E-Paper
India’s AI push needs a resource roadmap, not just more servers
Authored by - Manish Dubey, chief-practice and Amir Bazaz, head-practice, Infrastructure and Climate, Indian Institute for Human Settlements (IIHS).
India’s AI ambitions are now being poured into steel and concrete. By 2030, the country’s data centre capacity is projected to rise from roughly a gigawatt today to about 8 GW, backed by tens of billions of dollars of investment and a broader bet that the build-out will anchor India as an AI and cloud hub, generate steady hosting revenues, and create tens of thousands of skilled and semiskilled jobs. Political and industry signals already suggest that 8-9 GW is a milestone, not a finish line.
However, even that first 8 GW means acknowledging that AI will not be a light digital layer but a heavy industrial sector with major demands on power, water, and materials. To hit 8 GW by 2030, data centres will consume 50-70 TWh of electricity annually, roughly the demand of a small industrial state like Haryana or Chhattisgarh, or the combined load of a big urban pair like Mumbai and Pune.
Water demand, counting both cooling and power generation, could rise from about 150 billion litres annually in the mid-2020s to about 300-400 billion later this decade, nearly half of Bengaluru’s yearly municipal consumption. And the associated S 35–50 billion-wave of investment will drive demand for copper and aluminium cabling, high-capacity transformers and switchgear, steel, concrete, batteries, and server hardware in an economy that still imports much of its high-end electronics and many critical minerals. Every additional GW beyond 8 GW will magnify these pressures further.
The strain will show first in India’s AI hubs--Mumbai-Navi Mumbai, Chennai, Hyderabad, Pune, Bengaluru, and Noida--where grids are constrained, water is stressed, and urbanisation is accelerating. In the US, data centres already account for much of new load growth as residents protest water use. India is entering that same curve with less headroom, making a long-term lifecycle roadmap unavoidable.
The answer is not to pull back from AI, but to build smarter. AI clusters must be treated as permanent grid infrastructure: Embedded in national and state power plans, tied to specific renewable assets, and supported by advance investment in substations, transmission and storage. Approval for large campuses should depend on credible long-term power supply plans with minimum renewable energy shares and defined efficiency pathways, not just available local capacity.
Policy must also separate fresh and non-fresh water. If data centre demand runs into hundreds of billions of litres annually, it is critical to avoid draining fragile aquifers. In water-scarce basins, cooling should rely only on treated wastewater or closed loops, with strict caps on freshwater withdrawals tied to municipal demand and mandatory disclosure of use and recycling rates.
On materials, mining, manufacturing and recycling must align with long-term AI demand. The need for vast quantities of copper, aluminium, steel, concrete, batteries and servers--even as global AI and energy transitions tighten supplies--will require pushing domestic mineral exploration, scaling manufacturing of transformers, power electronics, cables and server assemblies, and establishing large-scale e-waste and server-recycling units near AI hubs to recover metals and reduce imports.
Governance will determine whether India’s AI push becomes a sustainable transformation or another infrastructure scramble. The country needs a 25-year framework to manage its AI resource footprint. Fragmented city-level control risks weak oversight, while overcentralisation ignores local realities. A hybrid model offers balance: a national mission to set standards and coordinate power, water and materials planning, paired with empowered state and urban authorities to enforce resource budgets, finance grid upgrades, and expand wastewater and recycling systems. Building that framework will also demand new skills. Alongside AI coders, India must cultivate grid planners, power engineers, water systems specialists, and recycling and environmental experts embedded within utilities, regulators, and city agencies. The IndiaAI Mission and state skill programmes should treat these capacities as core to the AI economy.
India’s AI expansion can drive lasting growth and resilient jobs only if it is anchored in a clear resource roadmap, one that treats power, water, and materials as core infrastructure, not afterthoughts. Integrating these elements into planning and governance now will determine whether India’s AI ambition endures as a sustainable technological transformation or stalls under the weight of its own resource demands.
This article is authored by Manish Dubey, chief-practice and Amir Bazaz, head-practice, Infrastructure and Climate, Indian Institute for Human Settlements (IIHS).
