Meta on Wednesday announced a landmark agreement with Reliance Industries to lease its first AI-enabled data center in India — a 168-megawatt facility that Reliance will build from the ground up at its industrial campus in Jamnagar, Gujarat. The deal closes the final gap in a three-layer India strategy that Meta has assembled since 2020: compute infrastructure to run its AI, a subsea cable to move the data, and a joint venture to sell Llama-powered AI products to Indian enterprises. For developers, businesses, and the hundreds of millions of Indians using Facebook, Instagram, and WhatsApp, the announcement means Meta's AI capabilities will be served from within the country rather than routed through data centers elsewhere.

Mark Zuckerberg framed the deal in terms of his company's broadest ambitions. "This world-class facility in Jamnagar will help us scale our AI infrastructure globally while deepening our long-term investment in India's economy," he said in a statement. Reliance Chairman Mukesh Ambani called it a "transformative moment for India's digital infrastructure," adding that it demonstrated India's readiness to compete at the frontier of global AI.

Why Reliance Builds and Meta Leases: Economics of 168 MW at Scale

The structure of the deal — Reliance builds, Meta leases — reflects a deliberate economic calculus. At 168 megawatts, a hyperscale data center requires an investment of roughly $2 billion to $4 billion in real estate and infrastructure alone before a single server is installed. By contracting a build-to-suit lease, Meta eliminates that capital outlay and shifts the construction risk entirely to Reliance, which already has the engineering expertise, land, and energy resources in Jamnagar to deliver the facility within two years.

Under a build-to-suit arrangement, the developer designs and constructs the facility to the tenant's exact technical specifications, then leases it back on a long-term basis. The tenant — in this case Meta — retains operational control of the hardware and workloads while the developer handles everything from design and construction to utility management, renewable power supply, network connectivity, and ongoing operations. Reliance said it would provide all of those services end-to-end. Meta, in turn, committed to covering the full cost of the energy and water the facility consumes.

This model has become the standard for hyperscale infrastructure deployments globally. The key advantage for Meta is not just cost savings but speed: rather than spending years standing up a real estate operation in India, it can have a dedicated facility operational within two years on a campus that Reliance is already developing into one of the largest data center complexes in the world, with a planned total capacity of 3 gigawatts.

Desalinated Seawater Cooling Replaces Freshwater in Water-Stressed Gujarat

Among the most technically distinctive elements of the Jamnagar facility is its cooling system. The campus will use desalinated seawater drawn from the Gulf of Kutch — the body of water that Jamnagar sits on — rather than freshwater, which is chronically scarce in Gujarat. Reverse osmosis desalination removes salts and minerals from seawater, producing water clean enough to run through a data center's cooling loops without corroding the infrastructure.

The choice is both operationally sensible and environmentally relevant. Hyperscale AI data centers are among the most water-intensive facilities ever built, consuming millions of gallons per day to prevent GPU clusters from overheating. In a water-stressed state like Gujarat, relying on freshwater at that scale would be environmentally indefensible and potentially impossible under local regulatory constraints. Jamnagar's coastal geography — the same geography that made it ideal for Reliance's oil refineries — makes seawater cooling a viable engineering solution.

The facility will run entirely on renewable energy. Meta separately contracted approximately 925 megawatts of new clean energy from two Indian providers: CleanMax will supply 837 megawatts of new solar and wind projects in Rajasthan and Karnataka, raising the cumulative capacity announced with that partner to over 900 megawatts; Fourth Partner Energy will provide 88 megawatts of solar and wind projects across Tamil Nadu, Karnataka, Maharashtra, and Uttar Pradesh. Together these agreements are roughly equivalent to two-thirds of India's entire installed data center capacity as of 2025, contracted in a single day.

How Does a 168 MW AI Data Center Work?

A hyperscale data center at 168 megawatts is, by the industry's own standards, a large but not extraordinary facility. The International Data Corporation defines a true hyperscaler as any facility using more than 100 megawatts of power and at least 5,000 servers. What distinguishes an AI-optimized campus from a general-purpose cloud data center is the density of specialized hardware: rows of GPU clusters — likely including Nvidia's current-generation accelerators — connected by optical fiber and designed to process the massive matrix calculations that train and serve large language models.

