Supply on a schedule
you can sell against.
For GPU clouds covering committed capacity and intermediaries reselling to a network or program: dedicated clusters handed over in dated waves, with root on every node.
Committed capacity, delivered in waves.
Cover a commitment or a slipped delivery with capacity that arrives on dates set at signing. What you sell on top is your business.
- Dated deliveryLarge clusters are handed over in monthly waves of 32 nodes, with dates counted from signing.
- A cost base you can planOne 60-month commitment under capacity you resell by the hour or the month.
- Root on every nodeBare metal with no hypervisor. Run your own scheduler, images and billing. Firmware and BMC stay with us.
From a signed commitment to capacity on the floor.
Step 1: Map your commitments
Tell us what you have sold and when it starts. We turn it into a wave plan of monthly waves of 32 nodes.
You getA wave plan against your datesStep 2: Sign and prepay into escrow
Proposal, then MSA. Dates run from signing, and the prepayment is held in escrow.
You getA dated scheduleStep 3: Receive each wave
Every wave is burned in before handover. You get root and bring your own OS image and scheduler.
You getRoot on every nodeStep 4: Sell it on
Resell by the hour or the month to your own customers, through your own channels and billing. We keep the cluster running.
You getOperations to an SLA
The terms operators plan around.
Every cluster follows the same contract. Rates are on the pricing page.
- Term
- 60 months, prepaid
- Delivery
- In monthly waves of 32 nodesDates counted from signing.
- Access
- Bare metal, root on every nodeYour OS image and scheduler. Firmware and BMC stay with us.
- Network
- RoCE v2 Ethernet, spine-leafNon-blocking on request.
- Exit
- Exit option in the contractYour proposal sets the terms.
Questions buyers ask first.
Something missing? Ask us directly. We usually reply within a few hours.
Yes, in bare-metal mode. You get root on every node with no hypervisor and run your own OS image and scheduler. Firmware and the BMC stay with us, and we monitor the cluster and replace failed parts.
Yes. The cluster is single-tenant and you decide how to sell it, by the hour, the month or on your own contracts. Billing your end customers is yours to run.
In monthly waves of 32 nodes for larger clusters, each burned in before handover. HGX B300 takes about 10 to 14 weeks from signing, and we set the wave dates in the proposal.
8 HGX nodes (64 GPUs) or one NVL72 rack (72 GPUs). Most clusters are 32 to 128 nodes; larger ones are scoped case by case.