{{i.name}}
{{i.spec}}
Per GPU-hour
{{i.avail}}
Get access
Spin up GPUs on live capacity.
Launch your own on-demand GPU cluster in minutes, reserve a dedicated cluster for your next training run, or call models by the token. All on infrastructure VOLT operates.
- GPUs
- B300 · B200 · RTX PRO
- Availability zones
- Europe · Middle East · Asia
- Billing
- Per GPU-hour or per token
- APIs
- OpenAI-compatible
On-demand GPU clusters
Your own cluster, on our live capacity.
Do it yourself: pick a GPU, pick a zone and launch. Scale up or tear down whenever you want. You pay per GPU-hour, with no contract.
- 1Choose your GPUs
From a single RTX PRO to 8× B300 nodes. - 2Pick an availability zone
Run where your data has to stay. - 3Launch
Bare metal, VMs, Kubernetes or Slurm, from the console or the API.
Pricing on request.
Reservations open for 2027
Reserve your cluster before it's built.
Three dedicated NVIDIA Blackwell clusters come online in 2027. Lock in the whole cluster, or just the slice your roadmap needs.
Up to
4,096
GPUs across three clusters
Single tenant
Physically dedicated GPUs, network and storage. Nobody else on your hardware.
Your size
From one 8-GPU node to the full cluster. Grow into it as your roadmap does.
Guaranteed capacity
Your GPUs are committed for the full term, not subject to spot availability.
Engineers alongside
Our team helps you onboard, benchmark and tune your workloads.
Availability zones
Where you can run
VOLT Compute runs in AI-ready datacenters today, and in our own AI Gigafactories as they come online. Choose the zone that matches your latency and jurisdiction needs.
VOLT AI Factory
AI-ready datacenter
Live today
The VOLT console
See and control every GPU you run
One self-service console for your whole stack, from bare-metal nodes to Kubernetes clusters, with live telemetry down to each GPU. Everything in the console is also available through the API.
One platform, every layer of the stack
AI infrastructure
- Bare-metal servers
- Virtual machines
- Managed Kubernetes
- Virtual clusters
- Slurm on bare metal or Kubernetes
Tokens and inference
- Model catalog
- Industry-standard APIs
- Per-token metering
- Serverless pods
- RAG
AI dev and MLOps
- Jupyter notebooks
- Hourly-metered inferencing
- Serverless inferencing
- Kubeflow
- Managed KubeRay
Apps and integrations
- Marketplace and app store
- One-click deployment
- GitOps
- Secrets management
- SIEM and SSO
Underneath: multi-tenancy, IAM and RBAC, quota management, advanced networking, API, CLI and Terraform, billing and metering, and white labelling.
Why VOLT Compute
Cost, control and customization
Three reasons teams move their AI workloads to VOLT.
Cost
We operate the full stack ourselves, from the facility to the GPU. That lets us price at a fraction of hyperscaler rates.
Control
Run mission-critical AI on hardware you can visit, operated by a Dutch company, with a traceable data path.
Customization
Any open-weights model or your own. You decide how it is served, scaled and secured.
Run the leading open models, or your own
- Kimi K3
- GLM-5.2
- DeepSeek-V4
- Qwen3.6
- Mistral Large 3
- Any open-weights model
Private or shared inference, OpenAI-compatible APIs, metered input and output tokens.
Deploy a bare cluster, or start from a template
- AI Studio
- Kubeflow
- NVIDIA Run:ai
- ClearML
Full observability out of the box: in-browser kubectl with zero setup, live CPU and GPU metrics, per-namespace usage, and GPU health down to temperature, power draw and throttle reasons.
Reserve a dedicated cluster
Tell us what you need and we'll come back with availability and pricing within 1 business day.
Live capacity
Capacity available now
GPUZoneAvailableBilling
Console access opens soon. Join the waiting list to be first in.
Need more? Reserve a dedicated cluster