Amazon Web Services (AWS), Microsoft, and Google have committed more than USD 50 bn to build cloud and AI infrastructure in India—the country’s largest-ever digital investment cycle rivalling UAE and Saudi Arabia. But while capital is accelerating, the physical reality is lagging: Power, water, and talent shortages now threaten to disrupt the pipeline.

The demand is locked in: 5G rollout, AI deployment and data-localization laws have pushed installed capacity from 590 MW in 2019 to 1.4 GW today with a projected five-fold increase to 8 GW by 2030. But the supply is fragile as India hosts nearly 20% of the world’s data but holds less than 6% of global capacity.

“This is not a normal incremental build-out, it’s hyperscaling,” Unaise Urfi, partner at KPMG in India, told EnterpriseAM. “The demand is real and rising. Where there is sustained demand, supply will follow. Hyperscalers don’t commit at this scale unless they see a long runway.” Urfi cautioned that long-term returns will now be determined by operational conditions — particularly around power, water, land readiness, and skilled manpower.

Power and water constraints

Power constraint: Hyperscale facilities require near-perfect uptime, but grid reliability across major hubs such as Mumbai and Chennai still falls short of the “five nines” (99.999%). Data centers could consume 2.6% of India’s total electricity by 2030, up from 0.8% today.

“Power is now the most existential constraint,” Urfi said. Operators continue to rely heavily on diesel generator back-up — that is increasingly costly, carbon-intensive, and exposed to regulatory scrutiny.

As hyperscalers tighten renewable-energy requirements across their global supply chains, power is shifting from a “basic utility input to a license-to-operate issue”, Urfi added.

AI-grade servers consume 5-6x more power than conventional servers, creating heat densities that legacy cooling cannot handle. “Power is not just needed for computation anymore — it is also needed for cooling at much higher densities,” Urfi said.

Cooling costs: This has pushed water availability to the centre of site-selection decisions. Legacy cooling systems rely heavily on water-intensive evaporation methods, making sustainability and water stewardship increasingly critical. An average 1 MW data-centre load can consume more than 25 mn litres of water annually for cooling. “The industry has long focused on PUE (power usage effectiveness), WUE (water usage effectiveness) is becoming equally non-negotiable,” Urfi said.

With most data centres concentrated in water-stressed urban markets — Mumbai and Chennai alone account for over 70% of India’s installed capacity — resource realism apart from connectivity or customer proximity, is beginning to shape where new capacity gets built.

Permits are faster, but land readiness lags

India’s permitting environment has improved materially over the past five years, reflecting greater recognition of data centres as strategic digital infrastructure. Approval timelines are shorter and procedural clarity has increased across several states.

Land readiness: However, bottlenecks have shifted. “The challenge today is less about approvals in principle and more about making land infrastructure-ready,” Urfi said, pointing to delays in power connectivity, environmental clearances, cooling-related permissions, and grid readiness. Large hyperscale campuses typically require 20-50 acres of contiguous land near power and fiber infrastructure, a combination that remains difficult to secure quickly in metro regions.

Talent is the next binding constraint

While India has a deep IT talent pool, data-centre operations require a different skill set altogether — spanning mechanical and electrical systems, advanced cooling, critical infrastructure safety, and data-centre infrastructure management systems.“As AI workloads grow, operations are shifting toward GPU-dense environments that require continuous monitoring and automation,” Urfi said. “Without proactive reskilling, talent constraints could slow commissioning and increase operating risk.”

Execution test: For now, hyperscalers with strong balance sheets and long-term capacity commitments remain insulated. But for smaller or merchant developers, delays or cost overruns could quickly strain returns.

The Bottom Line: India’s data-center ambition is no longer constrained by capital. The winner will be whoever can solve the power, water and skills equation first.

(** Tap or click the headline above to read this story with all of the links to our background as well as external sources.)

Tags: