In developmentThe workload comes first.
For customers who need compute capacity, not a facility. We are developing a managed GPU offering on dedicated energy and infrastructure. The U.S. pilot is the starting point.
Illustrative example · index of 6 Oct 2026Requirement explorer
Establish the cluster, storage and interconnect requirements alongside power and cooling.
The development scope follows the workload. Tell us what you intend to run, the control you need, your preferred location and the expected duration. You provide the workloads, software and data.
Requirement explorer, not a capacity quote. Live GPU availability and service launch dates are not currently published.
Treat the cluster as a system.
- 01Private capacity with a defined purpose.
Develop dedicated clusters and bare-metal infrastructure around training, fine-tuning or inference requirements.
- 02Treat the cluster as a system.
Align interconnect, network architecture and GPU-optimised storage with the workload, data and operating model.
- 03Connect compute to the physical design.
Bring rack density, cooling, power distribution and commissioning into the same deployment specification.
- 04Define the service before deployment.
Develop compute-management, support and service responsibilities with specialist providers and the customer.
Engage today.
Discussions · Pilot · Intended servicesWe welcome discussions with AI teams and enterprises about workload, scale, location and commercial requirements.
NowRequirements and development discussionsRequirements and development discussions
The pilot is being developed to bring dedicated energy, modular infrastructure and compute integration into one deployment.
In developmentIntegrated U.S. pilotIntegrated U.S. pilot
Dedicated and managed GPU capacity are intended services, subject to commissioning and agreed terms.
Live GPU availability and service launch dates are not currently published.
Intended service capabilityDedicated and managed capacity
Shape your compute requirement with us.
For dedicated GPU infrastructure, private AI compute or future managed capacity.