This is a sister volume to Dart Consultants' "India's AI Compute Stack" primer, and it exists because that report deliberately excluded an entire half of what a data centre actually is. A GPU cluster does not run on silicon alone — it runs on a transformer that steps grid power down to usable voltage, a switchgear yard that keeps that power segregated and protected, a cooling loop that removes the heat 120-kilowatt racks throw off, a building shell that holds all of it together, and the cable plant that connects the finished result to the rest of the world. None of that is optional, none of it is glamorous, and almost none of it shows up in a headline about "India's AI opportunity." This report is about that other half.
Pouring a warehouse-grade concrete shell to house a data hall is a scaling problem: pour the slab, raise the walls, hire the labour, pass the building inspection. Building — and then never letting fail — the transformer, switchgear and cooling plant that keep 50-100 megawatts of continuous, redundant power and heat-removal capacity alive without a single unplanned outage is a different kind of problem entirely, and it is the one almost no published statistic about "India's data-centre boom" actually measures. Every company report in this document is really answering one question: how much of this company's revenue sits on the engineered, currently-scarce side of that line, versus the commodity, competitively-bid side.
A note on sourcing: every figure in this report carries a source and a date next to it. Where published estimates disagree — and in this sector, on India's data-centre power-demand forecasts especially, they disagree by a factor of three or more — we show the disagreement rather than picking a number silently. Where we could not verify something, we say so. See the closing Notes section for the full methodology and data caveats.
