This is the section where the two halves of this report's own research diverge most clearly, and it is worth reading as a genuine structural finding rather than an artefact of combining two documents. The compute layer has one of the most concentrated anchor structures Dart Consultants has documented in any sector; the physical layer, sitting directly beneath it in the same value chain, has one of the most diffuse. Both matter; neither substitutes for the other, and conflating them is the single easiest way to misjudge either half's investment case.
Nvidia's data-centre segment revenue reached $75.2bn in Q1 FY27 (quarter ended 26 April 2026), up 92% year-on-year — the clearest evidence that chip demand remains the binding constraint on the entire compute layer, not manufacturing capacity anywhere in that chain. Nvidia is simultaneously reported to be narrowing, not widening, the field of manufacturing partners certified to build its newest rack-scale systems. Alongside it, the Government of India runs three concurrent, largely independent instruments: India Semiconductor Mission 2.0 pays for packaging/fab capability, not demand; the IndiaAI Mission pays for compute access, not India's own manufacturing capability; IT Hardware PLI 2.0 pays for assembly output, not R&D or IP. No single instrument, or simple sum of the three, tells a reader how much real value India's compute-layer companies are capturing versus passing through.
This report's own research (§3) found no equivalent single company or institution rationing access to transformer or switchgear manufacturing capacity. Instead: a genuinely diffuse, global capital-equipment scarcity (multi-year lead times reported in several markets) layers on top of state electricity boards, the Central Electricity Authority's grid codes, and a patchwork of state-level data-centre incentive policies — with no single dial a reader can watch the way IndiaAI Mission's 100,000-GPU target functions on the compute side. This report's own company research found a genuine, important caveat here too: at least one India-entity management team explicitly described its own domestic customer procurement as "normal," not scarcity-priced, directly contradicting the dramatic global lead-time narrative at the India-specific level — a distinction this report's individual company files check one by one rather than assuming by category.
A concentrated gatekeeper (Nvidia) can reprice an entire layer overnight with one policy change — which makes the compute layer's scarcity easier to see, easier to point to, and easier for a single company report to over- or under-state. A diffuse constraint (global equipment-manufacturing capacity) moves more slowly in both directions — harder to break into, but also harder for any one company's marketing to claim credit for. Every rating in §11 for a physical-layer company was checked against confirmed, company- specific evidence for exactly this reason: this report does not assume category membership (e.g. "power equipment") implies pricing power the way it might more safely assume for a confirmed NVQual-certified compute-layer partner.
