At a glance
India runs roughly 1.3GW to 1.5GW of operational data centre capacity, close to Australia’s 1,350MW, but growing far faster off a lower base.
Forecasts put India at 4GW to as much as 8GW by 2030, and the market itself on a path to about US$22 billion.
Mumbai holds close to half of national capacity; Chennai, Hyderabad and Bengaluru follow, anchored by subsea cable landings.
Hyperscalers and Indian conglomerates have pledged more than US$100 billion, led by Reliance and Adani, with Microsoft, Google, AWS and OpenAI committing separately.
Australian capital is already inside the build, through AirTrunk’s US$30 billion program, Sharon AI’s compute exports, Firmus hardware and AustralianSuper via NIIF.
India matches Australia’s capacity, and is growing far faster
When AirTrunk’s Robin Khuda used India PM Narendra Modi’s Melbourne visit to press Australian super funds on India, he was pointing at a market few of them track closely. This spotlight sets that market out.
India runs roughly 1.3GW to 1.5GW of operational data centre capacity as of mid-2026, close to Australia’s 1,350MW of live capacity but on a population fifty times larger and an economy compounding at about 6% a year. Bernstein put it at “about 1.5GW today” in June 2026. The capital is arriving faster than the grid can carry it.
Consensus estimates put installed capacity near 1.7GW to 2.0GW by the end of 2026, roughly 30% growth in a single year, and forecasts to 2030 run from a 4GW to 5GW base case to as much as 8GW to 9GW in the AI-accelerated cases carried by Bernstein and the Indian government’s own 5GW to 8GW target band.
The dollar figures need care, because three different numbers get conflated. The market’s revenue value is tipped to reach about US$22 billion by 2030, from roughly US$10 billion, on the Indian Brand Equity Foundation’s estimate. Genuinely committed capital since 2020 sits nearer US$15 billion. The eye-catching “US$100 billion-plus” and “US$180 billion” tallies are announced multi-year intent, dominated by two conglomerate pledges, and should not be read as deployed money.
Mumbai is the centre of gravity, anchored by the subsea cables
Mumbai holds close to 49% of national capacity and is set to take about 47% of all additions over the next five years, because it is the country’s financial capital and its principal cable landing point. Chennai is the second hub, with roughly 268MW and about 153MW under construction, and three new subsea systems landing across 2026 and 2027. Hyderabad, at about 138MW, and Bengaluru, near 119MW, are growing on hyperscaler self-builds, land availability and aggressive state incentives.
The connectivity is the reason. Four major submarine systems are commissioning at Indian landing stations: the Reliance-backed India-Asia-Xpress and India-Europe-Xpress, the 21,700km SEA-ME-WE 6, the Meta-anchored 2Africa at Mumbai, and Google’s Blue-Raman, which uses Sify as its Mumbai landing partner. That density of routes into Europe, the Gulf, Africa and Southeast Asia is what makes Mumbai and Chennai the natural homes for capacity, the same cable-and-land logic that decides where facilities land in Australia’s own market.
Reliance and Adani lead the pledges
India’s two largest industrial houses have pledged roughly US$100 billion each toward AI and data centres, dwarfing the pure-play operators.
Operator | Backing | Scale and pipeline |
Reliance / Jio (Digital Connexion) | Reliance, Brookfield, Digital Realty JV | ~US$11bn, 1GW campus in Andhra Pradesh; Reliance flagged a ~US$110bn AI plan |
AdaniConneX | Adani Group + EdgeConneX (50:50) | Targeting ~1GW; part of Adani’s ~US$100bn AI and data centre pledge |
STT GDC India | ST Telemedia (Singapore) | Largest by installed IT load, ~318MW; ~US$3.2bn to add 550MW |
Nxtra by Airtel | Bharti Airtel | ~400MW target by 2027, ~US$586m; renewable-powered by 2030 |
CtrlS, Yotta, Sify, NTT India | Domestic and Japanese | Established colocation across Mumbai, Chennai, Noida, GIFT City |
Source: Company disclosures and trade coverage, July 2026. Market-share rankings vary by source.
The conglomerate money changes the risk profile of the whole market, because Reliance and Adani can finance their own power and land in a way a pure-play operator cannot. Tata has entered too, launching a hyperscale arm, HyperVault, with OpenAI as its first customer for 100MW of AI capacity. The read-through matters for Australia: this is the same pattern of capitalised balance sheets underwriting the AI build that we traced in the regional JLL Asia-Pacific supercycle worth US$772 billion, and it is why CBRE promoted India into its top “Leading” tier of Asia-Pacific data centre markets alongside Australia in 2026.
Hyperscalers and AI labs have pledged more than US$40 billion
The foreign platforms announced most of their India commitments across late 2025 and early 2026, led by Microsoft’s US$17.5 billion and Google’s US$15 billion.
Platform | Commitment | Detail |
Amazon / AWS | Additional AI and cloud infrastructure, reported at US$8bn to US$13bn | Part of Amazon's larger multi-year India spend |
Microsoft | US$17.5bn | AI infrastructure, deploying to about 2029 |
Google Cloud | US$15bn | New AI hub and data centre capacity at Visakhapatnam, Andhra Pradesh |
OpenAI + Tata | 100MW, eyeing 1GW | First customer of Tata’s HyperVault, part of Stargate |
Oracle | Live regions | Mumbai and Hyderabad, both government-certified |
Source: Company releases and trade coverage, July 2026. Figures are announced multi-year pledges.
