Inside the Mind of a Tech Investor with Gavin Baker

The semiconductor and AI compute landscape faces unprecedented physical constraints, driven by a fundamental shortage of wafers and power. While historical memory cycles typically signal a time to sell, the current cycle represents a unique capacity-driven shift where supply remains tightly controlled by TSMC’s conservative expansion strategy. Investment success in this sector requires moving beyond traditional "all-you-can-eat" models toward usage-based pricing, as AI token consumption becomes a critical enterprise expense. Furthermore, the operational complexity of managing high-performance GPU clusters creates a durable competitive advantage for specialized "neocloud" providers like CoreWeave. Looking ahead, orbital compute offers a potential solution to terrestrial power and cooling limitations, likely gaining meaningful market share by the end of the decade as the industry transitions away from the current reliance on traditional terrestrial data center build-outs.