GPU Serverpowered by OpenStack

GPU performance for AI, simulation and rendering — on Swiss infrastructure. stepping stone offers dedicated NVIDIA RTX PRO 6000 Blackwell GPUs with up to 96 GB of vRAM: full access, no shared resources, AI-ready from the very first minute.

With stepping stone’s GPU servers, you get dedicated computing power for AI workloads, physical simulations and 3D rendering — hosted in Switzerland. Each virtual machine has full access to NVIDIA RTX PRO 6000 Blackwell GPUs with 96 GB of vRAM per GPU, and you can combine up to eight GPUs per VM.

The resources are fully dedicated: 1 to 256 vCPUs and up to 2 TB of RAM, with no CPU overbooking. GPU access is via PCIe passthrough — custom drivers are supported, and the latest drivers are regularly updated and maintained by stepping stone. AI-ready images are available to get you started quickly: Ollama, vLLM, Open WebUI and Hugging Face CLI are pre-installed, along with accompanying guides, so you can be up and running within minutes. stepping stone operates its GPU infrastructure entirely in Swiss data centres — your data and models remain in Switzerland.

Research teams, developers and companies that need GPU performance for computationally intensive workloads — without having to purchase their own hardware and without handing over data to US cloud providers.

Typical areas of application: LLM inference and training on local infrastructure, reinforcement learning and machine learning workflows, physical simulations (molecular dynamics, protein folding), 3D rendering and visualisation, AI prototyping using pre-installed tools and guides.

Swiss data centres. Dedicated NVIDIA GPUs. No overbooking.

NVIDIA RTX PRO 6000 Blackwell with 96 GB vRAM per GPU, up to 8 GPUs per VM. Full GPU access via PCIe passthrough — custom drivers supported. AI-ready images with Ollama, vLLM, Open WebUI and Hugging Face CLI pre-installed — plus guides to get you started straight away. Up-to-date drivers, regularly updated and maintained. Dedicated vCPUs and RAM with no overbooking. Personalised support from specialists who know the stack inside out. Personalised advice and operations provided by stepping stone in Bern.

Scope of services

Dedicated GPU infrastructure

NVIDIA RTX PRO 6000 Blackwell with 96 GB of vRAM per GPU, up to eight GPUs per virtual machine. Full GPU access via PCIe passthrough; custom drivers are supported. In addition, dedicated vCPUs (1–256) and up to 2 TB of RAM — with no overbooking. The latest drivers are regularly updated and maintained by stepping stone.

AI-ready from the very first minute

Pre-configured images featuring Ollama, vLLM, Open WebUI and Hugging Face CLI — pre-installed and ready to use. Accompanied by step-by-step guides and documentation to help you get up and running in minutes. No setup hassle, no driver headaches.

Supervised operation

Deployment, monitoring, maintenance and support on Swiss infrastructure, with personalised advice. Personalised support from specialists with in-depth knowledge of the GPU and AI stack. Can be combined with managed backup, monitoring and on-call support as required.

Related Products

GPU Server cost calculator

0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
0.00 CHF/Month
All prices are in CHF, excluding VAT.
190.00 CHF/Month

Frequently Asked Questions (FAQ)

Cloud Hosting ist eine schnellere, stabilere und zuverlässigere Alternative zum Shared Hosting. Das Managed Cloud Hosting bietet im Vergleich zum herkömmlichen Webhosting eine 4-fach höhere Geschwindigkeit und 20-fach mehr Ressourcen. Außerdem ist es vollständig verwaltet – das heißt, wir kümmern uns um alle technischen Belange, damit Sie sich auf die Erweiterung Ihrer Projekte konzentrieren können. Es ist die perfekte Lösung für Website-Besitzer, die die Leistung und Stabilität eines virtuellen privaten Servers (VPS) suchen, ohne die Komplexität der Verwaltung eines solchen.

Yes, we offer stoney backup.

To create ed25519 SSH public/private key pair, you need OpenSSH 6.5 or newer. ssh-keygen -t ed25519 -C "name.surname@example.com" Die Ausgabe von ssh-keygen sieht in etwa wie folgt aus: Generating public/private ed25519 key pair.
Enter file in which to save the key (/home/steppingstone/.ssh/id_ed25519): 
Enter passphrase (empty for no passphrase): 
Enter same passphrase again: 
Your identification has been saved in /home/steppingstone/.ssh/id_ed25519.
Your public key has been saved in /home/steppingstone/.ssh/id_ed25519.pub.
The key fingerprint is:
SHA256:DpS2gFLsSAfCuoBjgf7UEsQY3lKrDbEpnOn+rIlg+6k name.surname@example.com
The key's randomart image is:
+--[ED25519 256]--+
|+=Bo |
|=*X=. . |
|B%+o+ + |
|O+*o = . |
|o=o.. o S |
|o . o |
|.o . |
|+ = . |
|.E+= |
+----[SHA256]-----+
RC=0 steppingstone@lale1202:~$ cat .ssh/id_ed25519.pub 
ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAID5DuOj3fhVqQpeXTRa2Tf8OXm1OO9AmCKs312770odA name.surname@example.com

The first one mentioned uses decimal (SI) and the second one binary (IEC) unit prefixes. The actual sizes are shown in the table beneath.

DecimalBinary
SI unitMeaningIEC unitMeaning
Kilobyte (KB)1 000 ByteKibibyte (KiB)1 024 Byte
Megabyte1 000 000 ByteMebibyte (MiB)1 048 576 Byte
Gigabyte (GB)1 000 000 000 ByteGibibyte (GiB)1 073 741 824 Byte
Terabyte (TB)1 000 000 000 000 ByteTebibyte (TiB)1 099 511 627 776 Byte
Petabyte (PB)1 000 000 000 000 000 BytePebibyte (PiB)1 125 899 906 842 624 Byte
Exabyte (EB)1 000 000 000 000 000 000 ByteExbibyte (EiB)1 152 921 504 606 846 976 Byte

More information can be found on Wikipedia.

The bandwidth of stepping stone AG will be calculated with the 95%-rule.The 95%-rule is a common standard. 

Calculation of the 95%-rule

When you have ordered a bandwidth of 1 Megabit/s and are connected to a 100 Megabit/s line, the Committed Data Rate (CDR) is 1 Megabit/s with a burst rate of 100 Megabit/s. The Committed Data Rate will be invoiced independently of the usage, as this bandwidth is fixed reserved for you. The burst rate of 100 Megabit/s is “best effort” and will be invoiced according to the 95%-rule, which is calculated as follows: