The Urban Computer. The Most Energy-Efficient Real Estate in AI Infrastructure.

Umpireal invented the Continuous Flow Digital Junction™, a new category of civil infrastructure. The carriageway rises 2.5 metres, the pedestrian and cyclist underpass drops 1.5 metres, and the semi-excavated shell between them hosts 1 MW of AI compute and 4.5 MWh of battery storage in daylight basement rooms. One structure. Three infrastructure functions: transport, energy, digital. Delivered at approximately €6M per MW, the lowest cost basis of any compute environment on land, sea or in orbit, and the physical answer to the USD 5 trillion Momentum Tax™ the world pays each year at signalised junctions.

In plain terms: the road rises a little, people pass safely underneath, computer rooms sit in the space that creates — and the computer pays for the junction.

What it is for, and what it is not. The Urban Computer is a place to put compute that a city will allow. It is not a replacement for a training campus. Frontier models are built on campuses and will stay there. Today’s largest training sites run from about 100 MW to 1 GW, and the biggest projected runs are larger than any single campus, which is why the industry is working on linking a small number of very large sites rather than building one bigger box. Either way the unit is hundreds of megawatts tightly coupled. It is not a thousand 1 MW rooms. Urban Computers are for what happens next, and for almost everything else: inference, agents, fine-tuning, and regional or sovereign capacity, running next to the people and machines that use it. About 1 MW at a crossing, 10 to 15 MW at a full junction where the ring allows, at roughly €250 per kW per month, a market colocation rate rather than an environmental premium.

Campuses invent the model. Urban Computers run it, store it, and pay for the junction the city was going to buy anyway.