Country profile · Rank 8 of 108 · Primed
India compute capacity
India runs about 1.9 GW of live datacenter capacity as of 7 September 2026, a Gross Domestic Compute of $95B. Its identified energy resource could host about 271 GW of AI compute, a ceiling worth $16.3 to 21.7T at today's price of a gigawatt, about 6 times its $3,916B GDP. The bankable slice today, discounted mainly for physical conditions and governance, is $1.7T, or $1,180 per person. India is primed, ranked 8 of 108 countries by unlockable value.
Key numbers
| Live datacenter capacity | 1.9 GW | IT capacity, mid-2026, compiled from industry and national sources. |
|---|---|---|
| Live compute (GDC) | $95B | Live GW × $50B, the central price of an AI factory. |
| Resource ceiling | 271 GW | Identified hydro, geothermal, solar, wind and surplus fossil resource. |
| Ceiling value | $16.3 to 21.7T | Ceiling GW × $60 to 80B per GW, NVIDIA's all-in figure. |
| Ceiling as a multiple of GDP | 6× | High end over nominal GDP of $3,916B (IMF). |
| Unlockable value | $1.7T | Firm untapped GW × $50B × readiness. The bankable slice. |
| Unlockable per person | $1,180 | IMF population, 2026. |
| Built | 20.3% | Hydro and geothermal already installed, over the ceiling. |
| Readiness | 66% | Eight-part discount. Binding: physical conditions and governance. |
| Conversion score | 59 | 0 to 100. +5 since 19 Aug. |
| Tier | Primed | Readiness 65% or better with a ceiling at least 3× GDP. Ready, with real headroom left to build. |
| Rank | 8 of 108 | By unlockable value. 1 of 7 in South Asia. |
Three numbers, one scale
Ceiling is what the endowment could host; unlockable is the slice underwriters would sign today; live compute is what runs. Drawn on one linear scale on purpose: the gap is the point.
Latest signal
Projects on the ground
| Project | Scale | Status | Date |
|---|---|---|---|
| IndiaAI Mission GPU commons | 17k+ GPUs installed; 100k+ target | Live | 2025-26 |
| Reliance Jamnagar AI campus (Meta JV 168 MW) | $110B plan; 120+ MW live H2 2026 | Building | 2026 |
| OpenAI for India (Tata) | 100 MW scaling to 1 GW | Announced | Feb 2026 |
South Asia: how India compares
| # | Country | Unlockable | Live compute | Readiness | Tier |
|---|---|---|---|---|---|
| 8 | India | $1.7T | $95B | 66% | Primed |
| 27 |  Pakistan | $587B | under $1B | 42% | Sleeping Giant |
| 30 |  Nepal | $535B | under $1B | 56% | Sleeping Giant |
| 46 |  Bhutan | $326B | under $1B | 66% | Primed |
| 61 |  Afghanistan | $222B | under $1B | 23% | Sleeping Giant |
| 87 |  Sri Lanka | $108B | $1.5B | 50% | Emerging Upside |
| 95 |  Bangladesh | $41B | $2.0B | 49% | Long Road |
What compute costs right now
The price of the silicon a country would host, from the tape. Labeled buy-now prints from named venues.
Questions people ask
How much data center capacity does India have?
India runs about 1.9 GW of live datacenter IT capacity as of 7 Sep 2026, compiled from SemiAnalysis, Cushman & Wakefield, Rystad, Knight Frank and national reports. Valued at $50B per gigawatt, that is a Gross Domestic Compute of $95B.
What is India's compute net worth?
The ceiling is $16.3 to 21.7T: 271 GW of identified hydro, geothermal, solar, wind and surplus fossil resource valued at $60 to 80B per gigawatt, NVIDIA's all-in figure for an AI factory. That is 6 times India's $3,916B GDP. The bankable slice today, after a readiness discount of 66%, is $1.7T.
What holds India back from converting its energy into compute?
Readiness scores 66% on an eight-part discount stack (governance, stability, GPU access, grid, fiber, momentum, physical conditions and capital access). For India the binding discounts are physical conditions and governance. It has built 20.3% of its ceiling as generating capacity, and its conversion score is 59 of 100 (+5 since 19 August).
How does India compare with its neighbours?
India ranks 8 of 108 countries by unlockable value, and 1 of 7 in South Asia. The regional leader is India at $1.7T unlockable and $95B of live compute.
Cite this page
Free with attribution for research, journalism and personal use. Method, sources and every parameter: how the numbers are made.