This is a first-principles primer, not a broker initiation note with an industry preamble bolted on front. The order is deliberate: the physical and commercial reasons a chip becomes a working AI cluster come first; every fact that follows — margin, order book, valuation — is presented as a consequence of that difficulty, not as a separate story. If you finish Sections 1–4 and cannot explain why a company that bolts together an Nvidia-specified server rack earns a single-digit margin while Nvidia itself earns a 75% one, the primer has not done its job yet.
Bolting a GPU board into a chassis that Nvidia has already designed for you is a scaling problem: buy the rack, hire the technician, pass the burn-in test. Making sure that chip, and 71 others like it in the same rack, actually keep computing under 130 kilowatts of sustained load without throttling or failing — and being allowed to buy the chips in the first place — is a different kind of problem entirely, and it is the one almost no published statistic about "India's AI opportunity" actually measures. Every company report in this document is really answering one question: how much of this company's revenue sits on the scarce side of that line, versus the abundant 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 they disagree constantly — 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.
