Why The AI Bubble May Be Good

TL;DR

Even if the AI bubble bursts, the underlying infrastructure investment in computation will drive the next wave of economic growth.

Key Points

• 1.Circular investment problem: Nvidia invests in OpenAI, who buys Nvidia chips with Nvidia's money — creating artificial demand before real-world productivity gains arrive, which analysts don't expect until 2027 or later.
• 2.Google's hardware challenge to Nvidia: Google's new tensor-core chips offer 2.8x better performance-per-watt than Nvidia's best GPUs, and a $52B deal with Anthropic (supplying 1 million chips) signals a slow but serious erosion of Nvidia's dominance.
• 3.Historical precedent supports optimism: Railways, electrification, and 1990s fiber networks all saw massive capital misallocation and investor wipeouts — yet the infrastructure built real economic growth for the next era.
• 4.The real question isn't if, but who survives: Like past infrastructure bubbles, early investors will largely lose, but the few companies that endure will capture enormous long-term value — identifying them is the key challenge.