthropic where Dario had [21:27] like 15 co-founders, tons of dilution [21:30] from from different funding rounds and [21:34] you know we had talked about the SPV [21:35] thing yesterday like there are lots of [21:37] people on the cap table like Anthropic [21:39] is going the way of like a typical [21:41] American company where there are lots of [21:42] different stakeholders. They might be a [21:44] public company soon which means the [21:46] government has more oversight as the SEC [21:49] controls a lot of what they will do and [21:51] so you're in this odd scenario where you [21:55] you could have a similar level of [21:56] technology deployed in China from a very [22:00] closely held private company that is up [22:03] against a much more aggressive [22:07] government. If Deep Seek is able to [22:08] catch up you know there's some there's [22:10] some [22:11] some evidence that like China's on a [22:13] slightly shallower growth trajectory but [22:16] still you know if you take Dario's claim [22:19] of 6 months even if you extend that to a [22:21] year like something is going to happen [22:23] there where China is going to have their [22:25] own Mythos moment and what that looks [22:28] like and what we learn about it I I [22:30] think is going to be dramatic and [22:32] interesting and also it's just it's just [22:35] odd that like there's this backdrop of [22:36] all this stuff that's happening in AI in [22:38] America and then there's going to be you [22:41] know talks over like you guys want to [22:43] buy more apples. Taylor do you have any [22:45] any thoughts on on this and am I [22:46] directly correct on the timeline or you [22:49] know this [22:49] >> Yeah, I mean it seems like something's [22:51] going to happen here. There's a graph [22:52] that shows that open source is [22:53] increasingly getting like further and [22:55] further away from frontier models. But [22:56] yeah, I don't know it's just like so [22:57] much to tell like what's actually going [22:58] on in China like what like [23:00] >> what what happens when the CCP starts to [23:03] to have, you know, more control over a [23:04] company, like what does it actually look [23:05] like? [23:06] >> Yeah. [23:06] >> In in the instantiation. [23:08] >> Can they continue to be pro-competition [23:10] if they need to consolidate all of the [23:12] black wells that are coming into the [23:13] country into one data center that's [23:15] networked together that can actually [23:16] train a mythos level model? [23:18] >> Right. That's uh [23:20] That's That's a big question. [23:21] >> SpaceX [23:22] >> SpaceX [23:23] >> Google are in talks to launch data [23:24] centers in orbit. [23:26] >> Take it [23:26] >> The deal between the two tech titans [23:28] would give a boost to SpaceX business [23:30] ahead of a historic public listing. [23:31] Google is in talks for a rocket launch [23:34] deal as a search giant expands its own [23:36] efforts to put orbital data centers in [23:37] space. Launch deal would put the two [23:39] companies in partnership as they gear up [23:40] to compete on orbital data centers, an [23:42] unproven technology that SpaceX CEO Elon [23:45] Musk has said is the next frontier for [23:47] his rocket company. Google is also in [23:50] discussions about a potential deal with [23:51] other rocket launch companies. [23:53] >> Um they got to at least have a stocking [23:56] horse. [23:57] >> How are we not talking about blimp data [23:59] centers yet? [24:00] >> I know. [24:00] >> Sergey Brin has a blimp company. You [24:02] know, Bezos has a rocket launch company, [24:04] so there's a very logical line from AWS [24:07] to space data centers through Blue [24:09] Origin. But we got to put some GPUs on [24:13] the blimps. We got to start ma- mass [24:15] manufacturing the blimps. This could be [24:17] the thing. But I mean no no there's no [24:19] NIMBYism in the middle of the Pacific [24:22] Ocean if there's just a bunch of blimp [24:23] data centers floating around in a [24:25] circle. [24:25] >> Well, I think if they're already in the [24:26] ocean, you should go down as well, you [24:28] know, Atlantis map. [24:29] >> ling [24:30] >> Yeah, like I feel like none of the ocean [24:32] guys that are working on ocean robotics [24:34] are thinking about like putting at [24:36] the bottom of the ocean. [24:37] >> Yeah, I mean it feels easier than uh [24:40] than than uh space data center [24:41] insurance. The space is free, the land [24:44] is free, you don't run into the [24:45] regulatory stuff. Energy, little bit [24:47] tricky down there. [24:48] >> Geothermal? [24:49] >> Geothermal maybe. I don't know. [24:51] >> No, there's there's oil and natural gas [24:53] reserves on [24:54] >> You have to bring oxygen there cuz how [24:56] do you burn the oil and the natural gas [24:59] under water? [24:59] >> Big. [25:00] Diving bell. Tube, yeah, scuba scuba [25:03] tanks or something. Anyway. [25:04] >> Speculative technology has been at the [25:05] center of SpaceX's pitch to investors [25:07] ahead of its planned IPO this summer, [25:10] which is anticipated to be the largest [25:12] IPO of all time. Last year, Google [25:14] announced its own plans to launch [25:15] prototype satellites by 2027 