Buy the Rumor; Sell the News

Share this post

User's avatar
Buy the Rumor; Sell the News
The GPU market will be bigger than the oil market

The GPU market will be bigger than the oil market

What happens to the market for GPU compute when it grows larger than the oil market?

Dave Friedman's avatar
Dave Friedman
Jul 31, 2025
∙ Paid
10

Share this post

User's avatar
Buy the Rumor; Sell the News
The GPU market will be bigger than the oil market
2
Share

TL;DR: If GPU-hours eclipse crude oil futures in size by the late 2020s, the invisible input of the AI era—compute—will become the most consequential commodity on Earth, reshaping how labs, hyperscalers, chipmakers, and investors operate.

Note that this post makes extensive use of commodities trading jargon, which is likely unfamiliar to my readers who come from the technology world. All jargon is defined in footnotes.


Benchmarking the Unthinkable

Oil is the yardstick of commodity finance. At any given moment, roughly $650-700 billion of crude oil futures sit open on NYMEX and ICE. And even more if you count the options books layered on top.

Don Wilson believes GPU-hour futures will grow bigger than that within years, not decades1. That means billions of GPU-hours deliverable every quarter and enough open interest to rival oil’s liquidity. And he’s put his money where his mouth is: he provided seed investment for Compute.exchange, a Palo Alto, CA-based startup that is building an auction site for spot GPU prices.

Add to this the Trump administration’s recent AI Action Plan. One of its recommended policy actions is: “Ensure access to large-scale computing power for startups and academics by improving the financial market for compute….America has solved this problem before with other goods through financial markets, such as spot and forward markets for commodities” (emphasis is mine).2

So we have at least two important indicators here: (1) the imprimatur of Don Wilson, who has created liquid trading markets for all sorts of exotic assets; and (2) the imprimatur of the Trump administration. Taken together these two facts suggest, if not inevitability, a path dependency that is not easily ignored.

How to falsify this: if by January 2028 open interest in GPU-hour futures hasn’t broken $300 billion, or if installed Tier-1 GPUs remain under ~20 million units, the thesis is wrong.

But if Wilson is right, the world will change. Read on for the roadmap.

Keep reading with a 7-day free trial

Subscribe to Buy the Rumor; Sell the News to keep reading this post and get 7 days of free access to the full post archives.

Already a paid subscriber? Sign in
© 2025 Dave Friedman
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture

Share