Brad Gerstner: No AI Bubble, Semis Eat the Nasdaq & AI's Take Off Problem

Gerstner says the AI trade now depends on revenue conversion: Anthropic's monthly revenue rose from $2 billion in January to $11 billion in March, but Anthropic, OpenAI, and SpaceX together would need to increase roughly $100 billion of run-rate revenue by July to at least $180 billion by year-end. He argues that Microsoft, Google, and Amazon's infrastructure spending requires corresponding customer revenue.

Gerstner argues that AI's addressable market is large enough to support the investment, with capturing about 4% of knowledge-work demand—equivalent to $1.2 trillion—potentially covering the required capital expenditures. He also points to Uber's projected 20% growth and Snowflake's projected 30% growth without additional headcount as examples of how AI could expand margins by slowing hiring rather than requiring mass layoffs.

The projected AI buildout faces practical bottlenecks: planned compute additions could rise from 19 gigawatts in 2026 to 43 gigawatts the following year, but Gerstner expects actual deployment closer to 25 gigawatts because of permitting and grid delays, labor shortages, and sold-out power equipment. He also estimates that $100 billion to $110 billion of revenue could be supported by about 1.5 gigawatts, suggesting revenue growth may not require the full projected buildout.