Compute Exchange vs Opengrid: auctions or routing for GPU compute

Compute Exchange is one of a handful of companies trying to make GPU compute trade like a commodity. It runs auctions and forward contracts for blocks of GPU time. If you are comparing Compute Exchange with other ways to buy compute, this post covers what it does, who it suits, and how Opengrid takes a different route to a similar goal.

Up front: Opengrid is pre-launch. The market view ships first and the routing API is in development. Where this post describes Opengrid, it describes what we are building, not something you can sign up for and use today.

What Compute Exchange does

Compute Exchange runs GPU auctions and forward contracts. Buyers and sellers trade blocks of GPU time through an auction, or agree today on a price for capacity delivered later. The company is backed by Don Wilson, the founder of trading firm DRW.

That model fits how a lot of serious compute is bought. Most new NVIDIA allocation is reserved into 2027, and on-demand capacity has been sold out across GPU types at times. For a buyer who needs a large block of GPUs for months, negotiating a price for future delivery makes sense. So does running an auction, which lets sellers find the market price for capacity.

Compute Exchange is also provider neutral. It is a venue, not a cloud selling its own supply.

Where Compute Exchange stops

Compute Exchange is built around large, negotiated blocks of GPU time. That leaves some buyers outside its sweet spot:

• No API routing. There is no single API call that takes your job's constraints and places it on the best provider right now.
• Blocks, not jobs. If you need 8 GPUs for a training run starting tomorrow, an auction or forward contract is more process than you need.
• Partial view of the market. An auction shows the prices of what is listed in that auction. It doesn't show the whole market, and it doesn't produce reliability data on how each provider performs once your job is running.

That is a matter of focus, not a flaw. Compute Exchange is designed for the large-block end of the market.

How Opengrid is different

Opengrid starts from the job, not the block. You describe what you need in one request:

POST /v1/route · 8× H100 · US · < $2.50/hr · done within 12 hours

Illustrative request. The routing API is in development.

Then routing works in five steps:

1. Describe: GPU type, count, region, price ceiling and deadline.
2. Discover: live prices and availability from every connected provider.
3. Score: rank the options on price, location, reliability and performance.
4. Place: provision on the winning provider through its own API.
5. Bill: one invoice and one dashboard, whichever provider ran it.

Our first buyers are AI teams that are too big for one GPU and too small for a hyperscaler contract. They rent anywhere from 8 to a few hundred GPUs at a time, have no procurement team, and feel every dollar of the price spread. For them, the problem isn't negotiating a forward price. It's knowing where to run this week's job, at a fair price, on a provider that delivers.

Compute Exchange vs Opengrid at a glance

	Compute Exchange	Opengrid
Status	Live	Pre-launch. Market view first, routing API in development
How you buy	Auctions and forward contracts	One API request per job
Typical purchase	Large negotiated blocks of GPU time	On-demand jobs, from 8 GPUs up
Routes jobs by price, region, deadline	No	Yes, that is the core product
Fair-price benchmark	Partial	Yes, from the live market view
Reliability data on providers	No	Planned, from real routed jobs
Provider neutral	Yes	Yes

Which should you use?

Use Compute Exchange if you are buying or selling large blocks of GPU time and want price discovery through an auction, or a forward contract that locks in capacity for later.

Join the Opengrid waitlist if you buy compute job by job. You'd describe the work and its constraints, have it placed wherever price, availability and reliability are best, and get one bill across providers.

If you are weighing the trading side of the market more broadly, our explainer on GPU compute futures covers the new CME and ICE contracts. For the full landscape, see our comparison of GPU cloud aggregators and marketplaces.