A startup called Ornn just raised $33 million to do something that sounds simple but could reshape how the AI industry buys and sells its most precious resource: computing power.
The bet is straightforward. Just as oil traders use futures contracts to lock in prices and hedge against volatility, AI companies should be able to do the same with GPU capacity. Right now they can't, which forces them to rely on expensive long-term supply agreements and hope prices don't spike.
Backed by Andreessen Horowitz, Ornn is one of several startups and exchanges trying to build that missing infrastructure. The timing matters. Goldman Sachs estimates that between 2026 and 2031, roughly $7.6 trillion will flow into compute, power, and data center buildout globally. The financial plumbing to handle that scale, the bank notes, hasn't been built yet.
"We want to make financing AI way more seamless," Wayne Nelms, Ornn's chief technology officer, told Axios. CEO Kush Bavaria framed the effort as part of America's competitive edge against China and said the company doesn't work with Chinese AI labs.
Here's where it gets tricky. Unlike oil barrels sitting in a tank, compute is volatile in ways that defy traditional commodity logic. Each new generation of Nvidia chips offers better performance, which instantly devalues the previous generation. Unused GPU capacity also expires the moment it's not being used. There's no storage, no shelf life, making it nearly impossible to build standardized contracts around.
Ornn is working around that by becoming a pricing hub. The platform integrates with Bloomberg Terminal and other data services, letting traders check GPU prices through tools they already use. Lenders can benchmark compute costs. Buyers and sellers can hedge their positions. It's building the basic infrastructure first, letting the market develop from there.
Regulatory hurdles loom, but Ornn is operating under a de minimis exemption that gives it room to move. Larger exchanges like CME and the Intercontinental Exchange are working through approvals and plan to launch their own compute futures contracts tied to pricing benchmarks, including Ornn's index.
The skeptics have a point. Compute may never look like oil markets. The asset depreciates too fast, it can't be stored, and standardization remains elusive. But with $7.6 trillion potentially at stake over the next five years, Wall Street has already decided to try.
Author James Rodriguez: "This works only if the market can move past the fundamental mismatch between compute and physical commodities, and that's not a given."
Comments