Buyer
AI labs, research groups and enterprises that need blocks of 8, 32 or 72 GPUs on bare metal.
- Can post
- Submits RFQs, which post bids. Takes asks.
- Needs
- An account to submit. Funds verified by the exchange to fill.
The order book for GPU capacity
Submit an RFQ for a block of GPUs and it becomes a live bid at your price on GPU SLA Exchange. Whitelisted sellers and the house post asks. When they meet, the SLA is issued.
Prices are USD per GPU-hour. 72 GPUs equals one NVL72 rack. Verified funds earn the badge; unverified bids still show on the book.
GB300-1Y
No priceindicative · per GPU-hour
Top of book
Server read at · USD per GPU-hour
Seeded demonstration book. 72 GPUs equals one NVL72 rack. Full depth on the market page.
How it works
01
Create an account, pick the GPU and term, and name your price and quantity. The exchange posts a bid at your price on your behalf and shows it to you resting on the book.
02
Ask us to verify your funds. Once we have checked the proof and recorded the amount, your bid carries the verified badge, locks a 20 percent deposit, and can fill against an ask. Unverified bids still show, so sellers see real demand.
03
Only the house and sellers the exchange has whitelisted post asks. A fill issues the SLA on the spot, the deposit moves into escrow, and activation runs with a human in the loop.
Verified and unverified
Live version on the GB300-1Y ticker.
Capacity, racked
A filled SLA is a real block of GPUs with a real activation date. Follow it from print to proof of SSH in the public ledger.
Who is on the book
AI labs, research groups and enterprises that need blocks of 8, 32 or 72 GPUs on bare metal.
Neoclouds, SPVs and operators with GPUs to run or the funds to deploy them, reviewed by the exchange.
SLAX Market Making SPV, the exchange's own market maker with capital committed to racks.
Pricing reference
| Seller and product | USD per GPU-hour | Source | As of | Confidence |
|---|---|---|---|---|
| GMI CloudGB200 NVL72 | from $8.00per GPU-hour | GMI Cloud pricing page and blog | Medium confidence | |
| CoreWeaveGB200 NVL72, 4-GPU instance | $10.50per GPU-hour | CoreWeave GPU pricing page | High confidence | |
| AWSGB200 UltraServer capacity block (u-p6e-gb200x72) | $10.58per GPU-hour | AWS Capacity Blocks for ML pricing | High confidence | |
| Oracle CloudGB200 NVL72 (BM.GPU.GB200.4) | $16.00per GPU-hour | OCI Price List API, part number B110979 | High confidence | |
| Oracle CloudGB300 NVL72 (BM.GPU.GB300.4) | $18.00per GPU-hour | OCI Price List API, part number B112140 | High confidence | |
| Microsoft AzureND GB200 v6, 4-GPU VM (Standard_ND128isr_NDR_GB200_v6) | $27.04per GPU-hour | Azure Retail Prices API | High confidence |
No provider publishes a one-, two- or three-year committed price for a full NVL72 rack. The book exists to print one.
Methodology and sourcesWho is making the market
At launch the ask side of the book is made by SLAX Market Making SPV, a partner vehicle with capital committed to deploying NVL72 racks once an offtake SLA exists. Its asks are real inventory: a filled trade on those asks becomes a rack order.
Any neocloud or provider the exchange whitelists can post asks on the same tickers, and any holder can resell an SLA before activation. The spread narrows as more of them do. Trades and every balance movement print in the public ledger, so the history of the market is checkable by anyone.
Questions
Create an account with your work email, submit an RFQ, and watch your bid rest on the book. Ask us to verify your funds when you are ready for it to fill.