What a Data Centre means now

We don’t store Data.
We build Engines.

“Data Centre” is a name from a different era: rooms of servers holding files, running payroll, keeping a website online. That world still exists. It is not the world we build for.

01From storage to compute

Calling it a Data Centre is like calling a power station a shed.

What we build doesn’t hold information. It trains, infers and reasons, at a scale with no precedent. For thirty years a Data Centre had one job: keep information available, keep it safe, keep the lights on. An AI Engine is a different machine.

Then / The Data Centre

Built to keep.

Utilisation was modest. Density was low. A rack ran cool because it was barely working. The constraint was square footage, and the building was the product.

Now / The AI Engine

Built to think.

Every rack runs at sustained, near-continuous load, every watt doing calculation, not idling. The constraint has moved to power density, cooling capacity and the grid behind the wall. A facility built to store data cannot run compute at this intensity. It has to be engineered for it, from the ground up.

02The requirements

What an AI Engine actually requires.

The preconditions for compute at frontier scale are physical, and none of them can be added later. They are decided in the land, the grid and the design, long before the first rack arrives.

01 / Power

Power, at industrial scale.

Not a generator and a service agreement. Gigawatts, secured years ahead, from grids engineered to deliver them.

02 / Density

Density, not just space.

A single GPU rack now draws in a day what a legacy rack drew in a fortnight. Cooling, power delivery and layout are designed around the chip, not fitted around it after the fact.

03 / Uptime

Uptime as a physical guarantee.

When a facility is training a frontier model or serving inference at scale, downtime is not an inconvenience. It is compute-hours that never come back.

04 / Heat

Heat as an output, not a problem.

An engine running at this load produces heat by doing its job. We design that heat to be reused, not vented and wasted.

Why it matters for how we build
We don’t build Data Centres. We build what Intelligence runs on.

Every site in our pipeline is designed as an AI Engine first, not retrofitted into one. That changes what we look for in land, what we negotiate in power and what we specify in construction, before a single rack goes in. It is why our platform starts with power and origination, not a shell and a services brochure.