States should evaluate natural resources before pitching for data centres: NITI Aayog

The proposed framework would make resource verification a condition for data-centre approvals and promotion. States would need evidence that electricity, water and other infrastructure are available. This matters because large facilities can place pressure on public utilities and local resources.

It also shifts attention from individual projects to wider planning. States would assess entire campuses and infrastructure corridors, prepare 20-30-year seasonal water assessments, protect existing users and require developers to fund grid assets built exclusively for their facilities. They would also measure benefits after accounting for land, infrastructure, utility concessions and other support.

The framework comes as India’s industry expands rapidly. India had 271 data centres in January 2026, with capacity at about 1.57 GW in August. The presentation does not state whether the recommendations have been formally adopted or will bind States.

In general, a data centre is a facility containing computer servers, storage systems, networking equipment and cooling systems. It receives, processes and stores digital information. Cloud computing depends on such facilities because online services and data can run on shared remote infrastructure instead of only on a user’s device.

Artificial intelligence also relies on data centres to train and operate models. The article highlights Large Language Models, whose rapid progress has followed the accelerated addition of graphics processing units. These GPUs perform the intensive calculations AI requires, but they are extremely power hungry and need large quantities of freshwater for cooling.

That makes data centres both digital foundations and major infrastructure projects. India’s expanding industry could support cloud and AI growth, while also increasing pressure on electricity systems, water supplies, land and local ecosystems. The proposed framework seeks to align expansion with those physical limits.

India’s data-centre industry is already substantial and expanding quickly. The country had 271 data centres as of January 2026. Their installed capacity stood at about 1.57 gigawatts in August 2026, according to the Union government. States are competing actively to attract more investment.

Industry estimates put possible capacity at 4.5-6.5 GW by 2030. Electricity use was approximately 13 terawatt-hours in 2024, equal to about 0.8% of India’s total electricity demand. The CEEW projects consumption to rise nearly fivefold to 57 TWh by 2030, or about 2.6% of demand.

This growth explains the proposed planning rules. Maharashtra and Tamil Nadu lead the investment race, while several other States have announced major targets. Expansion could strengthen India’s digital and AI infrastructure, but it will also intensify questions about electricity, water, land and public support.

AI-focused data centres perform far more intensive computing than many ordinary digital workloads. Large Language Models depend on graphics processing units, or GPUs, for the calculations involved in developing and running them. The article links the rapid progress of these models to the accelerated addition of GPUs.

The key mechanism is scale. When many GPUs operate together, their combined electricity demand becomes very large. They also generate substantial heat, so facilities need cooling systems. The article specifically notes that these systems require very large quantities of freshwater. Electricity powers the computing, while water helps manage the resulting heat.

This combination makes AI infrastructure especially important to plan carefully. India’s data-centre electricity consumption was about 13 TWh in 2024 and could reach 57 TWh by 2030. Local objections in Maharashtra and Andhra Pradesh show why water and environmental effects are becoming central concerns.

Without long-term planning, data centres could compete with existing users for water and electricity. The proposed framework therefore asks States to prepare 20-30-year seasonal water assessments and protect people and activities that already depend on those supplies. It also links approvals to verified infrastructure availability.

Electricity grids could face new demands, while land and local ecosystems could experience greater pressure. The article records protests in Thane over an Amazon facility’s resource use and proximity to homes, schools and hospitals. Near Visakhapatnam, objections to Google’s proposed project cite water use and ecological concerns. These examples show how expansion can create local conflict.

The framework proposes planning campuses and infrastructure corridors rather than approving projects one by one. It also includes environmental safeguards, climate risks and governance. That approach could help States compare economic gains with resource costs before encouraging further investment.

The proposed framework says developers should fund grid assets built exclusively for their own data-centre projects. The principle is direct: infrastructure created specifically for one facility should not automatically become a cost carried by the wider public. It would make the project’s real economic cost clearer.

States are also asked to pitch benefits after accounting for land, infrastructure, utility concessions and other State support. That means employment, investment or other gains would be considered alongside the value of public assistance. Incentives could be linked to measurable outcomes such as jobs, renewable-energy use, efficiency and local procurement.

This matters as States compete aggressively for investment. Maharashtra and Tamil Nadu lead, while Gujarat, Uttar Pradesh, Andhra Pradesh, Odisha, Telangana and Madhya Pradesh pursue projects or targets. Requiring project-specific grid funding could improve accountability, although the presentation does not say whether the framework is binding.

A gigawatt, or GW, measures power capacity: the rate at which equipment can use or supply electricity at a particular moment. In this context, it describes the installed computing and facility capacity of India’s data centres. India had about 1.57 GW installed in August 2026.

A terawatt-hour, or TWh, measures energy consumed over a period. It captures how much electricity data centres actually used. India’s data centres consumed approximately 13 TWh in 2024. The CEEW projects that figure to reach 57 TWh by 2030, equal to about 2.6% of total demand.

Both units are needed because a larger capacity does not automatically reveal yearly consumption. GW describes the industry’s ability to operate computing equipment, while TWh shows its cumulative electricity requirement. Together, they illustrate both possible expansion and its effect on India’s power system.