India’s artificial intelligence ambitions are increasingly being measured in gigawatts. The country wants to host more hyperscale data centres, attract global cloud investment and build the infrastructure needed to train and run advanced AI systems. But the next constraint may not be access to processors or fibre networks. It may be whether the electricity system can deliver enough reliable power, in the right locations, at a price that makes these facilities viable.
The scale of the ambition is striking. In a July 2026 reply to Parliament, the Ministry of Power projected an additional 26.3 gigawatts (GW) of electricity load from AI data centres by financial year 2031–32. That is a projection of future demand—not installed capacity, signed power contracts or electricity already being consumed. The distinction matters because India’s ability to convert proposed projects into operating infrastructure will depend on grid connectivity, generation, transmission, land, cooling and financing. (The Ministry’s projection and the grid-planning questions it raises.)
The 26.3 GW number needs context
The government’s estimate is based on projects received from states and consultations with transmission utilities, distribution companies and data-centre developers. It is an important planning signal, but it should not be read as a forecast that every proposed facility will be built on schedule. In an August clarification, the Ministry told ETEnergyWorld that applications for grid connectivity had been submitted for around 17 GW of the projected 26.3 GW. Applications for the remaining 9.3 GW had not yet been filed with the relevant state transmission utilities, based on the position cited in that response.
That gap is not evidence that the projects will fail. It is evidence that a large demand projection and a completed grid connection are very different milestones. A data-centre developer still has to secure land, obtain approvals, arrange power procurement, fund the facility, build the campus and commission the equipment. Grid applications themselves are not equivalent to approved connections or energised capacity. (The Ministry’s clarification on the 17 GW of applications.)
The starting point is also important. The Ministry of Electronics and Information Technology said in an August 2026 parliamentary response that India’s installed data-centre capacity had grown from 375 megawatts in 2020 to about 1,575 MW. It also pointed to emerging investment destinations beyond the established hubs, including Andhra Pradesh, Madhya Pradesh, Chhattisgarh and West Bengal. The distance between today’s installed base and the scale implied by future projections is substantial. (Government update on data-centre capacity, efficiency and water measures.)
The bottleneck is not just generation
It is tempting to frame the challenge as a simple question of producing more electricity. But a data centre needs power delivered continuously, with high reliability and at a specific location. Generation capacity somewhere in the country does not automatically translate into usable power at a particular campus.
That is why transmission and distribution infrastructure matter as much as new generation. Substations, transformers, transmission corridors, redundancy and last-mile connections determine whether a proposed campus can actually operate. The Ministry has said that transmission infrastructure is being developed in phases alongside planned generation additions. The National Electricity Plan cited in the parliamentary response envisages substantial investment in the transmission system through 2032.
There is also a geographic question. India’s data-centre market is concentrated around Mumbai and Navi Mumbai, Chennai, Hyderabad, Bengaluru, Delhi-NCR and Jamnagar. These locations offer connectivity, customers, talent and established digital infrastructure. But adding large, concentrated loads to already busy urban power systems can create local constraints even when national generation capacity appears adequate.
The Indian Express reported that officials and sector stakeholders have been discussing grid stability, sharp changes in data-centre loads and the possibility of locating some facilities closer to renewable-energy sources or away from congested urban networks. The implication is significant: site selection may increasingly be determined by power-system readiness, not simply proximity to a technology cluster. (Reporting on grid stability and location considerations.)
Energy companies are positioning themselves for the AI build-out
The opportunity is already changing how power companies approach data-centre customers. Tata Power said in October that it had built a data-centre opportunity pipeline of more than 2.8 GW across 18 domestic and international operators, spanning distribution, transmission and renewable energy. The company described 29 active workstreams and a broader offer that includes grid connectivity, renewable power, storage, cooling and utility infrastructure.
The composition of that pipeline is revealing: around 430 MW of distribution-related upcoming load, 1,236 MW of transmission-linked opportunities and 1,150 MW of green-energy capacity under active engagement. These are different categories of opportunity, not 2.8 GW of commissioned data-centre capacity or guaranteed revenue. But they show how the value chain is expanding beyond the conventional electricity sale. (Tata Power’s announcement on its data-centre opportunity pipeline.)
