Back To Homepage

Field Notes / AI Infrastructure / October 2026

The real challenge behind a 100 MW AI data centre.

The AI revolution is often described in algorithms and chips. But behind every model sits an immense physical economy of electricity, water, cooling, land, connectivity, people and capital.

100 MWHypothetical continuous power draw
876 GWhElectricity over 365 days at full 100 MW draw
2M L/dayIllustrative freshwater demand for some cooling designs, not a universal requirement

Start with power.

A data centre drawing 100 MW continuously uses 2,400 MWh in a day, or 876 GWh in a 365-day year. This is an arithmetic illustration, not a forecast of Setupuram consumption. Actual energy use depends on the built and commissioned capacity, occupancy, load, power usage effectiveness and operation.

Then confront water.

Some evaporative cooling approaches may use roughly two million litres of water daily at this scale. Other approaches can use far less potable water for cooling. The comparison depends on climate, cooling architecture, measurement boundary and operating conditions. The serious question is not simply how much water a data centre uses, but which water, from where, and at what environmental cost.

Design for the place, not just the machines.

For the proposed Setupuram AI Data Centre, the concept is to investigate seawater-assisted heat rejection via heat exchangers and sealed loops, targeting no drinking water for primary cooling. That is a design objective at pre-feasibility stage, contingent on engineering, marine and environmental studies, and approvals. It does not mean zero environmental impact or zero water use across every activity.

Capital has to work with physics.

A project of this size is not only a technology investment. Grid availability, permits, marine works, operator selection, anchor demand, financing, construction sequencing and risk allocation must converge. That is what makes institution-building as critical as engineering.

The inclusive-growth question.

Can new AI infrastructure create local capability, employment, supplier opportunities and wider access to compute, while protecting the communities and ecosystems around it? I believe that should be part of the design brief from day one, not an afterthought.

Project note: The proposed first phase is 100 MW total facility power on approximately 200 coastal acres. Tech Innovations coordinates consortium development; Setupuram Limited is the landholding participant. The project remains at pre-feasibility stage. This article is a personal perspective and does not represent approvals, financing commitments or a completed facility.

Further reading: setupuram.ai · techinnovations.ai