privatecompute.com
Post the GPUs you have. Post the cluster you need.
The capacity board for dedicated GPUs and private infrastructure. Offers, requests, and two-sided notes, from a single node to a reserved rack, a cage, or the power to run it.
Three kinds of post
Every card on the board is one of these. The contract stays between the two parties.
On the board
Example cards wear a yellow label and are kept out of the search index. Live posts do not.
GB200 NVL72 rack, US East, reserved
72× NVIDIA GB200 NVL72
United States — East · United States
Example offer: one GB200 NVL72 rack on the US East coast, reserved term, delivery window in the first quarter of 2027.
Price on request
Seaboard Blackwell Desk · 4 Oct 202664× RTX 5090 for a two-week rendering burst
64× NVIDIA RTX 5090
United States — West · United States
Example request: a media team wants 64 RTX 5090s in the western United States for a two-week delivery window.
Price on request
Westframe Pictures · 3 Oct 2026Singapore inference cluster, 12-month term
16× NVIDIA L40S
Singapore · Singapore · Singapore
Example request: L40S or similar inference GPUs in Singapore, single tenant, for a 12-month production service.
Price on request
Lantern Inference · 2 Oct 2026Need 8× B200 in EU West for 90 days
8× NVIDIA B200
Europe — West · Netherlands
Example request: a studio wants eight B200s, single tenant, inside the EU, for a 90-day fine-tune starting in November.
Price on request
Kepler Studio · 1 Oct 2026MI300X nodes in Iceland, price on request
8× AMD MI300X
Reykjavík · Nordics · Iceland
Example offer: AMD MI300X 8-GPU nodes on renewable power in Iceland, bare metal, term and price discussed directly.
Price on request
Basalt Cloud Iceland · 30 Sept 202632× H200 SXM, single tenant, Dallas
32× NVIDIA H200
Dallas · United States — Central · United States
Example offer: a liquid-ready HGX pod in Dallas, InfiniBand between nodes, available on a monthly term from November.
$4.85 per GPU-hour
Northline Rack Co. · 28 Sept 2026Browse by hardware and place
Guides for the searches buyers already type, with the live posts that match.
Definition
What private compute means here
Private compute, on this site, means dedicated GPU and infrastructure capacity a buyer can reserve in a known place for a known term. A post is an offer (capacity available), a request (capacity needed), or both (a party that has some infrastructure and needs something else). Private Compute introduces the parties. The contract is theirs.
Questions the first posts raise
What is a private compute listing?
A public post about dedicated capacity. An offer says what you can make available. A request says what you need. A both-post says you have one piece and need another. The place, the term, and the tenancy are part of the post.
Does Private Compute launch the GPUs?
The board introduces the two parties. Billing, access, and the contract happen between them. There is no instance button and no wallet.
Who can post?
Operators with spare or upcoming capacity, teams with a requirement, and anyone who has one and needs the other. A cage, a storage cluster, or a fabric post is as valid as a GPU post.
Is my email public?
Your email stays off the public page. Inquiries are stored on your private manage link, which is the inbox. Email alerts are not turned on yet, so bookmark that link.
What are the example cards?
Example cards show the format of a finished post. They are fictional. Each one is labeled Example, and inquiries on them stay on the board as a demonstration. They are kept out of the search index.