Skip to content

Why Hexi

We watched AI wait for power. So we started building it.

Hexi Industries is UK veteran-founded. It started inside an AI lab, watching the same thing happen again and again: the models were ready, the chips were bought, and the work sat waiting for power and a building to put it in.

The grid can't move faster than the queue in front of it, and new connections in parts of the UK now take a decade or more. But plenty of sites already have spare power. So we build AI compute on it, and we're working towards the day compute brings its own.

Built by people who've seen the problem from both ends.

Our founding experience spans the armed forces, the nuclear sector and AI labs. Between them they cover how power is made, how compute is used, and how hard things get delivered on time.

From service
Planning that holds up when conditions change, logistics that arrive on the day, and one person accountable for the outcome.
From nuclear
Safety cases, engineering discipline and the habit of proving a design before anyone builds it.
From AI labs
How compute actually gets used, and what it costs when the GPUs are bought and the power isn't there.

More useful compute from every megawatt. Then more megawatts.

Each square below is drawn to scale: its area is its power. The first is what we deliver now. The last is where we're going.

  1. Today1.05 MW

    A Hexi pod on power a site already has. Factory-built, liquid-cooled, running in about 15 months instead of waiting years for the grid.

  2. Next100 MW

    Campuses of pods, designed and run as one. The engine already plans at this size: our Texas showcase is a 100 MW campus.

  3. Then470 MW

    Campuses on their own small reactors, so compute is no longer waiting on anyone's grid. One reactor the size of the Rolls-Royce SMR design.

Pod 1-7 IT load from the Hexi engine. Reactor size from the Rolls-Royce SMR design (470 MWe). Small reactors are our long-term goal, not something we sell today.

One engine, from the first pod to the first reactor.

The Hexi engine is a model of the real world a facility has to survive: grid connections and equipment lead times, cooling against hourly weather, outages, costs and cash. On top of it, an evolutionary search breeds thousands of candidate designs and keeps the ones that get compute running soonest at the lowest risk-adjusted cost.

The same engine sizes a single pod today and plans campuses at 100 MW. Every pod we deliver will make it sharper.

See how the engine works

We design from evidence.

Every pod starts from your site's own supply, weather and access, and every assumption carries a source and an owner.

We say what we don't know.

Estimates are marked as estimates until a manufacturer's quotation makes them firm.

We tell you when we're not the answer.

If cloud or colocation suits you better, the engine says so, and so do we.

It starts with the power you already have.

Check my site, free Configure your pod in 3D