idea schoolResearch
Research brief / P-03AI · Infrastructure
Research brief

AI's cloud has a physical bill.

Compute looks digital from the interface. The infrastructure behind it consumes electricity, water, land, capital and grid capacity.

Central question

How do siting, cooling and grid choices change the public cost of each new unit of compute?

The strongest version of the story is not that data centres are inherently good or bad. It is that their resource footprint depends on where they are built, how they are cooled, what electricity they use and how quickly local infrastructure can expand.

Hidden factor

The same amount of compute can have different local consequences depending on cooling technology, climate and grid mix.

That turns AI infrastructure into a planning question as much as a technology question.

Sources

  1. CEEW - data centres, power and water use in India
  2. IndiaAI - compute capacity

Limit

National projections should not be treated as the footprint of every facility. Site-level data is required for local claims.