Hyperscale and AI campuses
Campuses for the largest cloud and AI workloads, on confirmed power, engineered for the next two generations of accelerators.
Capacity on dates a cluster roadmap can rely on
Hyperscale cloud providers, frontier AI labs and neocloud operators need large blocks of capacity on dates they can plan around. The platform develops campuses only where grid capacity is confirmed and zoning is in place, so programme dates depend on construction, not on a grid queue.
- Single-tenant campuses and multi-building phased developments
- Transmission-level connections to existing extra-high-voltage substations where available
- Phased power ramps aligned with customer cluster roadmaps
- Master plans with room for expansion, on-site generation and storage
- Pre-let before build, through long-term lease structures
Engineered for density
Facilities designed today will host hardware from 2029 and 2031. Headroom is a design requirement.
Liquid cooling at the rack
Direct-to-chip liquid cooling as standard, with heat rejection sized for racks above 100 kW and a path to the densities of NVIDIA's Vera Rubin and Kyber-generation systems.
Scalable power trains
Electrical rooms, busways and distribution planned for higher densities and for the industry's transition to 800-volt DC architectures.
Efficiency and heat reuse
Efficiency that meets or exceeds local regulation, including Germany's PUE limit of 1.2 for new facilities, with heat recovery designed in.
Three ways to take a campus
Customers choose how much of the facility the platform delivers.
Powered shell
An energised building with grid connection, substation, structure and core infrastructure. The customer completes the fit-out to its reference design.
Build-to-suit
Designed and built to the customer's specification under a long-term lease.
Fully fitted
Ready-to-use white space with power, cooling and racks in place, conventional or modular.