What India Can Learn from the US About Data Centres

Last week, the Power Ministry began consulting data-centre developers, transmission utilities, and discoms as it works on a roadmap for meeting the sector's rapidly rising electricity demand, opening dialogue on reliable supply, dual-grid connectivity, green power, and grid stability. The government's own projections are striking: internal planning at the Ministry of Power points to electricity demand from data centres reaching 26 GW by 2032 and 35.7 GW by 2040, a near nine-fold rise within the decade. The ministry has just started drawing the map. This article sets out what that map must contain.

India's total operational data centre capacity crossed approximately 1.5 to 1.7 GW by late 2025, with cumulative long-term investment commitments projected to exceed $100 to $125 billion by 2030-32 (CBRE India Real Estate Reports; JLL Data Centre Intelligence). And yet the sector still lacks a binding national policy framework. India still lacks a single, comprehensive national framework governing the energy, water, and grid implications of data centre expansion. Growth has been shaped state by state, through Maharashtra, Tamil Nadu, Uttar Pradesh, and Telangana, through incentives and bilateral agreements, much like green hydrogen before it. The one thing a data centre cannot negotiate with is physics. The pressures building beneath this boom deserve to be named plainly.

The reason lies in the way data centres consume electricity. Three pressures stand out.

1. Relentless baseload. A 100 MW data centre operating continuously would consume approximately 876 GWh a year in IT load alone. Cooling and other facility systems add substantially to the electricity required beyond that IT load. Unlike most industrial loads, this demand runs at full intensity around the clock, with zero tolerance for interruption.

2. Geographic clustering. Mumbai alone holds between 45% and 50% of India's total data centre capacity, anchored by its subsea cable landings. The top four hubs, Mumbai, Chennai, Bengaluru, and NCR/Hyderabad, account for over 85 to 90% of total tier-I colocation supply (JLL India Data Centre Report; CBRE India). These are cities already stretched by the demands of urbanisation, and developers in Navi Mumbai have reported interconnection lead times of 18 to 36 months for dedicated 100 MW-plus substation setups (MSEDCL; Data Centre Association of India), with the queue only growing.

3. Clean energy bottleneck. The challenge is not generating green megawatts. India is doing that. The real challenge is delivering firm power to a specific substation at 3 in the morning on a windless night. Intermittent renewables alone cannot solve that.

The American preview

The United States is not a cautionary tale from a distance. It is a live demonstration of where India is heading, and the numbers make uncomfortable reading.

By April 2026, ERCOT, the Texas grid, was tracking approximately 410 GW of large-load interconnection requests, of which around 87% were from data centres (ERCOT Senate Committee Presentation, April 2026). In Q1 2026 alone, 198 GW of large load applied for interconnection (Ascend Analytics, May 2026), more than double the grid's all-time peak demand, arriving in a single quarter. PJM capacity prices have risen sharply as forecast demand growth, including from data centres, has outpaced new supply, hitting the regulatory cap. Last week, New York became the first US state to impose a moratorium: Governor Kathy Hochul signed Executive Order 62 on July 14, 2026, pausing state environmental permits for up to one year for new hyperscale data centres of 50 MW or above, and pursuing legislation to repeal their sales-tax exemptions (Governor Hochul's Office, July 14, 2026). As of May 2026, nearly 12 GW of data centre load requests were already in the NYISO interconnection queue, with more than 8 GW entering in 2025 alone, as cited in the Executive Order itself.

Public sentiment has turned sharply. Only one in three Americans now approves of the pace of data centre construction, and only 14% would be comfortable with a centre being built in their own community, according to a Reuters/Ipsos poll (June 2026), and dozens of state legislatures are weighing similar measures. India has the chance to avoid this entirely through better planning upfront. That advantage should not be squandered.

Four lessons India cannot afford to ignore

A. Demand flexibility from day one

Every new connection should be conditioned on curtailment capability from the start, paired with time-of-day tariffs that carry real incentives. A data centre that sheds 10 to 15% of its load at the evening peak is not a burden on the grid. It is an asset. This needs to be built in from the first approval, not added later when the grid is already struggling.