The architecture scales horizontally. Rather than adding more power to existing machines, hyperscale facilities add standardized nodes — each containing compute, memory, and networking — that can be brought online independently. Software defined at the infrastructure layer routes workloads automatically across the cluster, tolerating individual hardware failures without dropping jobs. At 168 megawatts, the Jamnagar facility would power roughly 1,000 to 2,000 high-density GPU racks depending on chip generation and thermal design, enough to run inference for millions of simultaneous user requests or to support large-scale model training.

Reliance's end-to-end service model — covering everything from power delivery to network connectivity — means Meta does not need to build separate procurement, operations, or facilities management teams in India. The facility plugs directly into Reliance's existing Jamnagar energy and logistics infrastructure, which already supports the world's largest oil refinery complex on the same site.

India's Data Center Market Draws $67.5 Billion from Hyperscalers

Meta's Jamnagar commitment lands in a market that has already absorbed the largest single-year surge of technology infrastructure investment in India's history. The country's installed data center capacity grew from about 375 megawatts in 2020 to around 1,500 megawatts by 2025, according to government data, a fourfold increase in five years. Industry estimates project that figure could exceed 7 to 8 gigawatts by the end of the decade.

The investment wave has arrived from multiple directions at once. Microsoft and Amazon together targeted $52.5 billion in Indian AI and digital infrastructure commitments; Google announced a $15 billion AI data hub investment in October; OpenAI signed an infrastructure partnership with Tata Group targeting 1 gigawatt of capacity; and Blackstone-backed AirTrunk committed $30 billion to build 5 gigawatts of Indian data center capacity by 2030. New Delhi has accelerated the inflow by offering zero-tax treatment through 2047 for foreign cloud providers on services sold overseas, provided those workloads run from Indian data centers.

The scale of investment reflects a structural shift, not a speculative cycle. India generates roughly 20 percent of global data but holds only 3 percent of global data center capacity, according to government analysis. AI workloads require processing close to the user to keep latency low enough for real-time applications. With over a billion internet users and mobile data consumption crossing 25 gigabytes per user per month, India represents a volume of inference demand that is simply not efficiently served from data centers in Europe or the United States.

Meta's Three-Layer India Stack: From Jio to Llama to Jamnagar Compute

Today's announcement completes an architecture that Meta has been building for six years. Each layer was added deliberately, and no single layer functions fully without the others.

In April 2020, Meta invested $5.7 billion in Jio Platforms, Reliance's telecom and digital services subsidiary. That investment bought Meta a distribution relationship with the company that operates India's largest mobile network — providing access to hundreds of millions of Indian users at the connectivity layer.

In August 2025, the two companies announced a $100 million joint venture to develop enterprise AI solutions built on Meta's open-source Llama models. The JV offered Indian businesses a full-stack platform for deploying AI across sales, marketing, information technology, customer service, and finance — a software and services layer sitting above the infrastructure. The cost efficiency was material: Llama's open-source architecture allows the JV to run high-performance AI at a fraction of the cost of proprietary alternatives, distributing that advantage through Reliance's existing relationships with thousands of Indian enterprises and small businesses.

What the JV lacked, until today, was its own physical compute infrastructure in India. Running Llama models at the scale of India's enterprise market — and simultaneously serving Facebook, Instagram, and WhatsApp to Indian users — requires a dedicated facility capable of handling inference and training workloads locally. The Jamnagar data center is that facility. The 2020 investment handled connectivity; the 2025 JV handled software and commerce; the 2026 Jamnagar agreement handles the compute layer that both of the previous layers run on.

Project Waterworth Provides the Subsea Backbone

The final component tying the stack together is connectivity at the intercontinental scale. Meta's Project Waterworth, announced in February 2025, is a multi-billion-dollar subsea cable project that will span more than 50,000 kilometers — longer than the Earth's circumference — connecting the United States, India, Brazil, South Africa, and other regions across five continents.

What distinguishes Project Waterworth technically is its fiber density. The cable will carry 24 fiber pairs, compared to the 8 to 16 pairs used in most other new submarine cable systems. Each additional fiber pair multiplies the data capacity the cable can carry, and 24 pairs represents roughly double the bandwidth of Google's recent Equiano and Dunant cables, which use 12 pairs each. Meta has built more than 20 subsea cables over the past decade, but Project Waterworth is its most ambitious, deploying first-of-its-kind routing at depths of up to 7,000 meters and enhanced burial techniques in shallow coastal zones to protect against ship anchors and other hazards.

