You don't need an enterprise IT budget
for a private model of your own.
A dedicated GPU instance in EU-Central starts at $0.79/hour, billed per minute. That's a machine you don't share with anyone else, not a seat on someone else's cluster.
Why smaller companies assume this isn't for them
Dedicated GPU infrastructure sounds like something only a large enterprise budgets for. The assumption is understandable and, for GPUwerk's pricing, wrong.
"We're too small for dedicated hardware"
A single DGX Spark instance is $0.79/hour on demand, billed per minute against a prepaid balance. That's the same rate whether you're a two-person shop or a two-hundred-person company; there's no minimum seat count or enterprise tier gating you out.
"We can't commit to a long contract to find out if it works"
There isn't one. You top up a prepaid balance, run the instance for as long as you need it, and stop it when you don't. A customer-requested stop drops the rate to $0.59/hour reserved instead of the full running rate, so idle time between projects doesn't burn the budget.
"We don't know where to start"
That's what first engagement is for: a scoped, fixed-cost way to work out model choice and setup with us before you commit to running it yourselves day to day.
What it costs, plainly
Prices are in USD, before tax, and billed per minute against a prepaid credit balance.
| Question | Answer |
|---|---|
| On-demand rate | $0.79/hour, single-node DGX Spark, billed per minute |
| Reserved (stopped) rate | $0.59/hour, a 25% discount on the on-demand rate, while the instance is kept allocated to you but not running |
| Minimum top-up | $20 |
| First top-up bonus | Pay $10, get $20 in credit (first top-up matched dollar-for-dollar, up to $10) |
| Contract length | None. You're spending down a prepaid balance, not signing a term commitment. |
Full pricing detail, including the two-node cluster rate, is at /pricing.
A low-commitment way in
If you'd rather not figure out model selection and setup alone, first engagement is a scoped starting point.
Project intake
Define what you actually need the model to do and roughly how much use it'll get.
Assisted setup
Get the instance running and the model or chat interface installed with someone who's done it before.
Handoff
Leave with an instance your team can run day to day, and a clear idea of what it costs to keep running.
Questions we get from small teams
Is dedicated GPU hosting only for large enterprises?
No. A single DGX Spark instance runs at $0.79/hour on demand, billed per minute against a prepaid balance, so a small team can run a private model for the hours it's actually used without committing to enterprise-scale spend.
What if we don't know which model or setup we need yet?
That's what the first-engagement offer is for: a scoped, low-commitment starting point to work out model choice, setup, and rollout before you're running production traffic.
Do we need our own IT team to run this?
You need someone comfortable with SSH and basic Linux administration, or you can book assisted setup as part of the first-engagement offer. There is no managed-service layer beyond the dedicated hardware and support included with it.
What happens if we stop using the instance for a while?
A customer-requested stop keeps the instance reserved at a 25% discount on the on-demand rate, so $0.59/hour instead of $0.79/hour while stopped, rather than billing the full running rate.
Related pages
See what a private model actually costs for your team.
Book a scoped first engagement, or deploy an instance directly and try it yourself.
Talk to us Deploy an instance