GPU capacity,
metered by the hour.
Bare-metal and instance GPU rental, tuned training clusters and inference endpoints, and wholesale capacity for channel partners. Everything priced in USD and delivered under contract.
Source, schedule, sell.
Pick by workload: from renting one eight-GPU server on demand to a wholesale contract for a whole cluster.
GPU rental
Bare-metal GPU servers and instances by the hour or month, scaling up and down as needed. You control whole nodes without building a machine room.
Training & inference services
Tuned clusters and inference endpoints for large-model training, fine-tuning and serving, billed by compute or by job.
Capacity wholesale
We lock in GPU capacity in bulk and resell it to cloud providers, integrators and channel partners at contract rates — fiat only, with no crypto settlement or custody.
Data center operations & advisory
Deployment, colocation and cluster operations, plus build-out and operations advisory for third-party data centers.
Five layers, from power to API.
The hardware layer is built on NVIDIA data center GPUs, with high-speed interconnect, tiered storage and our own scheduling and orchestration software.
NVIDIA, Hopper and Blackwell are trademarks of NVIDIA Corporation, named here only to describe hardware; no partnership or endorsement is implied.
Three main ways to buy.
Usage, monthly or annual reservation and wholesale contracts are the core; colocation, APIs and advisory are added as needed.
| Model | Billing | Commitment | Notes |
|---|---|---|---|
| On-demand | USD / GPU·hour | None | Start and stop any time; billed on actual hours |
| Reserved | USD / month · year | From 1 month | Dedicated nodes at a fixed rate, for continuous training |
| Wholesale | USD / contract rate | Long-term | For cloud providers, integrators and channel partners |
| Jobs | USD / compute · job | Per job | We run the cluster; you submit jobs |
| Colocation | USD / rack·month | By contract | Rack space, power and remote hands for your hardware |
| Data center advisory | Per project | — | Site selection, build-out and operations |
From quote to running.
Every customer goes through identity verification and end-use review. Export rules require it, and it keeps your workload out of trouble.
Requirements & quote
- GPU model, count, region, term
Identity & end-use
- Restricted-party screening
- Signed end-use statement
Contract
- Master agreement and order form
- Data processing addendum if needed
Activate & operate
- Access credentials delivered
- Support under the SLA
Six structural advantages.
Cost structure
Lower power and land costs at overseas facilities support better unit pricing.
Hardware to content
From GPUs at the bottom to AI content generation at the top, on our own capacity.
Idle hours absorbed
Our content business absorbs idle capacity, lifting utilisation and smoothing cash flow.
Compliance-first
USD pricing, services not investments, no crypto custody.
Distributed supply
Facilities in several regions spread risk and add elasticity.
US counterparty
You contract with Qovane Technologies Inc., a New York corporation, under New York law.
Frequently asked questions
How is compute priced and paid for?
Everything is priced in US dollars — per GPU-hour, by monthly or annual reservation, or on a wholesale contract — and paid by bank wire or ACH against a service contract and invoice. We don't accept cryptocurrency.
What is the minimum term?
On-demand instances have no minimum and are billed on actual hours used. Reserved capacity starts at one month. Wholesale capacity is contracted long-term.
What do you need from us before activation?
Company registration details, information on any shareholder owning 25% or more, and a signed end-use statement describing how and where the compute will be used. Export-control and sanctions rules require it; once verified, we sign the service agreement and activate.
Can you serve customers in China?
You can apply, but US export controls apply strictly. We deliver primarily as remote compute, review each request by country, chip model and service type, and have counsel sign off before activation. Some requests may not be serviceable.
How is our data protected?
Your data stays yours and is never used to train our models. Bare-metal nodes are dedicated per customer, networks are isolated by tenant, and data center partners have no access to customer data. A data processing addendum is available if you process personal data.
Which GPUs do you offer?
We are planning deployments of NVIDIA Hopper and Blackwell data center GPUs and later generations. Specific models, quantities and regions depend on availability at the time of quoting.
Can we resell access or mine cryptocurrency?
Reselling or transferring access requires our written consent, and cryptocurrency mining is not permitted. Partners who want to resell can apply to our channel program.
How many GPUs?
Tell us the model, count, region and term.