Earlier reporting indicated that Project Waterworth may land at Jio data center facilities in India — meaning the Jamnagar campus could be directly connected to Meta's global fiber backbone. That would allow data processed in Jamnagar to travel on Meta's own cable to its data centers in the United States and elsewhere, rather than relying on third-party carrier networks. A dedicated high-capacity subsea cable, a purpose-built AI compute campus, and a controlled enterprise AI distribution platform in one of the world's largest markets describe a degree of infrastructure independence that no other social media company has assembled outside the United States.

India's DPDP Act Makes Local Data Processing a Competitive Advantage

The Jamnagar facility's strategic value is further amplified by India's evolving data protection regime. India's Digital Personal Data Protection Act of 2023 began its phased implementation in November 2025 and is scheduled to reach full compliance by May 2027. The Act requires organizations processing Indian users' personal data to meet consent, security, and accountability standards enforced by a newly established Data Protection Board of India, with penalties reaching 250 crore rupees for serious violations.

While the DPDP Act does not mandate that all data be stored domestically, it creates powerful compliance incentives for companies that can demonstrate local data processing. A dedicated data center in India allows Meta to process and serve Indian user data within the country, simplifying compliance with the Act's requirements and reducing the legal complexity of cross-border data transfers. For Indian enterprise customers of the Reliance-Meta JV, the same local infrastructure means their business data can stay within India's jurisdiction — a consideration that has become commercially significant as Indian regulators increase scrutiny of foreign platforms' data practices.

India's Supreme Court reinforced that context in February 2026, warning Meta directly that it "cannot play with the right of privacy of this country" in a case stemming from the Competition Commission of India's 2024 fine of ₹213 crore over WhatsApp's data-sharing practices. A domestic data center does not resolve those disputes — which concern what data is shared among Meta's platforms, not where it is stored — but it signals a long-term commitment to operating within India's legal and regulatory framework rather than from outside it.

Frequently Asked Questions

Why is Meta building a data center in India?

Meta's Jamnagar facility will let the company run AI inference and serve its products — Facebook, Instagram, WhatsApp — from within India rather than routing requests through overseas data centers. That reduces latency for Indian users, supports compliance with India's emerging Digital Personal Data Protection Act, and anchors the compute layer needed to run the Llama-powered enterprise AI joint venture Meta operates with Reliance.

What companies are building AI data centers in India?

India has attracted a wave of hyperscaler investment totaling more than $67.5 billion in announced commitments. Microsoft and Amazon together pledged $52.5 billion in AI and digital infrastructure; Google announced a $15 billion investment; OpenAI partnered with Tata Group for 1 gigawatt of capacity; and Blackstone-backed AirTrunk committed $30 billion to build 5 gigawatts by 2030. Meta's 168-megawatt Reliance deal is the latest addition to that list, with India's installed data center capacity projected to grow from 1.5 gigawatts in 2025 to more than 7 gigawatts by the end of the decade.

What is the build-to-suit data center model and why did Meta choose it?

A build-to-suit data center is a facility designed and constructed by a developer — here, Reliance — to the exact specifications of a single tenant, who then leases the finished building rather than owning it. Meta chose this model because it eliminates the $2 billion to $4 billion in upfront real estate capital that owning a 168-megawatt facility would require, shifts construction risk to Reliance, and leverages Reliance's existing engineering expertise and energy infrastructure at Jamnagar. The facility will be ready within two years, faster than Meta could build a comparable owned facility from scratch in a new geography.

How does Meta's India data center connect to its global infrastructure?

The Jamnagar facility is designed to work alongside Project Waterworth, Meta's planned 50,000-kilometer subsea cable that will connect the United States, India, Brazil, and South Africa using 24 fiber pairs — roughly double the bandwidth of most other new submarine cables. Earlier reports indicated that Project Waterworth may land at Jio data center facilities in India, which would provide the Jamnagar campus with a dedicated, high-capacity link to Meta's global network rather than relying on third-party carrier infrastructure. That combination of local compute and owned intercontinental fiber represents the infrastructure architecture of a company treating India as a permanent strategic node, not a remote market.