OpenAI lowers its funding target: the clouds set the price

OpenAI has lowered the funding target for its next round, as TechCrunch reported on October 31. A Bluesky post from @carnage4life.bsky.social records more than $3 trillion in future spending commitments among the five biggest cloud companies. The smaller round says the market is cooling. The trillions say it is not. One is a measure of the industry. The other is a set of terms. Which one is the measure?

A reading that places the two numbers side by side looks like a contradiction until you see what the trillions actually are. They are not a valuation; they are a price for access, and the round has to be priced to fit them. The clouds and the investors behind them dominate the industry’s capital, and they want large stakes.

Jared Tame’s book Startups Open Sourced describes the same arithmetic. The author says a round has two numbers: how much money you want and how much of the company an investor wants. When investors demand larger stakes per dollar, the company brings in less. The round cannot give them the price the old arithmetic implied, so OpenAI brought in less.

The common assumption is that a smaller round proves the AI boom is cooling. That assumption fails. The capital has moved out of startup rounds and into the cloud commitments, so the round is no longer the right measure of the industry’s health. A smaller round no longer means less interest. It means the interest moved somewhere else, and the round no longer holds it.

If that holds, the next round OpenAI raises will be less a measure of its own conviction than a formality for access the clouds have already promised. The trillions already committed could keep multiplying across more labs and regions, with Google, Meta, Microsoft, Amazon, and Oracle deciding who gets access and on what terms. The round would become a formality. Access would become the real prize.

A funding round no longer measures the AI boom. The cloud commitment is the real measure, and the clouds (Google, Meta, Microsoft, Amazon, and Oracle) set the terms for every lab that wants in.