TBPN · Sep 28, 2026
With Evan Conrad, Co-Founder and CEO of SF Compute
If there is an AI bubble, Evan Conrad said, venture capitalists are the ones holding it.
Most talk of an AI bubble points at the companies building data centers. Conrad said the risk moves down the chain: cloud providers demand big deposits from AI labs, and the labs raise money at high valuations to pay them.
"So if you're looking for a bubble, it's held currently by the venture capitalists."
Evan Conrad, Co-Founder and CEO of SF Compute, on TBPN, runs a company that calls itself the inventor of compute markets and wants to make the future price of GPU time something buyers and lenders can know in advance.
The full interview is covered here so you can skip it.
Here are the 5 arguments that matter.
Key Takeaways
Computing power today looks like factories before the power grid, each running its own generator sized for peak demand
A cash-settled index is not a real compute market because it is not tied to actual machines, Conrad said
Too few cloud operators know how to run supercomputers for a market to work yet
An independent market operator has to run the data centers itself, down to building some of them
The AI bubble risk is held by venture capitalists, and knowing future GPU prices would reduce it
1. Compute Before the Grid
Asked to introduce the company, Conrad kept it short.
"We make a company called San Francisco Compute. We're the inventor of compute markets." — Evan Conrad
He compared the state of computing power to electricity before the grid.
"So our impression is that right now, compute kind of looks like at the early days before there was a power grid." — Evan Conrad
"And when that happened, every factory ran its own generator." — Evan Conrad
Each factory sized its generator for peak demand and wasted money in idle hours, he said.
"And so what happened in the power markets is that utilities came in, and they pooled demand between all of these different factories." — Evan Conrad
That improved every factory's economics and is how the country scaled up electricity, he said, and it is what compute markets are trying to do. Many new entrants are building an index price with a cash-settled future on top of it, a contract that pays out in money rather than in computer time. He said that approach falls short.
"Our impression is that by doing this, you're divorcing it from the actual technology under the hood because the index isn't connected to anything. Like, you're not actually buying the compute. You're sort of trading on this, like, magical number." — Evan Conrad
