AI & Data Centres

This industry exists because AI models, cloud computing and digital services all require physical infrastructure somewhere, servers, GPUs, cooling systems and the power to run them, and India's combination of a huge digital consumer base, a new data protection law nudging data storage inward, and genuine government ambition to build sovereign AI capability have together made building that infrastructure domestically a national priority rather than something left entirely to foreign cloud providers.

The reason this page bundles data centres and AI together rather than treating them separately is that they are, in practice, the same investment story viewed from two angles, data centres are the physical infrastructure, AI workloads are increasingly the demand driving that infrastructure's fastest growth, and government policy, the IndiaAI Mission chief among it, treats them as a single connected priority.

AI and Data Centres is India's newest large-scale industrial story, barely a distinct investment category five years ago, now attracting hundreds of billions of dollars in committed capital from Reliance, Adani, NTT and global cloud players, alongside a dedicated government programme, the IndiaAI Mission, specifically built to democratise access to the compute power AI development requires.

Unlike most industries covered on this site, this one is defined less by a single regulatory Act and more by a convergence of forces, hyperscale private investment, state-level land and tax competition, a new personal data protection law nudging companies toward local data storage, and a genuine national push to build sovereign, India-language AI models rather than relying entirely on foreign systems.

Nearly every headline about India's AI ambitions ultimately traces back to a single physical constraint: how much compute capacity actually exists, and how much power is available to run it. This is why the IndiaAI Mission's core achievement isn't a model or an app, it's aggregating and subsidising access to tens of thousands of GPUs, and why every major hyperscale data centre announcement, Reliance's $110 billion commitment, Adani's Google partnership, is functionally a bet that India's AI compute bottleneck is worth resolving at massive capital scale.

This is also why power keeps showing up throughout this industry's story, state incentive packages built around renewable energy requirements, data centres measured in megawatts rather than square feet, and long-range capacity forecasts ranging from 4 to 9+ GW by 2030 depending entirely on how aggressively AI workloads actually materialise, all reflect the same underlying truth, in this industry, the physical constraint of power and compute, not software cleverness, is what ultimately caps growth.

This industry's 'trade' story runs mostly in the direction of capital and technology inflow rather than goods export, hundreds of billions of dollars in committed investment from global players like Google (via Adani) and NTT, alongside domestic conglomerate capital from Reliance and Adani themselves, all flowing into Indian data centre construction.

The Semiconductors industry covered elsewhere on this site captures the chip-manufacturing side of this story; AI & Data Centres instead captures the demand and deployment side, India remains a major importer of the actual GPUs and specialised chips these data centres run on, even as it builds out the surrounding infrastructure domestically.

Data centre operators earn primarily on colocation and cloud service fees, renting out rack space, power and connectivity to enterprise and AI customers, with hyperscale facilities increasingly signing long-term capacity commitments directly with major AI labs and cloud providers rather than relying purely on spot demand.

GPU-as-a-Service providers, both government-subsidised and private, earn on compute rental by the hour, a model that scales revenue directly with actual AI development activity rather than requiring customers to make large upfront hardware purchases, the same underlying economics that made cloud computing itself such a durable business model.

Electricity is the single largest ongoing operating cost for any data centre, which is exactly why state incentive packages, Maharashtra's 60% electricity duty exemption, Tamil Nadu's five-year power tax subsidy, target this cost specifically rather than land or construction costs alone.

GPU and specialised chip costs dominate the capital expenditure side, Nvidia's latest Blackwell Ultra chips and similar hardware represent a genuinely enormous share of any AI-focused data centre's build cost, exactly why projects like Yotta's are described in the billions of dollars despite covering comparatively modest physical footprints compared to, say, a steel mill or cement plant of similar capital cost.

Over the next three to five years, expect continued hyperscale capacity build-out toward the 4-9 GW 2030 range, further IndiaAI Mission GPU expansion toward and potentially beyond the 100,000 target, and a National Data Centre Policy finalisation that could reshape how state and central incentives interact.

The real test is whether India's massive capital commitment to physical AI infrastructure, hundreds of billions of dollars across just a few announced projects, translates into genuine AI development leadership, sovereign models with real adoption, a thriving startup ecosystem using subsidised compute, rather than simply building infrastructure that ends up serving primarily foreign AI labs' training needs.