For utilities and infrastructure providers, the commercial proposition could include power procurement, transmission access, renewable-energy supply, storage, backup arrangements and cooling-related services. The company that can coordinate these pieces may become a more strategic partner to a hyperscaler than a supplier that simply sells electricity.
That shift is consistent with the wider transformation of AI infrastructure. As we explored in our analysis of why the next AI infrastructure battle is about moving data, compute performance depends on the systems around the processor. Power is another part of that same infrastructure stack—and potentially one of its most geographically constrained components.
Renewable power is necessary, but not a complete answer
The Ministry has said that the projected additional load is expected to be integrated into the grid and primarily served through renewable-energy capacity. That direction is important for India’s climate and energy-transition goals. But matching a data centre’s annual electricity consumption with renewable generation is not the same as ensuring that power is available every hour.
Solar and wind output varies. Data centres, by contrast, are designed to run continuously. The system therefore needs a combination of generation, transmission, storage, flexible demand, backup arrangements and grid balancing. Renewable procurement can help reduce emissions, but the operational design must still address reliability and the timing of supply.
There is also a water dimension. Data-centre cooling choices influence water use, while electricity generation itself can have water implications depending on the source. In its August update, MeitY referred to Indian standards covering Power Usage Effectiveness, Carbon Usage Effectiveness, Cooling Efficiency Ratio and Water Usage Effectiveness. It also cited the use of direct-to-chip liquid cooling, adiabatic cooling, immersion cooling and closed-loop systems. These metrics should become part of how projects are evaluated, not simply technical details buried in facility design.
For India, this creates a more demanding definition of sustainable AI infrastructure: not merely a renewable-energy purchase agreement, but transparent reporting on energy efficiency, water use, emissions, backup systems and the actual performance of the facility.
Where India’s next data-centre hubs may emerge
Established technology markets will remain important. Bengaluru, Hyderabad, Chennai and Mumbai have deep enterprise ecosystems, connectivity and access to skilled talent. Yet the next phase of expansion may favour locations that can combine reliable power, land availability, fibre connectivity, water stewardship and predictable approvals.
That could create opportunities for states seeking to attract digital infrastructure investment. But incentives and land parcels alone will not be enough. States will need to demonstrate that transmission capacity, substations, renewable supply and local resource planning can support large facilities without shifting costs onto households or other industries.
India’s wider AI opportunity is also not limited to owning the largest data-centre campuses. As we argued in India’s opportunity to win the AI economy without winning the model race, the country can create value through deployment, engineering, applications and industry-specific adoption. Reliable infrastructure is what makes those opportunities practical. The question is whether the economic benefits of hosting AI workloads—jobs, investment, technology capability and tax revenues—justify the energy and resource commitments required.
The real test is conversion, not announcements
India’s AI data-centre opportunity is real. So is the scale of the infrastructure challenge. The 26.3 GW projection tells policymakers and investors that the potential demand is large. The 17 GW of reported grid-connectivity applications shows that a significant part of that ambition has entered a formal process. Tata Power’s pipeline illustrates how energy companies are preparing to serve the market. None of these figures, on its own, proves that the full capacity will be built, connected and operated.
The next set of milestones will be more useful than another round of headline investment announcements: how much proposed capacity receives grid approval; how quickly transmission projects are completed; whether power procurement is affordable and reliable; how much new capacity is actually commissioned; and whether operators disclose credible energy and water-efficiency performance.
For data-centre developers, power availability will increasingly shape where they build and how quickly they can scale. For utilities, the AI boom creates a new class of large, sophisticated customers—and a reason to invest in grid modernisation, storage and integrated energy services. For state governments, it raises a difficult planning question: how to attract strategic infrastructure without compromising the reliability and affordability of electricity for everyone else.
India’s AI ambitions will not be decided by compute capacity alone. The competitive advantage will belong to the ecosystem that can connect compute to reliable power, efficient cooling, resilient networks and responsible resource use. The next AI race may be fought in data centres. But much of it will be won—or lost—outside their walls.
FutureIsNow | Signals over noise. Intelligence for what comes next.
