ComputeFinder

Guides

How to get GPUs this week

Where to look for GPU capacity now, how the offer types differ, which free routes exist and what to ask before you commit.

The short answer

Start with how fast you need it and how long you need it for. Capacity you can start today usually comes from providers that list on-demand access. Capacity you can plan around usually needs a conversation, a quote or an application. This guide shows the routes this site has checked and what to ask each one. It gives no prices, because a listed price here would be out of date, and it does not rank anyone.

1. Know the offer types

The glossary explains on-demand, spot, reserved and capacity block. The contract shapes guide shows which providers on this site say they offer which shape.

2. Providers by type

Every name links to its own page here, which lists what we checked on the provider's public site and links to the provider's price list.

Hyperscalers (9)

Very large general-purpose clouds. GPU capacity sits alongside many other services.

GPU clouds (53)

Providers focused on accelerator capacity, including regional and sovereign-minded clouds.

Marketplaces (8)

Sites where capacity from several hosts is listed in one place, often with prices set by hosts or bids.

Decentralised networks (10)

Networks that pool capacity from many independent operators, sometimes paid through a network token.

Public programmes (4)

Public schemes that allocate compute by application, not by the hour.

3. Free and funded routes

52 programmes are listed on free and funded compute. Eligibility, deadlines and amounts change, so read the programme's own page before applying.

National AI resources

Public supercomputing

Research grants

Cloud credits

Free tiers

4. Questions to ask before you commit

Sources

Built from facts already on this site, written on 3 October 2026 and due for review by 3 January 2027. Each fact keeps the check date shown on the page it comes from. This is general information, not advice. Something wrong? Tell us.