The AI chip race is now a full-scale infrastructure buildout, but the real money and power plays are far more complex than just silicon: it’s about control over energy, talent, and data, with China making aggressive moves on all fronts.
📊 11 episodes across 8 podcasts
⏱ 550 minutes of intelligence analyzed
🎙 Featuring: Andrew Feldman, Joe Weisenthal, Tracy Alloway, Gavin Baker
|
The Big Shift
The conversation around AI has lurched from software to hard limits: we are now discussing watts of power and wafers of silicon as the core constraints, not just models and algorithms. This isn't just about Nvidia anymore; it’s an infrastructure game where the biggest players are racing to secure scarce resources.
The evidence: Two guests this week, David Sambur from Apollo Global Management and Gavin Baker of Atreides Management, pointed out that the critical bottlenecks for AI development are now primarily physical. Cerebras CEO Andrew Feldman also highlighted the massive capex required — $500M over 5 years — just for his company's chip development, painting a picture of an industry where physical limitations dictate strategic advantage.
"I think capitalism is going to solve the Watts shortage absent big regulatory or political blowback, which I think is a real possibility."
— Gavin Baker, Founding Partner and CIO of Atreides Management on Invest Like the Best with Patrick O'Shaughnessy
Why it matters: This isn't a software pivot. It’s a full-stack, earth-moving play. Private equity and infrastructure investors are moving in, not just venture. The shift means that traditional operational excellence playbooks—securing supply chains, optimizing energy, managing large-scale physical assets—are becoming critical for participating in the AI boom. Expect to see deals focused on power plants, real estate for data centers, and advanced manufacturing rather than just SaaS multiples. The game is less about rapid deployment and more about control of the foundational layers. This structural constraint also creates unique opportunities for countries that can provide both the energy and manufacturing capabilities, notably China.
The Rundown
① China Outmaneuvered Trump on Trade.
Despite Trump's tough talk, China demonstrably held the upper hand in the recent Trump-Xi summit, leveraging control over rare earths and critical minerals essential for US industries (James Kynge on The Prof G Pod with Scott Galloway).
→ What it means for boards: Boards need to be asking their leadership teams about direct and indirect dependency on Chinese-controlled critical minerals, and what their mitigation strategy is for potential supply chain leverage or disruption.
② Anthropic Hits Profitability, Shakes AI Narrative.
Anthropic achieved EBITDA positive status in its most recent quarter, demonstrating strong ROI for AI infrastructure and challenging the notion of exclusively capital-intensive AI companies (Gavin Baker on All-In with Chamath, Jason, Sacks & Friedberg).
→ The Immediate Takeaway: This shifts the conversation from pure growth at all costs in AI to profitable, scalable growth, pressing other AI developers to demonstrate a clear path to profitability and ROI on their compute spend.
Signal Board
🔥 HEATING UP
• Chinese Financial Assets Investment: Chinese government bonds are increasingly seen as safe havens and diversifiers, attracting global investment despite past Western skepticism (Ozan Tarman and Aditya Singhal on Odd Lots).
• Watts and Wafers as AI Constraints: The physical limitations of power and advanced chip manufacturing are now the primary bottlenecks for AI development, driving massive capital investments (Gavin Baker on Invest Like the Best with Patrick O'Shaughnessy).
• Corporate Governance Reform in South Korea: Recent legal changes mandating fiduciary duty to shareholders across Korean corporations are opening up new long investment opportunities (Fahmi Quadir on Odd Lots).
• AI inference speed and token cost economics: The ability to run AI models faster and cheaper is becoming a competitive differentiator, directly impacting profitability and product viability (Andrew Feldman on Odd Lots).
👀 ON WATCH
• Anthony Karpathy (🆕): The addition of the former OpenAI and Tesla FSD lead to Anthropic to spearhead recursive self-improvement signals a new frontier in model development, focusing on internal feedback loops for unprecedented speed (Chamath on All-In with Chamath, Jason, Sacks & Friedberg).
• Long-term Hardware Innovation in Chipmaking: Despite massive investments, Andrew Feldman of Cerebras expects even bigger innovations ahead for wafer-scale engines, indicating continued disruption in hardware (Andrew Feldman on Odd Lots).
• US-China Geopolitical Chessboard: China's strategic leveraging of rare earths and critical minerals during the Trump-Xi summit indicates a new dynamic where China holds significant leverage over US industries (James Kynge on The Prof G Pod with Scott Galloway).
• Less trafficked parts of the market: Apollo's strategy of targeting large, complex deals that avoid crowded competitive situations suggests that private equity alpha is being found in overlooked, harder-to-execute opportunities (David Sambur on Dry Powder: The Private Equity Podcast).
❄️ COOLING OFF
• Nvidia CUDA ecosystem diminishing dominance: Nvidia's CUDA market share in leading frontier AI models has dropped significantly, with two out of three major models now built without CUDA (Andrew Feldman on Odd Lots).
• Direct-to-consumer is not the right distribution channel for niche products: For single, niche products like specialized socks, D2C is proving difficult to scale profitably, highlighting the need for aggressive B2B partnerships (Sarah LaFleur on How I Built This with Guy Raz).
• Informational alpha for short-sellers: The ability of short-sellers to impact price discovery through exposing fraud is diminishing, forcing a shift towards 'factor investing' rather than direct informational exploitation (Fahmi Quadir on Odd Lots).
The Debate
The future of US-China tech policy: Containment vs. Engagement.
🐂 The Bull Case for Containment: Andrew Feldman of Cerebras supports limiting technology diffusion to industrial rivals like China, even if it means foregoing certain markets, arguing this is essential for maintaining US technological advantage. "If you wanted to preserve value, you would have shot down the Wright brothers on their test flight," (Andrew Feldman on Odd Lots) he claims, emphasizing the need for strategic control over critical technologies.
