What Happens When the AI Boom Runs Out of Money
Summary
Ben Thompson and a guest analyze the geopolitical risks of US AI dominance, arguing that China’s optimal response to American military superiority would be destroying TSMC. They discuss the current AI equilibrium, the "railroad era" capital cycle, and the timing mismatch between massive infrastructure spending and revenue generation. The conversation also covers the shift from subscription to advertising models in consumer AI, the commodity dynamics of memory chips, and the critical importance of TSMC’s conservative capacity expansion strategy.
Topics discussed
US dominance in AI and national security risks
China's AI capabilities and the US competitive edge
AI recursion, capital curves, and funding models
Railroad history as a metaphor for AI investment
Verifiable vs. unverifiable AI domains and limits
AI applications in medicine and economic opportunity
Google's network effects and transaction costs
Apple's ecosystem and the Dropbox acquisition story
AI advertising models and consumer pricing challenges
Compute shortages, supply chains, and market dynamics
Memory chip cycles and TSMC's strategic positioning
Intel's foundry struggles and TSMC's risk management
Hyperscaler strategies: AWS, Graviton, and custom silicon
Apple's supply chain prowess vs. AI business models
SpaceX AI, IBM's legacy, and Microsoft's enterprise lock-in
AI's impact on social networks and content creation
Meta's advertising strategy and societal benefits
Meta's VR spending and antitrust concerns
Compute as a commodity and NVIDIA's circular financing
Hyperscaler threats, energy constraints, and future outlook
Listen ad-free on Castria