Shadeform alternative: Shadeform vs Opengrid for buying GPU compute

If you rent GPUs from more than one cloud, you have probably looked at Shadeform. It is one of the best-known GPU cloud aggregators: one account, one API and one console in front of 20+ GPU providers. This post covers what Shadeform does well, where it stops, and how Opengrid takes a different approach to the same problem.

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 Shadeform does

Shadeform is a Y Combinator company (S23) that aggregates GPU clouds. Instead of opening an account, signing a contract and learning a new API at each provider, you go through Shadeform once and launch instances across its connected clouds.

That fixes a real problem. The hardest part of using several GPU clouds is not finding them. It is the setup cost each time: a new account and contract, a new API and tooling, a new bill to reconcile. Shadeform takes most of that away for the 20 or so clouds it connects to.

Where Shadeform stops

Shadeform gives you access. It doesn't decide for you. You still look through the options and pick the instance yourself: which GPU, which provider, which region. Three things follow from that:

• No routing by constraint. You can't say "8× H100 in the US, under $2.50 an hour, done within 12 hours" and have the platform place the job. You translate your constraints into an instance choice by hand.
• No fair-price benchmark. Shadeform shows you its connected providers' prices, but not whether those prices are good compared with the whole market. That matters when the same H100 lists anywhere from about $1.25 to $11.07 an hour depending on who you rent from.
• About 20 clouds out of 209. SemiAnalysis tracks 209 GPU clouds. Aggregators reach about 20 of them, so most of the supply is outside the view.

None of that is a knock on Shadeform. It is built to give you access, and it does that. It just isn't built to make the decision for you.

How Opengrid is different

Opengrid starts from the request, not the instance. You describe the compute you need, and Opengrid decides where it runs:

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

Illustrative request. The routing API is in development.

Behind that request, routing works in five steps:

1. Describe: GPU type, count, region, price ceiling and deadline. One request.
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.

Because routing sits on top of a live view of the market, the same system that places your job can also tell you whether the price is fair. Over time, real routed jobs also produce reliability data on each provider.

Shadeform vs Opengrid at a glance

	Shadeform	Opengrid
Status	Live	Pre-launch. Market view first, routing API in development
One API across GPU clouds	Yes, 20+ clouds	Yes, the goal is the whole market
Who picks the instance	You do	Opengrid routes to the best option
Routes by price, region, deadline	No	Yes, that is the core product
Fair-price benchmark	No	Yes, from the live market view
Reliability data on providers	No	Planned, from real routed jobs
One bill across providers	Yes	Yes

Which should you use?

Use Shadeform today if you already know which instance you want and mainly need one account and one API across several clouds. It is live and it removes most of the setup cost.

Join the Opengrid waitlist if you want the decision made for you. That means describing the job and its constraints, having it placed wherever price, availability and reliability are best, and switching providers without changing any code when the market moves. GPU prices now move by double digits in weeks, so the right provider this quarter is often the wrong one next quarter.

For a broader look at the space, see our comparison of GPU cloud aggregators and marketplaces, or read why the same H100 costs so much more at some providers.