In June 2026, six weeks after entering India, AirTrunk announced a proposed investment of more than ₹3 lakh crore — about US$30 billion — to build over 5 GW of Indian data centre capacity by 2030. It is the largest commitment ever made to Indian digital infrastructure, and it is twice everything this registry counts in the entire country: 5,000 MW against 2,420.4 MW across 241 additive records. It is also, on its own arithmetic, an exceptionally useful number. Thirty billion dollars over five gigawatts is US$6 million, or ₹53 crore, per megawatt — which lands within 3% of the only lender-verified Indian construction cost in this file, NTT’s ₹51.4 crore per megawatt at Airoli. Two entirely independent sources, one a special purpose vehicle’s debt documentation and one a global chief executive’s government meeting, agreeing on what an Indian hyperscale megawatt costs. That is the most valuable thing in this record, and it is not the headline.
1At a glance
2US$6 million a megawatt
A programme announcement is normally useless for pricing. This one is not, because it carries two numbers that divide into each other.
| Measure | Value | Basis |
|---|---|---|
| Announced investment | US$30 billion | company announcement, June 2026 |
| Announced capacity | 5,000 MW | company announcement, June 2026 |
| Cost per MW | US$6.0 million | 30,000 divided by 5,000 |
| In rupees at ₹88.40 | ₹53.04 crore per MW | the registry’s fixed quarterly rate |
| Compare: NTT NAV2, Airoli | ₹51.4 crore per MW | ICRA rating report, October 2025 |
| Difference | 3% | two independent sources |
Set that against the other published costs in this file. NTT’s Bengaluru 4 campus works out at ₹35.7 crore per megawatt and its Kolkata campus at ₹80 crore. Both are real projects with published investments; neither is a market-wide benchmark.
The AirTrunk programme rate of ₹53 crore sits almost exactly on NTT’s Airoli rate of ₹51.4, and those two numbers come from completely different places. One is a credit rating agency reading a project company’s books. The other is a global operator telling a government what it intends to spend.
A new Indian hyperscale megawatt, land and finance included, costs about ₹53 crore.
Two independent sources agree to within 3%, which makes it the most defensible cost benchmark this registry has produced.
35 GW against a 2,420 MW registry
The scale of the announcement is best understood against what actually exists in this file.
| Record | Capacity | What it is | Kind |
|---|---|---|---|
| AirTrunk programme | 5,000 MW | a target for 2030 | announcement |
| This registry, counted | 2,420.4 MW | 241 additive records | buildings |
| AirTrunk’s India pipeline | 601 MW | five location cards | acquired pipeline |
| NTT, counted | 181 MW | 32 records | buildings |
| STT GDC, counted | 0 MW | portfolios do not add | portfolio headline |
5,000 MW is 2.07 times the entire counted Indian market in this registry. That does not mean the announcement is wrong — it means India has never built at that rate, and the market’s existing stock is much smaller than its press releases suggest.
For context, AirTrunk’s entire global portfolio at the Lumina acquisition was described as more than 3 GW of operating and planned capacity across 20 campuses in six markets. The India programme alone is 5 GW. Either the company expects to nearly triple its global footprint in one country, or the number is a ceiling rather than a plan.
The registry treats it as a ceiling. It is recorded, dated, sourced, and counted at zero — and it is flagged so that nobody adds it to the pipeline, the Lumina cards or any state total.
4The ₹3 lakh crore problem
One inconsistency is worth recording because it is the kind that propagates through secondary reporting.
| Figure | Source | Implied capacity cost |
|---|---|---|
| US$30 billion | Reuters, AirTrunk’s own release | ₹53.04 crore per MW |
| ₹3 lakh crore | Indian coverage of the same announcement | ₹60.0 crore per MW |
| US$30 billion at ₹88.40 | arithmetic | ₹2.65 lakh crore |
| Gap | ₹3 lakh crore is about US$33.9 billion | 13% apart |
Thirty billion dollars is ₹2.65 lakh crore at the registry’s fixed rate, not ₹3 lakh crore. The difference is about US$3.9 billion, or ₹34,800 crore.
This is not a scandal and it does not change the analysis. It is the ordinary result of a round number being restated in a second currency, and it is recorded here because every downstream analyst will pick one of the two and never notice they differ.
The registry uses the dollar figure for the per-megawatt calculation, because the announcement was made in dollars and the rupee figure has clearly been rounded upward.
5Who is doing this
The corporate structure behind the announcement is the reason it is credible, and it is unusual enough to be worth setting out.
| Entity | Detail |
|---|---|
| AirTrunk | Founded 2018. Acquired by Blackstone and CPPIB in December 2024 for an enterprise value of about A$24 billion / US$16.1 billion — then the largest data centre transaction in history. |
| Lumina CloudInfra | Launched by Blackstone in November 2022 as a standalone India hyperscale platform, with plans for 600 MW. |
| The India acquisition | AirTrunk acquired Lumina in April 2026. Both companies were already Blackstone portfolio entities, so the transaction consolidates one owner’s Indian position under its operating platform rather than introducing new capital. |
| Backers | Blackstone and the Canada Pension Plan Investment Board, with founder Robin Khuda retaining a minority stake. |
| Markets after the deal | Australia, Singapore, Japan, Malaysia, Hong Kong and India — six markets, more than 3 GW operating and planned across 20 campuses. |
The significance for the Indian market is that the $30 billion is not a startup’s ambition. It is the largest owner of data centres in the world consolidating an Indian position it has held since 2022 and now scaling it through an operator with 20 campuses of experience.
For anyone selling into this market, that is both an opportunity and a warning. AirTrunk’s entry changes who the biggest counted operator in this registry is, and it does so without a single facility that anyone can visit.
The founder’s line at the announcement is worth keeping because it is the strategy in one sentence: capital is mobile, and India is creating the conditions for it to thrive.
6Procurement
7Limits and open questions
- This is a programme target, not capacity. It counts zero in this registry and must never be added to the 600 MW pipeline, the Lumina location cards or any state total.
- The ₹3 lakh crore and US$30 billion figures differ by 13% at the registry’s fixed exchange rate.
- No site, state allocation, phasing or power approval is published at programme level.
- By 2030 is a horizon, not a schedule. The registry records it as a horizon.
- The 3 GW global portfolio figure comes from acquisition coverage, not from an audited count, and is operating plus planned rather than operating.
- AirTrunk’s Indian operating capacity is not established anywhere in the public record reviewed here.
- What would close it. A site list with acreage, substation status and capacity per site. Until that exists, this record is a statement of intent.
8Sources
Sources: AirTrunk announcement of the India investment programme, June 2026; Reuters reporting of 1 June 2026; Forbes and Moneycontrol coverage of the announcement and the government engagement programme; The Hindu BusinessLine and Times of India on the ₹3 lakh crore commitment; AirTrunk’s own announcement of the Lumina CloudInfra acquisition of April 2026; and Data Center Dynamics and Mingtiandi coverage of the acquisition.