🐻 The Bear Case for Engagement: Deutsche Bank's Ozan Tarman and Aditya Singhal highlight the significant capex investment in the Western AI stack, yet acknowledge China's parallel advancements in chip technology (Huawei's H100 comparable processors). They suggest the West underappreciates China's capabilities, advocating for a more nuanced understanding of their advancements rather than outright containment (Ozan Tarman and Aditya Singhal on Odd Lots).
Our Read: While the desire for containment is strong, the reality on the ground from trade data and China's internal tech progress suggests that full disengagement is becoming increasingly difficult, making a more complex, interwoven strategy inevitable.
The Bottom Line
The AI gold rush is transforming into a bare-knuckle fight for physical resources, strategic geopolitical positioning, and operational mastery—the smart money is already re-underwriting the playbook.
📖 Want the full episode breakdowns, guest details, and listen links?
Episode Guide
1. Odd Lots — "Why Cerebras CEO Andrew Feldman Built The World's Largest Computer Chip"
Runtime: 52 min | Host: Joe Weisenthal, Tracy Alloway | Guest: Andrew Feldman (CEO and Founder, Cerebras)
Recommended for: Anyone tracking the real bottlenecks and strategic plays in the AI hardware market.
Cerebras CEO Andrew Feldman discusses building wafer-scale AI chips, vastly larger than GPUs, to achieve unprecedented inference speeds. The episode delves into the critical role of G42 in Abu Dhabi as a major customer and investor, the long-term commitment required for hardware innovation, and the strategic implications of AI for white-collar jobs amidst US chip manufacturing challenges and export controls.
"It's 58 times larger than any other chip that had ever been."
— Andrew Feldman, CEO and Founder of Cerebras
2. Odd Lots — "Deutsche Bank's Ozan Tarman and Aditya Singhal on Understanding the Macro Risks"
Runtime: 29 min | Host: Tracy Alloway, Joe Weisenthal | Guest: Ozan Tarman (Vice Chair of Global Macro, Deutsche Bank), Aditya Singhal (Head of EM Trading Across Rates, FX and Credit, Deutsche Bank)
Recommended for: Investors and strategists needing a concise view on market resilience and hidden macro risks.
Deutsche Bank's Ozan Tarman and Aditya Singhal discuss the surprising market resilience driven by tech earnings and AI bets. They highlight the tension between AI-driven growth and geopolitical risks, contrasting Western AI capex with China's parallel advancements in chip technology. The conversation also covers the emerging narrative of robotics and investment opportunities in Chinese financial assets as safe havens.
"The Kool aid is to believe that this AI run may have another one to two years to go."
— Ozan Tarman, Vice Chair of Global Macro at Deutsche Bank
3. All-In with Chamath, Jason, Sacks & Friedberg — "SpaceX's $2T Case, Nvidia's Shock Selloff, America Turns on AI, Trump Pulls AI Order, Bond Crisis?"
Runtime: 102 min | Host: Chamath, Friedberg, Jason, David Sacks | Guest: Gavin Baker (Guest, Atreides Management), Gavin (Partner at Atreides Management, Atreides Management)
Recommended for: Anyone tracking extreme valuations, AI's self-improvement frontier, and surprising profitability shifts.
Gavin Baker from Atreides Management joins the All-In hosts to discuss Andrej Karpathy's move to Anthropic to lead recursive self-improvement in AI models. They also cover Anthropic's unexpected profitability, the "anti-humanist" psychological current fueling public AI skepticism, and major revelations from the SpaceX S1 filing, including its ambitious valuation and substantial AI deals.
"The success is extraordinary. It's undeniable. I think the fact that they are now they were ebit positive per the Wall Street Journal in the most recent quarter is a really important fact for kind of the whole AI narrative because now there's, you know, you could talk about circular funding, you could talk about roi, and we could go look at the ROIC of the hyperscalers. But if OpenAI and Anthropic are at, call it $100 billion of ARR now with 80% ish gross margins on inference, like the returns are there."
— Gavin Baker, Guest at Atreides Management
4. The Prof G Pod with Scott Galloway — "China Decode: Trump Promised to Be Tough on China. Xi Outplayed Him."
Runtime: 40 min | Host: Alice Han, James Kynge, Vox Media Podcast Network | Guest: Host-led discussion
Recommended for: Geopolitical strategists and business leaders with significant exposure to US-China relations.
This segment analyzes the Trump-Xi summit, concluding that China, for the first time, held the upper hand due to its control over rare earths and critical minerals. It explores discrepancies in post-summit readouts and the role of US CEOs like Jensen Huang of Nvidia and Elon Musk of Tesla in navigating access to the Chinese market and critical supply chains amidst growing dependencies.
"For the first time ever, the Chinese leader had the upper hand in this summit."
— James Kynge, Host at China Decode (Vox Media Podcast Network)
5. Invest Like the Best with Patrick O'Shaughnessy — "Gavin Baker - Watts and Wafers - [Invest Like the Best, EP.473]"
Runtime: 77 min | Host: Patrick O'Shaughnessy | Guest: Gavin Baker (Founding Partner and CIO, Atreides Management)
Recommended for: Deeply technical investors and strategists focused on the future of AI infrastructure and market dynamics.
Gavin Baker discusses "watts and wafers" as the primary constraints on AI development, predicting that capitalism will address the power shortage by 2027-2028, with orbital compute as a long-term solution. He highlights TSMC's critical role in preventing an AI market bubble and Elon Musk's "Terrafab" initiative, emphasizing the need for investors to master new AI technologies to remain competitive.
"I think capitalism is going to solve the Watts shortage absent big regulatory or political blowback, which I think is a real possibility."
— Gavin Baker, Founding Partner and CIO of Atreides Management