B. Embrace the off-grid data centre, but engineer it better than America

This is where India's most significant opportunity lies. S&P Global Commodity Insights and Wood Mackenzie estimate over 50 GW of planned behind-the-meter gas generation tied to hyperscale data centres across ERCOT, PJM, and MISO (S&P Global Market Intelligence). Roughly 2 GW is already operational, including xAI's Colossus in Memphis, which went live in mid-2024 using up to 35 mobile gas turbines capable of around 420 MW, bypassing the local utility queue and getting operational within roughly 100 days of the decision (Global Energy Monitor; Greater Memphis Chamber), and the Stargate campus in Abilene, Texas.

The commercial logic is simple. AI data centres running Nvidia H100 or B200 chips can earn USD 10 to 12 million per MW per year in gross infrastructure hosting revenue (Bernstein Research; Morgan Stanley AI Infrastructure Reports). Getting operational in two years rather than eight easily outweighs the cost of generating power on-site. Speed is what the business demands.

The important nuance is that the American build-out is optimised for speed, not efficiency. Turbines are sold out years ahead, so developers deploy whatever they can get. This is precisely where India's opening lies: take the model, and build it better.

Modern lean-burn gas engines, such as the Wärtsilä 34SG or the Jenbacher J624, deliver 42 to 48% electrical efficiency, well ahead of the simple-cycle turbines being deployed on-site across the US (Wärtsilä Technical Datasheet; Jenbacher Specifications). They hold that efficiency at partial load, synchronise in about a minute and reach full output within minutes. Gas turbines lose up to 15 to 20% of their capacity at Indian summer temperatures, while reciprocating engines derate by under 5% at the same conditions (ASME Digital Collection). Trigeneration, which pairs engines with heat recovery to drive the chillers, lifts total fuel utilisation beyond what any turbine setup can achieve, while taking 30 to 40% of the cooling load off the primary power supply (U.S. DOE Combined Heat and Power Technical Resources). In a country where cooling is a year-round necessity, not a seasonal one, this is not a marginal gain. It is a structural one.

India also carries two advantages into this that the US simply does not have. First, India has approximately 24.8 GW of installed gas-based generation capacity, of which 14 to 18 GW is stranded or running well below capacity, starved of affordable fuel, but with the sites, infrastructure, and expertise already in place (Central Electricity Authority; IEEFA, July 2025). Easing LNG prices are already prompting reactivation of approximately 17 GW of this idle capacity as data centre demand accelerates (Reuters, March 2026). Second, Section 9 of the Electricity Act, 2003 provides a clear legal framework for captive generation, allowing a data centre developer to set up on-site generation without a generation licence, subject to applicable requirements under the Act and the Electricity Rules, 2005 (Ministry of Power). The regulatory path that US developers had to build from scratch already exists in Indian law.

C. Decentralise deliberately

India should put policy weight behind the data centre and edge computing investments already taking shape in Bhubaneswar, Lucknow, and Guwahati, where state governments in Odisha, Uttar Pradesh, and Assam have signed MoUs with providers including Nxtra, STT GDC, and RailTel (STT GDC India; MeitY Data Centre Policy Framework). Data Centre Economic Zones in tier-2 locations, chosen for grid headroom and water availability rather than just fibre access, would spread the load more evenly across the national network. Northern Virginia illustrates the risks of allowing data centre growth to become heavily concentrated in a single region. One county carrying an outsized share of national capacity is the cost of leaving geography entirely to market forces.

D. Judicious planning and conditional access

The US is now moving toward dedicated large-load tariffs, minimum commitment obligations, and cost-allocation rules that make data centres pay their fair share of the infrastructure they require. New York has gone further with an outright moratorium. India does not need to reach that point, but it does need to make energy-efficiency standards, water norms, and power self-sufficiency conditions of connection from the very first application, not as corrections applied after the industry has already settled in.

The regulatory leverage India holds today is exactly what the US handed away during the early boom years and is now trying to claw back. The right time to use it is before a moratorium starts to look like the only option.

The window to act is now.

India has a genuine advantage here, the ability to watch another large market take the fast route, run into the consequences, and still have time to make different choices. The data centre boom is already here. The Power Ministry's grid roadmap is a real and necessary step. But a roadmap is only as useful as the rules written into it and the firm power capacity standing behind it. The wave is here. The question is whether the framework will be ready when it arrives.

This article is authored by Raman Chopra, CEO and whole time Director Caparo Power Limited.