Is Most AI Spending Circular?

Chamath Palihapitiya says most AI spending is circular. One percent of companies represent 90% of AI spending on tokens, and 80% of that 90% comes from unprofitable VC-backed companies.

VCs raise from pension funds, LPs, foundations, and family offices, then put the money into a startup. The startup spends part of it consuming other people's AI, shows revenue traction, and raises again from the same VCs. Chamath Palihapitiya says 75 cents of every dollar is a startup sending money back and forth so the companies can mark themselves up. He calls it a structural problem. Ken Griffin's point, in his telling, is that AI has to diffuse out of the startup ecosystem before it can scale as a predictable driver of long-term earnings. Breathtaking valuations sit on spending that has not spread into the wider economy.

The same pattern showed up in the early days of cloud. When AWS first offered EC2 and S3, the buyers of an EC2 instance or an S3 bucket were startups.