At a glance
- A rushed commissioning can undercut even a well-designed data centre facility.
- AI is pushing operators to weigh lifecycle economics over upfront cost, balancing energy use, water consumption and commissioning quality against rising rack densities.
- Grid transmission, not power generation, is becoming the critical constraint on India's next wave of hyperscale growth.
Ceremonies like this happen often now. A state minister breaks ground, a capacity figure makes the news, and a completion date goes into the press release. What gets far less attention is what happens after the ribbon is cut. A large share of that announced capacity still isn't operational, and on several campuses, commissioning timelines now stretch to 2028 and beyond.
This is rarely about engineering. Most of these campuses are well designed. What gets them is commissioning that slips, contractors juggling too many sites at once, and testing schedules cut short to hit a deadline.
Cloud adoption, AI workloads and enterprise digitisation are driving sustained demand for data centre capacity, but adding megawatts is only part of the challenge. The harder task is ensuring these facilities stay online every minute of every day. That task is only getting harder as campuses scale and workloads intensify.
Operators are building larger campuses against compressed delivery schedules, rising energy costs and growing pressure on the electricity grid. At the same time, AI is pushing rack densities to levels that conventional cooling and power architectures are unable to support. Every engineering decision, including cooling systems, electrical redundancy, commissioning and operations, now carries greater operational consequences.
Reliable 24/7 uptime is no longer achieved by adding more backup equipment, but by integrating cooling, power, modular construction, mechanical, electrical and plumbing (MEP) execution and operational discipline into one single engineering strategy.
The industry's focus has therefore shifted from building faster to building smarter.
Why cooling is no longer just about lowering PUE
Rooftop cooling plant serving a data centre campus. Illustrative.
Power usage effectiveness (PUE) remains the benchmark for operational efficiency, but achieving low PUE in India requires engineering for local climate rather than replicating global designs. High ambient temperatures, seasonal humidity and increasing AI workloads demand cooling systems that balance efficiency with resilience.
Air-side economisers, which use outside air for free cooling, deliver meaningful energy savings in cooler climates elsewhere. Their role in India, however, is structurally limited. High humidity along the coasts and dust in major urban centres reduce the number of hours when outside air can be used without compromising equipment reliability. For most operators, economisation supplements rather than replacing mechanical cooling.
The rapid adoption of AI is also changing the equation. High-density GPU racks generate significantly more heat than conventional enterprise workloads, making liquid cooling increasingly difficult to avoid. Although the technology requires higher upfront investment, its superior heat transfer allows operators to support greater compute density whilst controlling long-term energy consumption.
"Many decisions are taken based on upfront cost rather than lifecycle effects. Equipment that appears cheaper initially can actually cost more over five or ten years."
— Sandeep V. Dandekar, data centre expert and advisor.
The industry's sustainability priorities are evolving as well. Whilst operators have spent years driving down PUE, water consumption is emerging as an equally important metric. Adiabatic cooling systems can improve energy efficiency, but they also increase dependence on water, an increasingly scarce resource in several Indian markets.