India AI Digest — Saturday, June 6, 2026

• AirTrunk, the Blackstone-backed Australian operator, said it will invest about $30 billion to build roughly 5GW of AI data-centre capacity across India by 2030 — a single-operator commitment of the same order as analyst projections for the entire Indian market, and so far mostly a 2030 pledge plus a Maharashtra land letter.
• Delhi national-security AI firm Innefu Labs raised a $30M Series B led by Singapore's Panthera Growth Partners, putting a profitable, government-deployed defence-AI builder on a pre-IPO growth track.

---

INFRA · COMPUTE · CAPITAL · June 5, 2026

AirTrunk commits $30B to build 5GW of India AI data-centre capacity by 2030

AirTrunk, the Australian hyperscale-data-centre operator Blackstone acquired in 2024, said on June 5, 2026 that it plans to invest about $30 billion to develop roughly 5GW of new AI and cloud data-centre capacity across India by 2030. The announcement followed a meeting between AirTrunk CEO Robin Khuda and Prime Minister Narendra Modi. Maharashtra exchanged a letter of intent for land at the Raigad Pen Growth Centre for a campus of about 3GW, which AirTrunk sizes at roughly ₹2 trillion (~$21 billion). That sits on top of an existing pipeline of about 600MW across Mumbai, Chennai, and Hyderabad, built after AirTrunk entered India via its acquisition of Lumina CloudInfra. The numbers are AirTrunk's own, framed around a state-government engagement; the ~600MW pipeline is the only part already on the ground.

What this means. The number to hold this against is the one Avendus Capital published two weeks ago. Its data-centre report projected India's total built capacity growing from 1.6GW to about 5GW by 2030, with roughly $23 billion of investment driving 650,000–700,000 GPU deployments — a sizing for the whole market (covered in the May 28 digest). AirTrunk's stated plan, for one operator, is ~5GW of new capacity and ~$30 billion. The single-operator commitment is of the same order as a credible analyst projection for the entire country's 2030 build — and on the headline figures, larger.

That gap is the tell. A forward commitment that exceeds a whole-market sizing is a commitment, not capacity. What exists today is the ~600MW pipeline and a state letter of intent for land at Raigad; the rest is a four-year plan announced alongside a PM meeting. Blackstone has the capital and AirTrunk has an operating India base, which puts this above a pure signing-ceremony announcement. But 5GW by 2030 is a number to track against energized megawatts, not to bank.

India angle. Hyperscale capacity sited in India is the physical substrate for data-residency-bound workloads. BFSI and government deployers that cannot route regulated data to US or Singapore cloud regions need in-country capacity to run AI at all; a 5GW build, if it lands, is the kind of supply that makes in-country inference and training a default rather than an exception. The second-order question is unit economics — whether domestic capacity at this scale shifts the cost of running Indian AI workloads against the current pattern of offshoring to foreign cloud regions.

Both reads depend on terms AirTrunk has not published: pricing, third-party access models, and whether this capacity serves a small set of anchor tenants or the broader market. Who gets GPU access on what terms is the part that determines what this means for Indian builders, and none of it was announced. Treat the access story as forthcoming, not current.

Behind the news. This lands in a run of large India data-centre commitments rather than starting one. Google broke ground on its gigawatt-scale Visakhapatnam AI hub — $15 billion over 2026–2030 with AdaniConneX and Nxtra — in late April (covered in the April 30 digest). Uber announced its first India data centre with the Adani Group on May 13 (covered in the May 15 digest). And the same backer behind AirTrunk turned up weeks earlier on the cloud layer: the Competition Commission cleared Blackstone's controlling-stake investment in Mumbai AI-cloud startup Neysa Networks, a $1.2 billion financing earmarked for a 20,000-GPU build (covered in the May 22 digest). AirTrunk is the largest single line in that sequence — and the one whose headline figure most outruns what is built.

