For years, securing land was the primary headache for data centre operators in India. Not anymore. Sunil Gupta, chief executive of Yotta Data Services, puts it bluntly: "Securing adequate, reliable, and scalable power has become increasingly challenging." The primary operational bottleneck has shifted decisively from land acquisition to power infrastructure.
This isn't just about finding enough megawatts. It's about the sheer complexity of delivering them. "The most significant delays come from constructing high-voltage substations and transmission lines, which require their own land acquisition, right-of-way clearances, and regulatory approvals across utilities and local authorities," Gupta explains. Sridhar Pinnapureddy, chief executive of CtrlS Datacenters, estimates that addressing these bottlenecks will require net investment running into "a few lakh crore rupees".
The timing mismatch is brutal. Data centres can be built in 12-18 months. Solar farms take 18-24 months. But transmission infrastructure? That still takes 3-4 years. Sabyasachi Majumdar, senior director at CareEdge Ratings, points out the obvious problem: "Regulators, discoms, transmission utilities and private companies need to plan long-term". When your facility is ready but your power connection isn't, you're burning capital without generating revenue.
The fundamental challenge isn't just more data centres—it's a different kind of data centre. AI workloads have completely rewritten the power requirements. Conventional data centre racks operate at 8-12 kW. AI racks? They need 50-60 kW. That's a 4-6x increase in power density.
This isn't incremental growth. It's a step change that demands heavier-duty power distribution equipment and significantly stronger grid connectivity. A single hyperscale AI campus can require 100-300 MW of continuous power. When several facilities cluster together, the cumulative impact can overwhelm existing grid assets. The power ministry has informed Parliament that data centres are projected to add 26.3 GW of load to the national grid by FY32. For context, India's current operational data centre capacity is just 1.12 GW as of June 2025.
The renewable energy transition adds another layer of complexity. In Q1FY26, India curtailed 300 GWh of renewable electricity because the transmission network couldn't carry it where it was needed. Vijay Agarwal, managing director at Equirus Capital, notes the obvious implication: "That challenge becomes more acute when AI campuses require hundreds of megawatts of uninterrupted power around the clock".
The power crunch is playing out differently across India's data centre hubs. Odisha illustrates the challenge vividly. The state has enough infrastructure to support up to 80 MW of data centre load, but Hemant Sharma, Odisha's additional chief secretary for industries, acknowledges that "scaling to hyperscale campuses will demand entirely new transmission infrastructure". This matters because Odisha has attracted significant investments, including HCLTech's ₹14,257 crore AI-ready data centre and Adani Group's ₹800 crore facility.
Maharashtra, India's largest data centre hub hosting roughly 50% of national capacity, faces its own constraints. MSETCL's existing transmission network struggles to meet the 500 MW to 1 GW power capacity demands that AI developers need within a year. The state is responding aggressively—public and private firms have poured ₹1,500 crore into upgrading Maharashtra's power infrastructure over the past four months alone. Bajel Projects Ltd, Tata Power, and PowerGrid Corporation of India Ltd are building 400kV and 765kV substations to remove transmission constraints in critical regions like Pune.
Chennai, India's second-largest data centre hub, faces a different problem. A senior Tangedco official explains that apart from power infrastructure shortage, "land for new substations is scarce". Existing data centres are accommodated, but "expanding capacity takes years". This land constraint directly limits the ability to add the substations needed for new data centre capacity.
The financial burden of power infrastructure doesn't fall evenly. The cost-sharing model varies across states, but typically follows a pattern: state distribution companies (discoms) fund substations—unless built inside a data centre campus—while developers pay for transmission lines, step-down transformers, and internal campus cabling.
For operators like CtrlS Datacenters, this means substantial capital expenditure requirements beyond the data centre facility itself. They must front significant capital for transmission infrastructure before having guaranteed power delivery, creating working capital challenges and increasing project financing complexity. The average cost of setting up data centres has already risen from Rs 40-45 crore per MW to Rs 60-70 crore per MW. Add power infrastructure costs, and the economics get even tougher.
The scale of investment is staggering. Over the past 12 months, domestic conglomerates, Big Tech firms (Google, Amazon, Microsoft, Meta), and standalone operators have committed over $250 billion to build new data centre capacities across India. The additional "few lakh crore rupees" for power infrastructure represents a substantial percentage increase in total investment requirements, extending payback periods and potentially affecting internal rate of return calculations.
Here's the irony: state incentives have largely converged, but grid delivery capabilities haven't. Maharashtra, Karnataka, Tamil Nadu, Telangana, and Uttar Pradesh all offer similar packages—power tariff subsidies, electricity duty exemptions, stamp duty waivers, capital investment support. But the actual ability to deliver reliable power varies dramatically.
This has shifted the competitive landscape. Vijay Agarwal puts it directly: "Power reliability and interconnection throughput are now the primary differentiator—ahead of land and headline incentives, because incentives have largely converged across states while grid delivery hasn't".
Data centre operators are adapting their strategies accordingly. Site selection now prioritizes locations with robust grid infrastructure over those with the most attractive incentive packages. Operators are seeking deeper partnerships with state utilities and transmission companies to secure power infrastructure commitments. Some are considering vertical integration into power generation and transmission to ensure reliable supply. Others are diversifying across multiple states to mitigate grid-related risks.
The market is segmenting along grid capability lines. Established hubs like Mumbai and Chennai face grid constraints but offer connectivity advantages. Emerging locations with better grid infrastructure are becoming increasingly attractive. Specialized AI hubs are developing dedicated power infrastructure. Edge computing locations are distributing facilities to reduce individual power requirements.
India's data centre ambition is undeniable. The government projects peak electricity demand to reach 388 GW by March 2032, with AI infrastructure emerging as one of the fastest-growing contributors. The National Electricity Plan envisages developing the transmission system with an estimated investment of about ₹9.16 lakh crore by FY32. Grid modernization measures including STATCOM, battery energy storage systems, and synchronous condensers are being planned to enhance reliability and facilitate renewable integration.
But the timeline pressure is real. Data centres are being built faster than the grid can support them. The 300 GWh of renewable curtailment in Q1FY26 is a warning signal—it demonstrates that current grid infrastructure cannot support the simultaneous expansion of both renewable energy and AI data centres without substantial modernization.
The companies that succeed will be those that treat power infrastructure as a core competency, not an afterthought. Early engagement with utilities, long-term planning, and strategic partnerships with transmission companies will separate the winners from the losers. The $250 billion in committed investments depends on it.
