Meta picks Reliance to anchor its first India AI data centre, betting on a market the hyperscalers have circled for years
On 10 June 2026, Meta confirmed what Indian telecom and retail executives had been told in private for months: the company will build its first AI data centre on Indian soil, and it will do so inside a joint framework with Mukesh Ambani's Reliance Industries, anchored by a 168-megawatt facility that both sides describe as expandable. The announcement, carried first by Reuters at 07:15 UTC and elaborated within the hour by The Indian Express, lands as the most concrete step yet by a US hyperscaler to move serious AI compute into a country that, until now, has hosted their users but not their silicon. The deal is modest in capacity terms — a single mid-sized campus rather than a hyperscale region — but the signal value is large. Meta is telling New Delhi, and the rest of the industry, that the era of treating India as a downstream market for foreign-built models is ending.
The bet, in plain terms, is that India's AI demand will be served increasingly by compute that sits inside its borders, under its data-protection regime, and on power grids and fibre rings that Indian conglomerates are best placed to wire up. Reliance brings land, power-purchase agreements, and the Jio customer base that the partners will need to make a campus of this size pay back. Meta brings the capital, the model portfolio, and a regulatory permission slip that few non-Indian firms currently hold in full. Both sides are buying optionality. The question is who ends up holding the leverage when the next round of expansion is signed.
The deal as announced
The structure, as described in the 10 June coverage, is a partnership rather than a build-and-sell arrangement. Reliance will provide the real estate, the power, and the network reach; Meta will deploy the compute and the software stack. The 168-megawatt figure, flagged by TechCrunch in its 07:05 UTC write-up, refers to the initial campus and is described as expandable. The Indian Express frame puts the deal inside a broader Indian government narrative about hyperscalers localising compute, and notes that the facility will support Meta's global AI workloads — not only traffic from Indian users. That last clause matters: it implies Meta is reserving the option to use Indian capacity for non-Indian inference and training, a hedge against capacity squeezes in its northern Virginia and Oregon regions.
The most consequential line in the Indian Express write-up is the one that the business wires are likely to under-weight: this is, in their phrasing, an AI-powered data centre built via the Reliance partnership, not merely a Meta campus on Reliance land. The distinction is not semantic. A traditional build-to-suit would leave Reliance as a landlord. The partnership framing leaves room for revenue-sharing, model fine-tuning on Indian datasets under Indian compliance, and a long-tail arrangement in which Reliance's enterprise salesforce — embedded across the country's largest retail and small-business network — becomes a distribution channel for Meta's APIs. Neither side has disclosed commercial terms; the public statements are deliberately short on numbers and long on verbs.
The timing is not accidental. New Delhi has spent eighteen months signalling, through a combination of the Digital Personal Data Protection Act, the IndiaAI Mission, and a quieter stream of state-level subsidy packages, that it expects the next wave of AI infrastructure to land on Indian terms. Meta's announcement is the first unambiguous US response.
Why Meta, why now
Meta's relationship with the Indian market has been unusually complicated for a company of its scale. WhatsApp, the platform on which Meta's India presence rests, has more than 500 million users in the country but has spent most of its history as a free-riding messaging layer that did not, until recently, generate proportionate revenue. The 2021 privacy-policy fight, the long standoff with the Reserve Bank of India over payments-data localisation, and the recurring complaints from the Ministry of Electronics and Information Technology about content-moderation standards have together kept Meta on the defensive in Delhi.
Compute is the cleanest answer to most of those headaches. A campus on Indian soil gives Meta a domestic legal entity against which the Indian state can enforce local compliance, a physical answer to data-localisation demands, and a hiring story — construction jobs, operations engineers, site reliability staff — that buys political goodwill that WhatsApp's user-growth charts never quite delivered. The fact that the partner is Reliance, rather than a US co-investor or a pure-play Indian data-centre operator, says more about the politics of power and land in India than any of the press releases admit. Hyperscale campuses in Gujarat need power purchase agreements that the state electricity regulator, the central grid operator, and a state government with its own industrial-policy ambitions must all sign off on. No Indian firm navigates that trilemma as fluently as Reliance.
The other reading, harder to evidence but plausible, is that Meta is hedging against a future in which US-headquartered AI services face regulatory or geopolitical friction in India. If Washington and Delhi ever diverge on technology controls — a scenario that the current administration's export-control posture makes less hypothetical than it was two years ago — Meta's Indian compute footprint gives the company a domestic story to tell. That is the kind of optionality that does not appear on a slide, but is implicit in any 168-megawatt commitment made in a market where Meta has historically been a free-tier consumer play.
