Although nascent, India’s data centre sector is on the brink of a new era, where demands of artificial intelligence (AI), cloud and digital infrastructure are redefining power, speed and global competitiveness. In a panel discussion organised by tele.net in Mumbai, key stakeholders gathered to deliberate on the evolving landscape of future-ready data centres. Key takeaways from this discussion include the following perspectives…
Digital economy, cloud adoption and AI are driving a surge in demand for the country’s data centre industry, which provides substantial headroom for growth despite the current scale being small, relative to global markets. India offers immense opportunity – land, security, fibre, low latency and above all, reliable, skilled talent and low-cost green power. India has a national grid that is robust and expanding. Companies can integrate diverse renewable assets such as wind farms in Maharashtra and Karnataka, solar power in Rajasthan, pumped hydro in Andhra Pradesh and distributed batteries and active participation in the Indian Energy Exchange (IEX) to dynamically balance and match them in real time in 15-minute blocks, delivering service level agreements (SLAs) close to 95-100 per cent at very competitive prices.
The sector has expanded from roughly 300 MW seven to eight years ago to about 1.2–1.4 GW today, with single operators exceeding 300 MW, and maintaining 25–30 per cent compound annual growth rate even on a larger base point.
Since AI/machine learning (ML) compute can be location‑agnostic, compared with low‑latency cloud and enterprise workloads, location diversity must expand beyond Mumbai and Chennai. Real estate is not a major challenge. As global demand grows, scale and power availability become decisive and India’s offering looks increasingly strong, the AI revolution depends on how fast data centres can be built and powered. The focus is renewable power, fibre and complete end-to-end solutions.
Gearing up for the future
Future‑ready design requires moving from legacy 3 kW environments to scalable 3–150 kW deployments, making 24×7 green power or renewable‑energy credits a non‑negotiable prerequisite for sustained expansion. Further, it drives power usage effectiveness (PUE) toward internationally cited benchmarks through liquid and immersion cooling, while planning for power constraints that can emerge after initial growth cycles, as seen in markets such as Malaysia.
Round‑the‑clock green power is positioned as essential for data centres because platform‑level sustainability pledges target carbon‑free operations by 2030, large global tech clients impose 100 per cent green power requirements or demand a five‑year roadmap to reach it. India’s low‑cost renewables can reduce the total cost of ownership when delivered efficiently to the meter.
High‑density AI is redefining specifications, with rack power leaping from 5-20 kW to 100–150 kW and cooling architectures transitioning from air to liquid and immersion even as facilities remain largely mixed‑use rather than neatly segmented into “AI‑only” or “cloud‑only” designs. The governing principle is flexibility; halls and entire floors must accommodate workloads from 5 kW to 150 kW across power, cooling and security so that operators can serve enterprises and hyperscalers, whether cloud or AI, without rebuilds or disruption. In practice, this means reserving open space and upstream capacity for step‑ups, as illustrated by a campus planned at approximately 30 MW with approximately 16,000 graphics processing units that scaled IT load to approximately 50 MW, by adding generators and chillers into pre‑planned areas, a path that is far harder on constrained city parcels with limited substation intake around 50 MW. The decisive capability for future‑ready facilities is to expand IT power and switch cooling strategies mid‑operations as customer needs evolve, changing design and thermal approach on the fly, while protecting reliability and service levels.
Net net, the emphasis is on aligning installed generation with contracted availability so that delivered green energy matches workload density and uptime expectations at scale.
Embracing energy security
Sustainability‑linked finance at the holding‑company level drives renewable expansion, aligning capital commitments with green milestones embedded in perpetual bond guidelines and reinforcing growth in clean energy procurement for operations and customers. That said, the commercial case for green power hinges on delivered cost at the meter, or it will not be adopted at scale, even amid policy and environmental pressures to maximise clean supply. State‑level policy frameworks shape renewable adoption strategies, with some jurisdictions enabling up to 90 per cent renewable pass‑through, while others cap it near 10 per cent; yet siting decisions still weigh multiple factors as evidenced by hyperscalers choosing constrained states when broader economics and ecosystem advantages align. Unless interstate transmission is made easier, storage is permitted and interstate subsidy charges are rationalised, the price can converge with standard grid tariffs, reducing urgency for customers despite net‑zero roadmaps.
Power strategy and grid realities
Grid readiness in Tier 1 cities benefits from sustained CTU and STU build‑outs, with hubs such as Mumbai enjoying rare levels of infrastructure depth that reduce connection friction, compared to markets facing multi‑year interconnection queues; and campus requirements have expanded from tens of MWs to around 500 mobile value-added services (MVA) per site, implying localised demand as high as 1,000 MVA, where neighbouring campuses cluster. National grid integration under a unified framework since 2013 further reduces the immediate need for on‑premises combined‑cycle plants or small modular reactors in data centre yards, underscoring a structurally different context from markets that solve fragmented grids.
Even with these advantages, operational prudence remains essential, as curtailment episodes in the past have forced facilities to shift to diesel during peak evening hours, highlighting that energy security must be engineered into plans because policy responses in a large democracy can change under stress. The practical takeaway is to pair cost‑effective green procurement with risk‑mitigation architectures that ensure continuity through policy variation, local network strain and rising density and scale of modern workloads.
Liquid cooling investments
Liquid immersion cooling is moving from pilots to mainstream investment, with operators converting in‑construction facilities where floor‑to‑ceiling and layout constraints allow. Large‑scale AI loads already run on liquid cooling in international campuses and dedicated innovation hubs in India are advancing readiness for high‑density deployments.