as part of [25:18] a moonshot initiative called Project [25:22] Labs, [25:23] >> name. [25:23] >> to build those satellites, tiny racks of [25:26] machines, and have them in satellites, [25:27] test them out, and then start scaling [25:29] from there, says Sundar. There's no [25:31] doubt to me that a decade or so away [25:33] we'll be viewing it as a more normal way [25:35] to build data centers. A little bit of [25:38] an interesting timeline [25:43] >> Elon's like, I thought we were saying [25:44] >> There's no doubt to me that a decade or [25:47] so away we'll be viewing it as a more [25:50] normal way. [25:52] >> Yeah. [25:52] >> A decade or so away. [25:54] >> He's being a little cautious in his [25:55] language on his timelines, but [25:57] >> Google was an early investor in SpaceX [25:59] and owns 6.1% [26:01] of the company. [26:03] >> Yeah. [26:03] >> Not bad. Um the other the other funny [26:07] dynamic from SpaceX's deal making [26:10] recently is given that the OpenAI [26:13] startup fund was uh an early investor [26:16] in, I believe, Cursor's seed round, [26:19] >> True. [26:19] >> million round. OpenAI, at least the [26:22] startup fund, will have some meaningful [26:24] exposure to SpaceX. [26:26] >> SpaceX an OpenAI backed company. That's [26:29] the way people will introduce it. OpenAI [26:30] backed SpaceX IPO. [26:32] >> Well, yeah, I thought it was it was [26:33] quite interesting having that happen [26:35] while the whole trial was going on. [26:37] >> Everyone owns everything. Everyone has [26:39] diversified into every every company at [26:42] this point, basically. We can flip over [26:44] to the other financial news that's [26:46] making headlines on the cover of the [26:48] Wall Street Journal. Inflation [26:49] accelerated 3.8% [26:52] last month. [26:54] Not good news. Uh Joe Weisenthal has a [26:57] couple posts, but uh one is just man, [27:00] and he says uh the headline, "US April [27:03] producer prices rose 6% year-over-year. [27:06] The estimate was 4.8%, so obviously the [27:09] closure of the Strait of Hormuz is [27:11] having an effect." We were talking about [27:12] Diet Cokes. You would think that the uh [27:15] the caffeine doesn't come from the [27:16] Middle East, the the delicious caramel [27:18] flavor, whatever else they put in this [27:20] uh [27:21] it does not come from the Middle East, [27:22] but aluminum actually, there are [27:24] smelters in the Middle East. Aluminum is [27:26] in shorter supply because of the closure [27:29] of the Strait of Hormuz, and so Diet [27:30] Coke is rapidly going out of stock, and [27:33] so you need to load up if you're [27:36] prepared. [27:36] >> you got four or five pallets? [27:39] >> I have not actually gotten any. I'm not [27:41] stockpiling Diet Coke yet, but who [27:43] knows, maybe prices go up. That's what [27:45] inflation represents, and Wall Street is [27:48] getting more anxious about it, so we can [27:50] read through this. Surging energy prices [27:52] have pushed investors' inflation [27:54] expectations to multi-year highs. [27:56] Tuesday's hot readout on consumer prices [27:59] is intensifying Wall Street's anxiety [28:01] about inflation. Uh even before the [28:03] latest release of the consumer price [28:05] index, inflation expectations had been [28:07] climbing, a potential trouble spot in [28:09] the market where stocks have largely [28:12] shrugged off the US-Iran conflict and [28:14] the resulting energy shock. And so [28:16] investors bet on inflation by buying and [28:18] selling both ordinary US Treasuries and [28:20] those that hedge the risk of inflation [28:22] called TIPS, uh Treasury [28:25] Inflation-Protected Securities. The gaps [28:28] between the yields of those two types of [28:30] bonds, known as the break-even rate, [28:32] recently hit the highest level since [28:34] October of 2022, [28:36] according to Federal Reserve data, [28:37] suggesting that investors expect annual [28:40] inflation to average around 2.7% over [28:43] the next 5 years. Now, that's It's [28:45] entirely doom and gloom, but the risk [28:47] here is that the AI boom, the the GDP [28:51] numbers are sort of concentrated in AI [28:54] CAPEX and data centers. It's not evenly [28:56] distributed. The real economy is only [28:58] growing at 0.1% while the real economy [29:01] is suffering from 2.7% inflation over [29:04] the next 5 years. That's a recipe for [29:07] stagflation, stagnation plus inflation, [29:10] uh a very rough thing for any economy to [29:13] go through and something that uh the [29:16] market and the economy and the [29:18] politicians should be focused on. [29:20] >> In some ways, you know, who who knows [29:22] how how history will view Powell, but [29:24] like he landed he landed the plane. [29:27] >> Good. [29:27] >> And then, you know, we get war in the [29:30] Middle East and uh and now Kevin Warsh [29:33] in some ways looks like he'll have just [29:35] as tough of a job. [29:36] >> Potentially. Potentially. [29:37] >> Leave us five stars on Apple Podcasts [29:39] and Spotify. Sign up for the newsletter [29:40] at tbvn.com and we will see you tomorrow [29:42] at 11:00 at 11:00 a.m. Pacific sharp. [29:44] >> We love you. [29:45] >> Goodbye. [29:45] >> Goodbye.