What to watch. Construction start and first energized megawatts at the Raigad Pen campus, and net operational capacity added across the Mumbai–Chennai–Hyderabad pipeline beyond today's ~600MW by mid-2028. Those are the figures that convert a 2030 commitment into delivered capacity.

What this is not. Not 5GW of Indian AI capacity. It is a ~600MW operational base, a state letter of intent for land, and a four-year, ~$30 billion plan to build the rest.

See also: Avendus projects 700,000 GPUs and ~$23B India AI data-centre opportunity by 2030 — published/2026-05-28.md, CCI clears Blackstone's controlling stake in Neysa — published/2026-05-22.md, Uber to build first India data centre with Adani — published/2026-05-15.md.

Source: TechCrunch, June 5, 2026. → link Also: Bloomberg; Data Centre Magazine.

Confidence: High that the commitment was announced at this scale; the capacity itself, beyond the ~600MW pipeline, is a forward plan.

---

DEFENCE · FUNDING · SOVEREIGN AI · June 5, 2026

Innefu Labs raises $30M Series B for national-security AI, ahead of a planned IPO

Innefu Labs, a Delhi national-security AI firm founded in 2010 by Tarun Wig and Abhishek Sharma, raised $30 million (~₹286 crore) in a Series B led by Singapore growth-equity firm Panthera Growth Partners on June 5, 2026. The round mixes primary and secondary capital and funds sovereign-AI R&D, overseas expansion, agentic-AI and robotics work, and domain-specific LLMs for high-trust environments, ahead of a planned IPO. Innefu says its platforms power deployments including India's National Terrorism Data Fusion Centre, and it reported ₹103 crore in FY25 revenue and ₹34.2 crore in net profit. The financials and deployment claims are the company's own, carried through trade-press coverage of the round.

What this means. The signal is sectoral maturity, not the cheque size. A $30 million growth round is mid-sized; what makes it notable is the profile of the company taking it. Innefu reports revenue, profit, and a named production deployment in a government counter-terrorism system — a defence-AI firm raising growth capital pre-IPO, not a pre-product story raising on a thesis. India's defence and intelligence AI cohort has run mostly on pilots and government contracts; a profitable, production-deployed builder moving toward public markets is evidence the category is maturing into fundable businesses. The financials are self-reported and unaudited in this coverage, so the substance rests on Innefu's own disclosure plus the named deployment.

The structural tension worth naming is the capital source. A national-security firm building LLMs for high-trust environments — on-premise, in-country, controlled-data systems by definition — is taking its growth capital from a Singapore-based foreign investor. That is not a contradiction in itself; foreign growth equity into Indian deep-tech is the pattern the market has been trying to build. But sovereign-AI capability funded by foreign capital in a defence-intelligence vertical is the part that carries geopolitical weight, and it is worth watching how the IPO and the overseas-expansion plan handle it.

India angle. This is a defence and govtech read, narrow by design. The data-residency and sovereignty constraint is the whole product: domain-specific LLMs for intelligence and counter-terrorism work cannot run on cross-border commercial APIs, which is what makes an in-country builder with controlled-data deployments structurally defensible against frontier-model commoditization. For the Indian deep-tech capital ecosystem, the round is a company-side instance of a capital-supply build-out already visible on the fund side.

Behind the news. The fund-side capital for exactly this category formed in the prior weeks. Piper Serica launched its ₹800 crore Bharat Tech Fund and Shastra VC closed a $100 million third fund on May 20–21 — both naming defence-tech among their target sectors (both covered in the May 22 digest). Innefu's Series B is the company-side counterpart: capital flowing into a defence-AI builder at growth stage, against a backdrop of dedicated domestic vehicles raising to fund the same thesis.

What to watch. The IPO timeline and any draft filing, which would make Innefu one of the first Indian defence-AI firms to test public markets, and the first named market for its stated overseas expansion — the cleanest signal on whether a sovereign-AI positioning travels outside India.

Source: TNGlobal, June 5, 2026. → link Also: Business Today; CIOL.

Confidence: High on the round; medium on the revenue and profit figures, which are company-reported.

---

Position movements