Cabling and supply chain
Connectivity sits immediately behind power as foundational for data centre performance, with early industry growth in Mumbai and Chennai anchored by submarine links and new undersea systems connecting Malaysia, India, Singapore and Thailand extending global reach, making intelligent cabling a paramount priority across campuses and routes. Beyond network fibre, the incoming underground high‑voltage circuits feeding GIS, 220 kV, 132 kV, or 110 kV, are candidates for embedded sensing, while advances such as superconducting grids that target near‑zero resistive loss and smaller cable footprints exemplify the kind of innovation the sector seeks to adopt at scale.
India’s cabling practice typically overbuilds redundancy due to frequent fibre cuts from digs and roadworks, but the next efficiency frontier is route hardening – deeper burial at 2–3 metres, continuous point of presence (PoP) to meet-me-room (MMR) route monitoring and predictive/preventive maintenance to raise SLAs without endless duplicative runs. Inside the rack, intelligent cabling that tracks every connection, deters tampering and alerts on patch changes is seen as critical for security and compliance. However, cost remains the barrier, with most enterprises not adopting, and adoption likely to surge if competitively priced, and locally made systems become available.
Supply‑chain readiness varies by component; imported, customised long‑lead items such as generators and chillers can take 12–15 months, while locally manufactured alternatives may arrive in roughly four months, and OEM‑backed chips and servers remain a global constraint beyond domestic control. Despite rapid industry growth, over half of critical inputs such as chillers, generators and IT servers are still imported, complicating sub‑six‑month delivery targets even after contracts are signed, while cabling demand at a 30 MW site can span 1,800–2,000 racks, with approximately 24 copper nodes each, rising optical speeds from 10G toward 150/200/400G, denser protected risers and buyer whitelists that fix vendor choices. Standardised “run‑of‑the‑mill” designs can compress procurement timelines, but future‑ready innovations, liquid and immersion cooling, direct‑to‑chip and two‑phase require close OEM co‑design and inherently add time, even as Indian execution has demonstrated speed with 30 MW builds in nine months, substations in 9-12 months and generators in three months.
Pre‑build approvals
Pre‑construction approvals – especially environmental clearances, remain a major bottleneck, with some projects failing to launch, making proactive licensing teams essential during the land‑to‑readiness window. India’s federal structure creates implementation gaps between central policies and state execution, even as environmental safeguards rightly demand diligence. Single‑window “corridor” models are proposed and already emerging in hubs such as the Navi Mumbai Airport area and Nagpur, streamlining permissions for power, water, building and fire to accelerate data centre timelines.
Becoming a regional leader
In Europe, there is hardly any power available and the data centre industry faces limitations in growth because of that. The United States of America (the US) is almost following the same track now, where power is becoming a major issue. India’s construction and operating costs are significantly lower than those in the US, Europe and Japan, with sufficient land and power availability strengthening its competitiveness. As other regions such as the Middle East and Africa mobilise, accelerating execution on both private and public fronts becomes critical to retain the advantage.
Although India meets core prerequisites to position itself as a global hub, it is still small on a global scale; yet it is poised for outsized growth. However, localisation mandates remain a material risk that must be addressed through industry advocacy and collaboration with global intermediaries to unlock sustained international demand at scale. A cleaner, more uniform approval process and a common national data centre policy are encouraged, but not at the expense of strict enforcement on data privacy, environmental safeguards and sustainability. Robust regulation is viewed as essential to prevent harmful overexpansion and to build long‑term trust with global customers, even if it entails longer timelines to comply with.
A credible path to regional leadership begins with aligning capacity to four distinct demand segments including edge, enterprise, cloud and AI, while designing infrastructure that remains relevant over the next few years to avoid obsolescence after the first utilisation cycle.
Edge deployments will progressively shift from connectivity and caching, likely just a small fraction of use, to high‑density GPU inferencing between devices and local sites, making retrofitted mobile switching centre (MSC) locations and new edge facilities’ long‑term assets, even if their immediate role appears limited today.
Enterprise demand is surging as regulators press banks, exchanges and insurers to upgrade, yielding 10–50 MW request for proposal with resilience requirements that diverge from typical colocation designs and often push toward Tier 4 specifications, necessitating purpose‑built architectures.
Cloud growth is expected to continue, but shifting policy and geopolitics are accelerating momentum toward sovereign and local cloud, with new offerings and potential mandates that could require Indian start-ups and major government contracts to run on Indian platforms, thereby reshaping siting strategies and the broader ecosystem.
At the same time, the AI wave is redefining scale as multi‑gigawatt cluster announcements make even 1 GW appear small, exposing gaps between ambition and grid availability and highlighting a stark cost contrast between historical US build economics and lower Indian cost structures, which will still demand execution models capable of delivering accelerated capacity at scale. This trajectory extends into sovereign AI and inferencing workloads, making it imperative to decide where such compute will ultimately reside and to configure infrastructure that can host it without compromising performance or resilience.
Sustained leadership also hinges on competitive business models and a re‑examination of what “long term” means in an era of rapid disruption, where flexible designs enable shifts between technologies with lower incremental investment, while protecting return on investment on committed capital expenditure (capex). The broader market context shows a handful of countries dominating installed capacity (roughly half in the US, around a fifth in China, followed by Japan, South Korea and the European Union), with India advancing toward roughly 3 GW, yet the larger opportunity lies in serving many other nations that lack infrastructure, technical depth and the scale for viable financial models. As data centres become mission‑critical national infrastructure and the digital revolution outpaces previous industrial transitions, more countries seek AI‑enabled facilities that they cannot feasibly build or operate on their own timelines, creating a window for India to offer solutions at cost and speed others may struggle to match.
Finally, policy must explore granting “embassy‑like” status to data centres so that foreign data can be hosted in India without privacy or security concerns, positioning the country to serve not only the domestic demand but also international storage and compute.