These are commitments, not switched-on megawatts, and the AWS line shows why the distinction matters: Amazon’s headline India number was revised between December 2025 and June 2026, so the AI-and-cloud-specific US$13 billion is the more useful figure than the top-line total. India has also dangled an unusual lure, offering foreign cloud providers zero tax through 2047 on services sold overseas if the workloads run from Indian data centres.
The demand is proven; the grid is the constraint
India is a data-localisation market. The Digital Personal Data Protection Act 2023, whose rules were notified in November 2025, plus the Reserve Bank of India’s payments-data localisation, keep a growing share of data onshore. Layer on 5G, the expansion of Global Capability Centres, and the government’s own IndiaAI Mission, which has crossed 34,000 subsidised GPUs and targets 100,000 by the end of 2026, and the domestic demand base is real rather than speculative. On policy, the draft National Data Centre Policy 2025 offers up to twenty years of conditional tax exemption, full electricity-duty relief and single-window clearance, and states from Maharashtra to Tamil Nadu, Telangana and Uttar Pradesh compete with their own incentive schemes.
The binding constraint is transmission. India’s grid is still about 68% coal, with a grid carbon intensity near 670 grams of CO2 per kilowatt-hour against Australia’s roughly 525, so the sustainability story is a real weak point. More pressing for operators, the wires lag the megawatts: roughly 300 gigawatt-hours of clean power was curtailed in early 2026 for lack of transmission, renewable projects take eighteen to twenty-four months to build while the lines to move that power take three to five years, and tens of gigawatts of auctioned renewable capacity sit without signed offtake. The AI training workload that fills these campuses is not latency-sensitive, so it can locate wherever power and land are cheapest, the logic set out in our explainer on what an AI data centre is. India has the cheapest land and labour of the major markets. What it still has to build is the power path to the door.
Australian capital is already inside the Indian build
Australian operators and capital already reach India in several directions.
Australian party | How it reaches India | Detail |
AirTrunk | Building in India | US$30bn / 5GW-plus by 2030; Lumina CloudInfra acquired 2026 (~600MW across Mumbai, Chennai, Hyderabad); 3GW Raigad hub LOI with the Government of Maharashtra, reported near US$21bn |
Sharon AI | Selling compute into India | US$1.25bn agreement with Indian provider ESDS, contracted to run on about 8,200 NVIDIA B300 GPUs served from an Australian data centre, by Sharon AI’s account |
Firmus (Sustainable Metal Cloud) | Hardware deployed in India | HyperCube immersion units inside ST Telemedia facilities, part of a roll-out across Singapore, India and Australia |
AustralianSuper | Institutional capital | A further A$500m into India’s NIIF in July 2026, lifting total India exposure to about A$3.3bn |
Source: Company disclosures and trade coverage, July 2026.
AirTrunk is the clearest case of Australian-built capital turning acquirer, a platform founded in Sydney and sold to a Blackstone-led consortium for A$24 billion that is now one of the largest single builders of Indian digital infrastructure, as we covered when it became the only Australian operator in the 2026 data centre M&A top 10. Sharon AI runs the other way, keeping the GPUs, the data and the certification in Australia while serving an Indian customer, a model we set out in its sovereign AI factory buildout. Firmus places its immersion-cooling hardware inside Indian facilities through its ST Telemedia partnership. Even the Canadian manager Brookfield, active across Australian infrastructure, has named India data centres a “big opportunity” and sits inside the Reliance Digital Connexion venture. The capital flows both ways across this corridor, and Australian names are on both sides of it.
India and Australia, side by side
India | Australia | |
Operational capacity | ~1.3GW to 1.5GW | ~1,350MW live |
2030 outlook | 4GW to 8GW | 3,100MW+ |
Leading hubs | Mumbai (~49%), Chennai, Hyderabad, Bengaluru | Sydney, Melbourne |
Hyperscale regions | AWS, Microsoft, Google, Oracle | AWS, Microsoft, Google, Oracle |
Grid mix | ~68% coal; ~670g CO2/kWh | ~43% coal; ~525g CO2/kWh |
Governance | DPDP Act, RBI localisation, tax holiday to 2047, IndiaAI sovereign compute | Hosting Certification Framework; no mandatory localisation |
Core constraint | Transmission build-out | Grid connection and generation |
Source: IBEF, CBRE, IEEFA and primary disclosures, July 2026.
The two markets are not really competing for the same megawatt. Australia offers a cleaner grid, regulatory predictability and sovereign certification, and it is a mature market doubling toward 3,100MW. India offers scale, cost and speed off a lower base, with a grid it is racing to firm. Where the next AI dollar lands depends on the workload: training that chases the cheapest power and land tilts toward India, while latency-sensitive and sovereign workloads still favour a certified, contracted Australian megawatt.
What to watch
India’s build turns on power and policy. The demand is already proven.
Catalyst | Why it decides the outcome |
AirTrunk Raigad financial close | The 3GW Maharashtra hub is at letter-of-intent stage; financial close turns intent into a project |
Transmission build-out | The wires lag the megawatts; RE curtailment and unsigned PPAs are the real brake |
Reliance and Adani execution | The US$100bn-plus headline rests on two conglomerate programs proceeding on schedule |
DPDP Rules compliance timeline | Consent-manager and breach-notification obligations phase in through 2026 and 2027, deepening localisation demand |
Australian super fund allocations | Whether AustralianSuper’s move is followed tests how much Australian institutional money backs the corridor |
India is absorbing more capital than its grid can currently power. Australia offers the cleaner grid, the slower build and regulatory predictability, while its flagship operator now spends its largest new commitments offshore. The question for Australian operators and investors is not whether India matters, it is which side of the build they want to be on. The capital is already moving. Australian names are already on both sides of it.