New Delhi, Dec. 26 -- India's data centre expansion in 2025 reached a scale the industry had not planned for. According to a report by CBRE, India's operational data centre capacity surpassed 1.5 gigawatts of IT load during the year, with another 1.3 GW under development, making it one of the fastest-growing data centre markets globally.

What made this cycle different was the composition of demand. Growth was driven not only by consumer internet and traditional enterprise systems, but by deep investment in artificial intelligence across core business operations. Banks built private AI platforms for fraud detection and compliance, manufacturers deployed vision systems and digital twins, telecom operators integrated AI throughout networks, and IT services firms reserved compute for global AI contracts. These workloads were persistent, power-dense, and latency-sensitive, pushing planned infrastructure well beyond earlier assumptions.

Venkat Sitaram, Senior Director and Country Head, Infrastructure Solutions Group at Dell Technologies India, calls 2025 a structural inflection point. "Physical constraints, especially power and cooling, surfaced faster than the industry anticipated. The biggest non-technical bottleneck was not capital or intent, but the ability to secure reliable, scalable, and cost-effective power in the timelines enterprises now demand," he says.

When land became the first bottleneck

As construction accelerated, suitable land quickly emerged as a limiting factor. CBRE data shows that Mumbai, Chennai, Delhi-NCR, and Bengaluru account for nearly 90 percent of India's operational data centre capacity, leaving little room for expansion without complex zoning, environmental, and planning approvals. Industrial land prices in these clusters rose sharply, forcing companies to reassess location, latency, and connectivity trade-offs.

Srividya Kannan, founder and CEO of Avaali Solutions, says enterprises underestimated the role of regulatory fragmentation. "The most significant constraint wasn't scarcity itself, but inconsistent rules across states in land acquisition, power allocation, and environmental clearances. Enterprises assumed that demand visibility would automatically trigger supply-side responsiveness. Instead, timelines stretched 18 to 24 months longer than projected," she says.

This forced CIOs to rethink cloud architecture and geography. Tier-2 cities gained interest, but gaps in fibre redundancy, grid resilience, and ecosystem maturity limited how quickly workloads could shift.

Power, not GPUs, broke first

By mid-2025, power availability overtook hardware as the primary constraint. A KPMG report on India's data centre growth warns that power infrastructure must scale significantly to support the country's digital ambitions, noting that energy planning, grid coordination, and renewable integration are now central to sustaining data centre growth.

"Demand for large data centres can exceed 100 megawatts for a single facility," says Vineet Mittal, Senior Vice President at Ziroh Labs. "That's equivalent to powering a mid-sized Indian town. Without assurance of power at a reasonable cost, profitability at scale becomes impossible."

AI also altered enterprise cost structures. Ankush Sabharwal, founder and CEO of CoRover, notes that power is no longer a background consideration. "AI scales in months, energy infrastructure scales in years. That mismatch caught the ecosystem unprepared. Power consumption now often exceeds hardware depreciation in total cost of ownership," he says.

Cooling exposed physical limits

India's climate intensified the challenge. High-density AI workloads significantly reduce the effectiveness of traditional air-cooling systems in warm climates, increasing both operational risk and energy consumption unless alternative cooling methods are adopted.

Extended heatwaves in 2025 exposed these limits across multiple regions. Water-based cooling solutions proved difficult to scale in water-stressed zones, while air cooling struggled to support sustained GPU workloads. Liquid and hybrid cooling moved from experimental deployments to operational necessity.

"The cost of cooling can wipe out margins for decades if not designed correctly," Mittal says. "GPUs are extremely sensitive to thermal instability, which leaves very little room for error."

Sustainability became a strategic choice

Sustainability shifted from reporting to operational viability. Enterprises increasingly prioritised data centres with direct access to renewable energy, low power usage effectiveness (PUE), and transparent energy metrics as sustainability goals intersected with uptime, cost, and regulatory risk.

Venkat Sitaram says the shift was unavoidable. "Advanced cooling, energy-efficient servers and system-level optimisation became enablers of scale, not optional features," he says. Long-term renewable power purchase agreements, on-site solar, and battery storage entered mainstream enterprise planning.

A shift toward smaller, owned infrastructure

With hyperscale capacity constrained in core clusters, many enterprises built mid-sized or modular data centres closer to campuses and industrial zones. Hybrid and enterprise-owned infrastructure models gained traction as firms sought predictability in power availability, latency, and governance.

Suresh Bachwani, CTO at Orient Technologies, sees this as a mindset change. "Data centre growth is no longer purely a capital deployment exercise. It is an infrastructure orchestration challenge requiring coordination across utilities, policymakers, and technology planners."

What 2025 revealed

The data centre crunch of 2025 made one thing clear: India's digital ambitions are constrained by physical systems. Land availability, grid readiness, cooling capacity, and energy economics now shape enterprise technology strategy as much as cloud architecture or AI roadmaps.

As Sabharwal puts it, "2025 marked the end of abstract cloud thinking. Compute strategy is now inseparable from energy, location, and infrastructure realism."

As India enters 2026, the focus is shifting from building more to building smarter-through efficiency, distributed architectures, and tighter alignment between enterprise technology planning and national infrastructure capacity.

Published by HT Digital Content Services with permission from TechCircle.