northsea-1Request early capacity

Europe’s next-generation
AI infrastructure.

Run, train and evaluate models.
We’re building specialized compute
with a path to North Sea renewable power.

Request early capacity Explore the offering
Built for serious AI workloads.

Frontier labs Neolabs Agentic platforms Industry

Run. Train. Evaluate.

Your models. Our infrastructure.

Production inference.

Put models to work. Serving built around your model, latency requirements and production load.

You bring
Your model and service requirements
We deliver
A serving endpoint, with agreed throughput and latency
Discuss inference
Model versions, evaluations and run records connected through one software layer to different compute architectures.
ModelProduction endpoint

Specialized silicon.
One software platform.

Your models connect through the northsea-1 software platform to three specialized compute architectures.

More compute for your budget.

Hardware selected around your model, throughput and latency requirements.

We handle the hardware.

We port, optimize and operate your models across supported architectures.

Your workflow stays together.

Versions, evaluations and deployments stay connected as the fleet evolves.

Rent capacity.
Keep control.

Direct compute for teams running their own stack.

By the hour
For experiments and variable workloads.
Dedicated clusters
A cluster allocated to your team’s training and production workloads.
Reserved capacity
Agree an allocation and term around your recurring demand.
Discuss capacity
A customer's software stack connects directly to a Northsea-operated compute cluster.
You controlModels, software & jobs
We operateCompute, network & cooling

Renewable power.
Greater control.

We’re building toward long-term renewable supply and direct North Sea integration.

North Sea offshore wind beside a coastal datacentre campus, illustrating the planned link between compute and renewable power.

Less exposure to power-price spikes.

We’re pursuing long-term renewable supply agreements to make compute costs easier to forecast across a multi-year deployment.

Meet your renewable-energy requirements.

Our plan includes documented power sources and renewable coverage, giving procurement the evidence it needs to assess and approve capacity.

Before we talk.

Can we bring our own stack?

Yes. Our direct-capacity offering is for teams that want to operate their own workloads. Managed inference, training and evaluation are separate options for teams that want us to handle delivery.

Which models and hardware will you support?

We qualify specific model and hardware combinations against workload requirements. Tell us your model family, size, memory needs and latency or throughput targets; those determine what is suitable.

How will commercial terms work?

Direct capacity can be scoped by accelerator hours, a dedicated cluster or a reserved allocation. Managed work combines the required compute with a defined delivery scope or recurring support. We agree the compute budget and service requirements with you before work starts.

Bring us
your workload.

First capacity targeted for Q2 2027.

Subject to financing, hardware procurement and deployment. Timing may change. Enquiries are non-binding and do not reserve capacity. Delivery commitments require a separate signed agreement.

1 / Your workload2 / Your team

What do you need?

Select all that apply. Leave open anything you’d like to discuss.

Use cases
Capacity preferences