Reserved capacity, sold three ways.
Every service runs on the same 256 GPU block, on 12 to 36 month terms. We do not sell tokens and we are not a marketplace.
Three ways to buy
One ledger across all three, so capacity is never sold twice.
Enterprise reserved cluster
Reserved pods from one node up, run the way a US team expects.
- →From 8 GPUs up to 128
- →Reserved from 12 months
- →Managed Kubernetes on bare metal
- →Named support and low latency to US hubs
Capacity partner agreement
Wholesale blocks behind your brand.
- →Blocks from 128 GPUs to full 256 GPU blocks
- →White label and “powered by” friendly, non exclusive by default
- →Backline L2/L3, API and ops handoff
- →Take or pay, 24 to 36 months
Dedicated AI factory
A full block for one customer, designed with you.
- →256 GPUs, 32 nodes, one fabric
- →Topology, storage and network handoff to your spec
- →Inference, training or both on one contract
- →Take or pay, 24 to 36 months, with first right on the next block
Included with every contract
Low latency network
A Spectrum X fabric in the POP and fast paths to the Central US and US South hubs.
Managed Kubernetes
Clusters on bare metal, GPU ready on day one, with your images and schedulers.
Storage tiers
Local NVMe plus parallel file and AI object storage sized to the cluster.
Named support
A named team, runbooks and an escalation path, not a ticket queue.
Monitoring and replacement
Node health monitoring and hardware replacement handled for you.
Capacity planning
Visibility into the next block, so growth is a contract amendment, not a new search.
Cross border AI
Built for AI that crosses borders. Teams serving users, data and operations across the Americas run training and inference close to where it is used, with low latency to US hubs.
What we do not sell
We would rather lose a deal on the first call than on the compliance review.
Not offered
- ✕Tokens or per call model APIs
- ✕Marketplace or best effort hosts
- ✕Spot or preemptible capacity
- ✕US data residency, FedRAMP or US only hosting
Where to go instead
- →Need US hosting? Ask about a split design with a future US site
- →Need tokens? Run your own serving stack on a reserved cluster
- →Need a short test? Start with one reserved node
How an engagement runs
- 01
Request
Size, buyer type and start window. That is enough for a first answer.
- 02
Term sheet
Delivery window, term and commercial structure in writing.
- 03
Contract
Order form with residency disclosed and terms in writing.
- 04
Delivery
Topology design, burn in and acceptance on our side.
- 05
Operate
Named support, monitoring and capacity planning for the term.