Reliance's strategic position
For Reliance, the partnership is the second of three signals it has sent in the past year about where it intends the next leg of growth to come from. The first was the Jio-AirFiber fixed-wireless push. The third, still unfolding, is the gradual build-out of an enterprise services business modelled on the kind of platform play that Reliance's telecom arm has long threatened to become. AI infrastructure is the missing layer. Indian enterprises that want to fine-tune large models on proprietary data have, until now, had to send that data to facilities in Singapore, the US, or the Gulf. A Meta-branded campus in Gujarat, even one governed by a partnership rather than a Reliance-only build, makes that loop shorter and politically easier to defend.
There is a counter-narrative worth holding up against the corporate communiqués. The 168-megawatt figure is genuinely small by hyperscaler standards. A single modern US hyperscale campus will draw between 300 and 1,000 megawatts once fully built, and the largest training clusters in operation today consume far more. Meta's first Indian campus will not, on its own, change the balance of compute globally. Indian AI researchers will still rely on overseas training capacity for the largest models. The data-localisation story is real, but it applies to inference and to fine-tuning workloads much more than to frontier-scale training. Readers who treat this announcement as proof that India has joined the front-rank of AI infrastructure are reading further into the press releases than the engineering warrants.
A more honest read is that Meta and Reliance are laying the first course of a structure that neither side has finished drawing. The 168 megawatts are a foundation; the building on top of it is the part both companies are still negotiating, and the part that will determine who ends up setting prices for Indian AI capacity in the back half of the decade.
What the rest of the field does next
The announcement lands in a market that is no longer empty. Google has had Indian cloud regions in production for years. Microsoft has spent 2024 and 2025 expanding its Pune and Hyderabad capacity under its own brand. Amazon Web Services runs Mumbai and Hyderabad regions, with a long-promised third site in the works. NVIDIA has signed multiple Indian partnerships over the past twelve months that effectively give Indian firms access to its top-end accelerators through hosted arrangements.
The Meta–Reliance structure is distinctive in two ways. First, it is a social-media company buying compute into a country where its user base is huge but its direct enterprise presence has been thin. The other hyperscalers already had enterprise distribution in India; Meta is, in effect, building the channel now. Second, the Reliance partnership sets a template that competitors will have to answer — a US model-owner paired with an Indian conglomerate that controls land, power, and last-mile distribution. Microsoft could replicate the pattern with Tata or with Adani. Google could try a similar structure with Bharti Airtel, though the fit is less natural. AWS would face the same optionality question Meta has answered: partner, build solo, or stay sub-scale.
For Indian policy makers, the deal is a vindication of the localisation push, but also a warning. The 168-megawatt campus is a single site in Gujarat. India has signed off on a target of multiple gigawatts of data-centre capacity by 2030, much of it under various state-level industrial-policy frameworks. Whether the Meta–Reliance structure becomes the template — US model-owner plus Indian conglomerate — or whether one side or the other ends up dominating will shape how that capacity is priced, who can access it, and how much of the value-add stays inside India rather than repatriated through licensing fees.
Stakes and the next twelve months
The proximate stakes are commercial. Meta gains a domestic presence it can show to the Indian state and a partnership that gives it room to grow. Reliance gains a marquee US partner at exactly the moment its own AI ambitions need a model-owner of Meta's scale. The secondary stakes are regulatory. The Indian government will be watching whether the partnership produces the kind of local hiring, local supplier participation, and domestic intellectual-property development that its industrial-policy frameworks ask for. If it does, the template will be repeated. If it does not, expect the next round of approvals for foreign AI capacity to come with strings attached.
The third set of stakes, less visible but more durable, is geopolitical. India is the only large democratic market in which the United States, through its hyperscalers, retains a credible claim to the AI stack. China has built its own. The European Union is regulating more than building. The Gulf states are buying in. If India ends up running on a US-model-owner-plus-Indian-conglomerate template, the geopolitical map of AI infrastructure for the next decade starts to look a little less like a US-China duopoly and a little more like a tripolar arrangement in which the third pole is India-shaped.
What remains genuinely uncertain is the price of compute inside that arrangement. The announcements are silent on what the Meta–Reliance campus will charge for inference, for fine-tuning, for access to fine-tuned weights, and for the data-localisation premium that Indian enterprises will be obliged to pay. Until those numbers are public, the deal is a frame as much as a facility — a story the two companies are telling the Indian state, the US government, and each other. The facility, when it is built, will tell the rest of the story.
This publication framed the deal as a structural shift in how US hyperscalers enter the Indian market — through partnerships with Indian conglomerates that control land and power — rather than as a one-off capacity announcement. The wires led with the megawatt figure; the more durable line is the partnership shape.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- http://reut.rs/4xkvcxU