Investor Paul Kedrosky questions trillion dollar plans for OpenAI and Anthropic IPOs

In short: A venture capitalist is warning that possible trillion dollar valuations for OpenAI and Anthropic rely on very optimistic profit expectations.

What's going on

Paul Kedrosky, an investor, laid out his concerns in a recent episode of The Ezra Klein Show about the prices being discussed for potential IPOs (an IPO is when a company first sells shares to the public). His core point is that skepticism about these numbers is not the same as saying AI is useless.

Kedrosky argues that investors may be mixing up two different things, the value of AI infrastructure and the value of the companies making the most advanced AI models. Infrastructure is the expensive “picks and shovels,” like data centers and computer chips that power AI. The model companies are the “miners” trying to make steady profits on top of that spending.

He says the business side of “frontier models” (the biggest, most expensive AI systems) could be fragile. If customers can get results that are good enough from cheaper options, such as DeepSeek, the top labs may struggle to keep charging enough to cover rising costs.

The valuations being discussed are also unusually large by historical standards, according to Kedrosky. Recent reporting cited a possible Anthropic IPO valuation near $2 trillion, with some investors discussing about $1.5 trillion after reviewing financials. OpenAI was reported to be seeking funding at about $1.4 trillion, with a possible listing later.

What to watch

Kedrosky also warns that the risk may spread beyond these companies. If they cannot justify private market valuations, it could affect their ability to raise money and could hurt suppliers that depend on their spending. At the same time, he says AI adoption could still grow, even if today’s leading AI labs do not end up earning enough to match the prices being discussed.

Source: NYTimes