India’s data-centre market is entering a phase in which power availability and physical infrastructure are becoming as important as cloud software. Recent industry estimates and company statements point to a rapid increase in capacity over the next decade, with artificial intelligence emerging as one of the main reasons for the expansion.
Microsoft India and South Asia president Puneet Chandok has said India currently has around 2GW of data-centre capacity and could reach 12GW to 14GW by 2035. Other state-level plans are even larger. Andhra Pradesh, for example, says it has secured agreements associated with a 6.5GW data-centre capacity target, with projects moving toward execution.
These figures should not be treated as a single national forecast. They represent different definitions, timelines and project stages. Some announced capacity is operational, some is under construction and some depends on future investment and power availability. The common point is that India’s infrastructure requirements are growing quickly.
AI changes the power equation
Traditional cloud services already require large server facilities, but AI workloads can create much higher power density. Training and inference systems use large numbers of accelerators, memory devices and high-speed networking equipment. Cooling becomes more complicated as racks consume more electricity and generate more heat.
This makes the location of a data centre increasingly dependent on electricity. Fibre connectivity remains important, but operators also need reliable grid access, suitable land, water or alternative cooling systems, and a regulatory environment that can support large industrial projects.
The growth of AI also changes how capacity is used. Conventional enterprise workloads may have relatively predictable demand. AI inference can grow rapidly as applications become more widely adopted, while training clusters can create very high short-term loads. Operators therefore need infrastructure that can accommodate changing computing patterns.
Andhra Pradesh is becoming a major test case
Andhra Pradesh has positioned itself as a major destination for large data-centre projects. The state says it has agreements covering 6.5GW of targeted capacity. Google is separately developing a large AI data-centre project in the Visakhapatnam region, while other companies have announced additional infrastructure plans.
The scale has raised questions about land, water and electricity. Those concerns are not unique to India. Data-centre expansion in the United States and Europe has also triggered debates over grid connections, water use and the impact of large computing facilities on local infrastructure.
For India, the opportunity is substantial. Large data centres can create demand for construction, electrical equipment, cooling systems, networking and operations. They can also improve the availability of local cloud infrastructure and reduce the need to process every workload outside the country.
Power may become the limiting factor
The next phase of the market will depend heavily on power planning. Data-centre operators cannot simply add servers when demand increases if the electricity connection cannot support the additional load.
That is why government officials and technology companies are increasingly discussing data centres in the same conversation as renewable energy, transmission infrastructure and industrial corridors. Locating computing facilities near reliable generation can reduce some infrastructure pressure, although it does not remove the need for strong network connectivity and resilient power distribution.
Cooling is another major issue. Microsoft says its Hyderabad cloud region uses air-cooled chillers and is designed for zero water use for cooling. Other operators are experimenting with different cooling technologies as rack power density increases.
India’s data-centre expansion is therefore becoming an infrastructure story rather than simply a cloud story. The country has strong demand for digital services and a large technology workforce, but turning that demand into AI capacity requires power, land, cooling, networks and capital at a scale that is only now becoming visible.