One machine. One price.
Billed by the minute.
One rate, one whole dedicated DGX Spark in EU-Central. No shared GPUs, no egress fees, no fine print, and stopped instances don't bill. If you would rather own the hardware, we sell and install it too.
On-Demand
- Nodes
- 1× DGX Spark
- Memory
- 128 GB unified
- Storage
- 1 TB NVMe
- Billing
- Per minute
- Commitment
- None
Reserved
- Nodes
- 1× DGX Spark, 24/7
- Saving
- ~30% vs on-demand
- Extras
- Fixed hostname and SSH port
- Billing
- Monthly
- Commitment
- Cancel monthly
On-Premise
- Nodes
- Yours, in your office
- Delivery
- EU-wide, installed
- Managed
- Updates · support · training
- Test first
- In our cloud, $50 free on first top-up
What a dollar buys, compared
A whole dedicated machine changes the math: no per-seat fees, no per-token pricing, no egress. Rough comparison for a team running a private assistant or serving an open model.
| Option | Monthly cost | Data residency | Trains on your data | Tenancy |
|---|---|---|---|---|
| GPUwerk Reserved | $1,490 flat | EU-Central, dedicated | Never, can't | One machine, one tenant |
| ChatGPT Enterprise (50 seats) | Per seat, quoted on request | US provider | No, by their contract | Shared SaaS |
| A100 80 GB cloud instance | Hourly, varies by provider | Varies by region | No | Virtualized, shared host |
| Own DGX Spark (via us) | Hardware + service fee | Your building | Never, can't | Physically yours |
We quote no numbers for the other options. ChatGPT Enterprise is sold on a negotiated quote with no published list price, and A100 hourly rates differ by provider and region. Ask those vendors for a figure and put it next to ours.
Billing FAQ
How does billing work?
Per-minute, against a prepaid credit balance. You're only billed while an instance is running, stopped instances cost nothing. Your first $100 top-up is credited with an extra $50, and if your balance runs out your instances are stopped automatically, never silently overdrafted.
Are there egress or ingress fees?
No. The hourly rate is the whole price, ingress and egress are included. (The 200 Gb/s ConnectX-7 ports on the Spark are a node-to-node interconnect, not your internet uplink.)
What happens to my data when I stop vs. terminate?
Stopping copies your /workspace off the node into our storage and hands the machine back to the pool. A stopped instance costs nothing. When you start it again we place you on whichever Spark is free and restore /workspace onto it, so your files come back, but the physical machine may be a different one and anything you installed outside /workspace does not survive the stop. Put your data, models and scripts under /workspace and use a setup script for the rest.
Terminating deletes the stored archive too. That is why it asks you to type the instance name to confirm.
How much does on-premise cost?
Hardware plus an optional monthly managed-service fee covering monitoring, updates, support and training. Every setup is sized to your workload, email us with your country and use case and you'll get a written quote, not a sales call.
Top up $100, compute with $150.
The $50 first top-up bonus makes an evaluation day on a dedicated Spark even cheaper.
Deploy a Spark