Dedicated GPU server hosting, one whole machine for you.
A dedicated GPU server is a physical machine that only you use: no other tenants on the GPUs, the memory or the drives. Ours has eight NVIDIA RTX PRO 6000 Blackwell GPUs with 768 GB of GPU memory, full root access and a 99.5% SLA, hosted in a data centre near Sofia, inside the EU.
FROM $1.70 PER GPU-HOURPOWER, COOLING, INTERNET AND RACK INCLUDED+359 882 490 689WHATSAPP, ANY TIME
A dedicated GPU server, not a slice of a shared one.
Eight RTX PRO 6000 Blackwell Server Edition cards with 96 GB each, in one chassis. Nobody else runs jobs on them during your term.
Your own operating system and your own firewall. We deliver Ubuntu Server 24.04 LTS with NVIDIA drivers, or you install any OS from your ISO through the BMC.
No per-hour metering and no traffic charges. Power, cooling, rack space and a 1 Gbps unmetered uplink are included.
The server runs at Neterra SDC Stolnik near Sofia, Bulgaria, so your data stays inside the European Union.
When a dedicated GPU server is the better choice.
Cloud GPU instances are billed by the hour and suit short tests and jobs that run for a few days. A dedicated GPU server is rented for a month or longer at a fixed price, and suits work that runs all the time. It is usually the better fit when:
- The GPUs are busy most of the month, for example serving a model to users, running inference behind a product or fine-tuning in long runs. Paying by the hour then costs more than a fixed monthly price.
- You need predictable capacity. The machine is reserved for you for the whole term, so there is no queue and no risk that the GPUs are taken when you need them.
- Your data must stay under your control and in the EU. You run the operating system yourself, and we have no accounts on it.
- You resell capacity. GPU clouds and resellers can offer the server to their own customers, under the same acceptable use rules as ours.
If you only need GPUs for a few hours, an hourly cloud instance is simpler. For a steady workload, a whole server for the month removes the hourly price risk. Our guide to published prices for RTX PRO 6000 servers in Europe compares monthly offers side by side.
The specification, as you see it on the machine.
Full details and a sizing guide for large models are on the RTX PRO 6000 server page.
| GPUs | 8 × NVIDIA RTX PRO 6000 Blackwell Server Edition, 96 GB GDDR7 with ECC each, 768 GB in total |
| CPUs | 2 × AMD EPYC 9754, 256 cores and 512 threads |
| Memory | 1.5 TB DDR5-4800 ECC |
| Storage | 2 × 4 TB Samsung 990 PRO NVMe, RAID 1 or your own layout |
| Network | Dedicated 1 Gbps uplink, unmetered (10 GbE port), with a dedicated static public IPv4 address |
| Control | Full root, BMC with remote console, virtual media and power control |
| Data centre | Neterra SDC Stolnik near Sofia, Bulgaria: power A + B, UPS and generators N+1, cooling N+1 |
From $1.70 per GPU-hour, everything included.
The dedicated GPU server costs from $1.70 per GPU-hour, which is about $9,930 a month for all eight GPUs, excluding VAT where it applies. The price is fixed for the term, and longer terms get better rates.
- Power for any load, cooling and rack space in the data centre.
- Internet at 1 Gbps with no traffic charges, and a dedicated IPv4 address.
- Replacement of failed parts and support under the SLA.
- No setup fee and no charge for reboots or reinstalls.
See the full breakdown on the GPU server pricing page.
How to rent the dedicated GPU server.
- STEP 01
Ask for a quote
Tell us your workload, the term and the preferred start date. We confirm availability and send a quote, usually within one business day.
- STEP 02
Sign one agreement
A Master Services Agreement with the SLA attached, signed electronically. The first month is paid in advance by bank transfer.
- STEP 03
Get root access
A clean install, full hardware diagnostics, your SSH keys and BMC access, typically within 5 business days.