Delivering AI Factories for the Intelligence Age
Discuss your IT capacity needs
We deliver purpose-built AI factories in record time
We bypass Europe’s grid constraints and bring inference capacity faster to market than anybody else by plugging prefabricated data centers into existing power connections or building on-site power plants.
With Calcore
~18 months
Site selection to RFS
Live
Conventional data center
4 to 6 years
Grid connection to RFS
0
18 M
36 M
54 M
72 M
Designed for the densest compute
Proprietary design to host the new generation of hardware.
Density-first architecture
Our reference design is built for today's 150+ kW liquid-cooled racks and engineered for the next generation of AI accelerators. Every layer - floor loading, distribution, cooling loop, containment - is sized for what is coming.
Direct liquid cooling
Direct-to-chip liquid cooling on closed loops, with dry coolers rather than evaporative systems. Higher heat capture, higher rack density, and near-zero operational water consumption.
Networking for frontier workloads
High-bandwidth fabric for large-scale distributed training, and low-latency connectivity for inference at the edge of demand. Fiber routes secured as part of site origination.
A power chain designed with the generation asset
Because we co-develop the energy asset and the datacenter as an integrated project, the electrical architecture runs from the generation unit to the chip as a single design: on-site substation, BESS, and distribution sized together.
Standardized, modular, and built to be repeated
Speed comes from optimization and repetition. We build from a standardized, modular reference design - prefabricated blocks, factory-assembled and commissioned before they reach the site. Each new project reuses a design that has already been engineered, permitted and built. That's how we deliver in 18 months where the industry standard is measured in half-decades.
Electrical and cooling enclosures
Skid-mounted, pre-wired and pre-tested electrical and cooling modules, including transformers, switchgear, UPS systems, chillers, CDUs, pumps and heat exchangers, that eliminate months of on-site electrical and mechanical work from the critical path.
kW / rack
Customed datahall design
Each data hall follows a standard design tailored to customer hardware requirements, with appropriate floor loading, power distribution, cooling and containment.
Replicable across the platform
We build from a standardized, modular reference design - prefabricated blocks, factory-assembled and commissioned before they reach the site. One design, deployed across our European pipeline.
For AI workloads
Our sites are built for three core use cases: large-scale training, latency-sensitive inference and sovereign AI. Each places different demands on the facility, but all three depend on the same fundamentals: reliable power, secure infrastructure and capacity located close to demand.
AI model training
Hyperscale capacity
100 MW+ campuses, deployable in phased, prefabricated blocks.
High-bandwidth interconnects
Ultra-high-speed, low-latency networking for distributed training across thousands of accelerators.
High-density power and cooling
Infrastructure designed for sustained, high-power GPU/accelerator clusters.
Large-scale inference
Low-latency connectivity
High-performance network infrastructure to serve real-time AI applications.
Proximity to demand
Geographically distributed capacity close to end users and major demand hubs.
Power efficiency at scale
Optimized power distribution and cooling to minimize operating costs per token.
Sovereign & sensitive AI deployments
European sovereignty
Sites located in Europe, supporting applicable data-residency and regulatory requirements.
Dedicated infrastructure
Isolated capacity and controlled environments tailored to customer security requirements.
Enterprise-grade SLAs
High availability, robust physical and cybersecurity controls, and Tier 1 operational services.
Our pipeline of AI Factories in Europe.
- NRD #1 RTBNormandy, France Q4 2027 22 MW
- GRD #1 RTBGironde, France Q1 2028 15 MW
- FRA #3 In developmentFrance Q3 2028 60 MW
- FRA #4 OriginationFrance Q4 2028 60 MW
- FIN #1 OriginationFinland Q4 2028 60MW
- FRA #4- Second Phase OriginationFrance 2029 40 MW
- FRA #3 - Second Phase OriginationFrance 2030 210 MW
- SWE #1 - Second Phase OriginationSweden 2029 60 MW