Inside the Mind of a Tech Investor with Gavin Baker

The AI-driven semiconductor landscape faces a unique capacity cycle defined by physical constraints rather than traditional market bubbles. Taiwan Semiconductor’s measured expansion, driven by the need to preserve legacy and avoid market volatility, acts as a primary bottleneck for global compute supply. While historical memory cycles typically signal a sell-off, current demand for high-bandwidth memory suggests a sustained growth phase. Coding has emerged as the dominant AI application, shifting enterprise spending toward usage-based models that prioritize token consumption. Meanwhile, specialized "neocloud" providers offer a durable competitive advantage through superior cluster management, outperforming hyperscalers in efficiency. Looking ahead, orbital compute represents a critical infrastructure evolution, utilizing solar power and space-based cooling to circumvent the terrestrial power and thermal limitations currently hindering data center growth.