When the lights must stay on: India's data centre boom becomes a power story

Four years ago, India's total data-centre capacity stood at roughly 375 MW, enough to power a mid-sized city, but modest by global standards. By 2025, that figure had crossed 1,500 MW. The quadrupling reflects more than digital ambition; it marks a structural shift in how India's economy runs. Cloud adoption accelerated by the pandemic, the rollout of 5G, and an enterprise pivot toward AI-ready infrastructure have collectively turned data centres from niche assets into critical national infrastructure.

The numbers ahead are starker. Electricity demand from data centres is projected to hit 13.56 GW by 2031–32. Annual consumption could climb from roughly 13 TWh today to around 57 TWh by 2030, approximately 3% of India's total electricity demand. To put that in context: the sector currently accounts for about 0.84% of national electricity use. Within this decade, it could consume as much power as several large industrial states combined.

India's data-centre story is no longer just about compute capacity. It is becoming a power-infrastructure story, with AI-grade GPUs at the centre of that shift.

The Geography of Compute

Mumbai leads the market, followed closely by Chennai, Bengaluru, Hyderabad, and Delhi-NCR. Each hub reflects a different pull: Mumbai for financial services and submarine cable connectivity, Bengaluru for enterprise tech demand, Hyderabad for large campuses with state-backed incentives. Since 2019, India has attracted roughly $94 billion in data-centre investment commitments, a figure that signals long-term confidence from hyperscalers, domestic operators, and global infrastructure funds alike.

For enterprise CIOs evaluating where to place workloads, this investment momentum matters. A well-capitalised data-centre market means better SLAs, more resilient infrastructure, and greater options for hybrid and multi-cloud architectures. But it also raises a question the industry has been slow to answer publicly: can the grid keep pace?

The Industrial Load Nobody Planned For

Data centres are no longer simply backend IT facilities. They are becoming a distinct industrial load class in national grid planning, comparable, in aggregate draw, to large manufacturing clusters. The distinction between MW and TWh matters here: installed capacity (MW/GW) describes how much power a facility can draw at any moment; annual consumption (TWh) describes the electricity actually used over time. As data centres run around the clock, even moderate installed capacity translates into enormous annual consumption.

AI workloads have made this dramatically more acute. An AI-grade data centre, dense with GPU racks, high-performance networking, and precision cooling, can consume five to ten times more power per unit than a traditional enterprise data centre. Inference and training workloads do not rest. The result is that AI compute is, in practical terms, a long-term power commitment dressed up as digital infrastructure.

The New Variables in Enterprise Planning

CIOs and infrastructure leaders are rethinking their location calculus. Latency and connectivity still matter, but grid reliability, power density, and sustainability credentials have entered the conversation. States such as Telangana, Maharashtra, Uttar Pradesh, and Gujarat are competing aggressively, offering policy incentives, land parcels, and grid-upgrade commitments to attract hyperscale and colocation investments. Enterprises, in turn, are signing long-term renewable power purchase agreements to meet sustainability targets and hedge against future tariff volatility.

A Global Race India Is Joining Late — but Fast

The US, Europe, and China are already grappling with AI-driven power constraints. Northern Virginia, the world's largest data-centre market, has faced moratoriums on new connections. European operators contend with both grid limits and strict sustainability mandates. China has directed data-centre growth toward regions with surplus hydropower.

India is at an earlier stage but is scaling rapidly. Unlike mature markets that are retrofitting existing grids, India must expand computing capacity and power infrastructure simultaneously, a more complex challenge, but also an opportunity to build smarter from the start.

The Bottleneck Nobody Is Talking About Enough

India's digital infrastructure ambitions — from sovereign AI to smart cities to financial inclusion, all run on the same substrate: reliable, affordable, scalable electricity. The investments are flowing, the demand is real, and the policy intent is strong. But as GPU-dense facilities multiply and AI workloads grow heavier, the sector's defining constraint may prove to be neither silicon nor software.

India added more data-centre capacity in five years than it had built in its entire digital history. What it builds next may depend less on capital or code, and more on whether the grid can keep up.