Watts, Wafers, and the Future of AI Infra | Gavin Baker

Sarah Guo is the founder of Conviction and one of the most closely watched investors in AI. In this conversation, Sarah and Patrick discuss what the people at the frontier of AI actually believe, why some researchers think exponential intelligence may be only a few years away, and how she’s navigating an investing environment that can’t be backtested.

Neil Movva, co-founder of Sail Research, joins Patrick to explain why the next era of AI may be defined not by faster chatbots, but by background agents that work autonomously for hours, days, or even weeks—and what it will take to make that intelligence cheap enough to become abundant. Neil lays out Sail’s “token factory” strategy across software, chips, data centers, and power; the lessons he learned chasing “speed of light” performance at Nvidia; why he’ll buy almost any chip at the right price; and why unreliable, distributed data centers powered by sources like solar and wind could actually be an advantage. They also discuss transformers and memory, the future of AI training data, the economics of the chip shortage, open versus closed models, and Neil’s belief that there will always be demand for more intelligence.

Ben Thompson joins Invest Like the Best for a wide-ranging conversation about the economics, geopolitics, and business models shaping the AI era. They discuss why overwhelming U.S. dominance in AI could be dangerous, whether the massive AI infrastructure buildout can generate returns before the capital runs out, and what the railroad boom can teach us about today’s spending cycle. Ben also explains why Google may increasingly resemble Berkshire Hathaway, how AI could turn intelligence into a commodity, why Amazon may have the deepest moat in technology, what Apple gets right by staying focused on hardware, how AI threatens Microsoft’s core business, why the current compute shortage may have saved Intel, and why Google and Amazon could ultimately become Nvidia’s most important competitors.

Eric Vishria, General Partner at Benchmark, joins Invest Like the Best for a wide-ranging conversation on how AI is reshaping software, infrastructure, and venture investing. He explains why the competitive frontier for SaaS has shifted, why companies can destroy value by faithfully executing an outdated plan, and what the cloud wars can teach us about the emerging AI market. Eric also discusses Fireworks and Sierra, the importance of understanding AI’s “jagged edge,” why some AI companies are effectively “selling magic,” and why he believes the market could support many enormous winners. The conversation also covers energy as a potential bottleneck to intelligence, the story behind Benchmark’s early investment in Cerebras, the future of robotics, what he looks for in founders and board partnerships, Benchmark’s expansion into growth investing, and the lessons he’s learned from backing companies capable of generating extraordinary returns.

In this episode of Invest Like the Best, Patrick O'Shaughnessy sits down with Gavin Baker to dissect the intense AI and semiconductor market sell-off of July 2026. Describing the period as "2022 packed into a single month," Gavin explains why public market panic contrasts sharply with the booming fundamentals seen on the ground in Silicon Valley. They dive into the accelerating operating cash flows of hyperscalers, the soaring spot prices for GPUs, and why open-source AI models are actually driving massive infrastructure demand. Gavin also unpacks the game theory behind Long-Term Agreements (LTAs) in the memory market, NVIDIA's strategic dominance, SpaceX's massive data center ambitions, and the looming regulatory risks for AI compute.