Ben Thompson discusses the geopolitical and economic implications of the AI race, warning that US dominance could be dangerous and that AI's capital intensity risks a "running out of money" scenario similar to historical railroad bubbles. He argues the current equilibrium—where US frontier labs lead but China stays 6–9 months behind—is favorable, but warns that the industry is burning through free cash flow, debt, and now equity (e.g., Google's issuance), and that compute shortages may ironically force Big Tech to work with Intel/Samsung, breaking TSMC's monopoly.
Summarized by Podsumo
US winning the AI race outright could trigger a dangerous game-theoretic response from China (e.g., blowing up TSMC), and US dependence on Chinese manufacturing is underappreciated.
AI's capital cycle is accelerating: companies have exhausted free cash flow and debt, and are now issuing equity (e.g., Google) and using circular financing (e.g., NVIDIA's backstop deals), risking a bust if revenue doesn't catch up.
OpenAI's pivot to advertising was overdue; Thompson calls it a repeat of Dropbox's consumer-to-enterprise struggle but at 100x scale.
TSMC's conservatism in expanding capacity has created acute compute shortages, which will ultimately force Big Tech to help rivals Intel/Samsung catch up—solving both the supply and geopolitical dependency problems.
Meta has the most interesting setup among frontier AI players because its ad platform is a perfect verification loop for AI-generated content, and it benefits from the human desire for connection in an AI-saturated world.
"We are dependent on a company that is 60 miles offshore of our greatest geopolitical opponent who thinks it's theirs."
— Ben Thompson on TSMC
"If you think about it, the most interesting thing about the internet is free distribution. Bandwidth is a commodity… Commodities change the world."
— Ben Thompson
"The risk doesn't disappear, it just moves. Sometimes that risk manifests in not making money, not in losing money."
— Ben Thompson on Big Tech's foregone revenue due to compute shortages