Skip to content

Green DC builds high-density AI capacity outside congested areas

Density, footprint, sovereignty: three constraints that most available colocation does not lift at the same time. Here is what the facility enables, and on what horizon.

Tell us about your needs

The first site is under development. We are not yet selling AI capacity.

What we offer interested organisations: take part in the sizing. We design the facility around real requirements. Telling us yours means shaping what gets built, and holding priority access when the time comes.

Positioning

A complement to hyperscale, not a replacement

Model training is geographically flexible. Inference is not: it remains latency-sensitive, and it will carry the bulk of production workloads. That is what makes edge deployments close to users relevant, alongside hyperscaler infrastructure.

Capacity spread across several mid-sized sites also reduces concentration risk: the risk of depending on a single geographic area, whether Paris or Marseille.

For whom

Who this is for

AI data centre operators

AI and HPC developers and operators that need high-density infrastructure as close as possible to users, for training and above all inference.

Organisations with sovereignty requirements

Public bodies, defence, healthcare, energy and industry seeking sovereign, secure capacity that hyperscale does not always cover.

Why Green DC

What the siting changes for you

01

Density and cooling designed for AI

Sized from the outset for loads an order of magnitude higher, without the structural limits of legacy air-cooled sites.

02

24/7 availability

Backed by local energy generation, independent of any constraints on the electricity grid.

03

Edge proximity

Which reduces latency and the risk of concentrating data in a single geographic area.

04

Built-in regulatory compliance and security

Security-by-design approach, ISO 27001 and 27017 alignment, Uptime Institute frameworks, NIS2 requirements and sustainability reporting.

Technical characteristics

What Green DC will enable

Density per rack

Sized for AI clusters, beyond the 40 kW per rack these loads require and that most of the existing base cannot deliver.

Refroidissement

Direct liquid, built into the design rather than added afterwards.

Energy efficiency

A PUE target well below new-generation benchmarks, which sit under 1.33.

Eau

Closed loop, very low consumption by design.

Power supply

Dual-sourced by design: local energy as primary supply, public grid as redundancy. Grid-light architecture, not off-grid.

Modularité

Capacity grows module by module, each matched to one generation of processors and carrying its own standalone infrastructure.

Let's talk about your needs

Tell us your requirements: they will feed into the sizing of the facility.

Tell us about your needs