Meta has partnered with Reliance Industries to build a 168 megawatt AI data center in Jamnagar, Gujarat. The site will go live within two years, powered by renewable energy and cooled with desalinated seawater. Meta Reliance marks Meta’s first major AI infrastructure footprint in India.
Key Takeaways
- 168 megawatt AI data center in Jamnagar, operational within two years
- Meta covers all its energy and water costs, Reliance runs the infrastructure end to end
- India is moving from 375 MW in 2020 to over 8 GW of data center capacity by decade’s end
A partnership that extends five years of Meta Reliance ties
The announcement formalizes a joint venture between Meta and Reliance Industries to build a 168 megawatt AI data center on the Jamnagar industrial site in Gujarat. The timeline is aggressive. The facility is promised to be operational within two years.
Reliance will handle the entire build. The Indian conglomerate takes on design, construction, renewable power supply, connectivity, and day-to-day operations. It is a fully integrated end-to-end role that positions Reliance as the reference infrastructure partner for foreign hyperscalers in the country.
The Meta Reliance alliance is not a first encounter. Meta already invested 5.7 billion dollars in Jio Platforms back in 2020, taking a strategic stake in Mukesh Ambani’s telecom ecosystem. Five years later, the relationship moves from the distribution layer down to raw infrastructure. For Meta, this is one more front in an offensive AI strategy, going as far as blocking Apple Intelligence in its iOS apps.
The financial value of the deal was not disclosed. The specific AI workloads that will run on the site remain unspecified, and Meta has not confirmed whether additional Indian campuses are planned beyond this first one.
Renewable power and desalinated seawater to cool the GPUs
Choosing Jamnagar is not random. The site sits on the coast, which lets Meta Reliance feed its cooling system with desalinated seawater instead of freshwater. In a country under chronic water stress, that is a political argument as much as a technical one.
The power supply will be fully renewable. Meta commits to covering all its energy and water costs across the perimeter of its own operations. The approach contrasts with the dominant U.S. model, where hyperscalers are reactivating gas or nuclear capacity to absorb soaring AI demand.
In parallel, Meta has secured close to 1 gigawatt of renewable capacity in India through separate agreements with CleanMax and Fourth Partner Energy. The company is therefore building a dedicated energy mix, partly insulated from the Indian grid, to absorb the predictable consumption of its future GPU clusters.
This hybrid design (renewable plus desalination) also serves a reputational purpose. Landing a massive data center in South Asia without worsening pressure on water and grid spares Meta the political backlash now slowing down several of its U.S. projects.
Also on Horizon:
- China cuts 12,000 university degrees made obsolete by AI
- 79% of parents already use AI to raise their kids
- GPT-5.5 Instant beats doctors on health answers
India becomes the central stage of the AI data center race
India’s data center capacity is on an exponential curve. It rose from 375 megawatts in 2020 to 1.5 gigawatts in 2025, and projections exceed 8 gigawatts by the end of the decade. That trajectory turns India into one of the three or four global markets that really matter for AI compute. On the end-user side, ChatGPT Images 2.0 exploded in India from launch.
The fiscal lever is massive. New Delhi grants tax exemptions running through 2047 for foreign cloud providers that execute international workloads from Indian facilities. It is a rare and aggressive setup, designed to vacuum global AI capex into the country.
Meta Reliance is therefore not landing on empty ground. Microsoft, Amazon, Google, OpenAI and Uber have each announced significant Indian investments recently. Among independents, Blackstone-backed AirTrunk plans to deploy 30 billion dollars for 5 gigawatts of capacity by 2030.
In the short term, the auction is fought on renewable power purchase agreements and access to coastal land with grid connection. Players able to bundle infrastructure and energy in a single deal (the Reliance model) take an immediate edge over pure-play hyperscalers.
In the medium term, the risk is polarization. One export-focused Indian hub on one side, a domestic market fenced off by sovereignty rules on the other. The real ability of Meta Reliance to serve local AI needs (languages, content, agents) will be the key indicator to watch over the next six months.
Follow the story on Horizon.
