India’s data centre ecosystem is entering a new phase. What used to be a market centred on basic cloud hosting has evolved into one where power availability, cooling efficiency and regional reach matter as much as capacity itself. Operators are no longer just adding more buildings – they are redesigning how those buildings work, from how much power a single rack can draw to how quickly heat can be removed when workloads spike. At the same time, the geography of demand is widening. Users outside the metros now expect the same responsiveness as those in major cities, pushing operators to build smaller, distributed sites closer to end-customers. These shifts place new pressure on engineering choices, vendor readiness and long-term planning. At tele.net’s eighth annual conference on “Data Centres in India”, industry leaders shared their views on the trends shaping the data centre industry, challenges being faced and sector outlook. Key takeaways from the discussion…
Key trends shaping the industry
India’s data centre market is undergoing rapid scaling as workloads shift from traditional cloud computing to artificial intelligence (AI) and graphics processing unit (GPU)-intensive environments. AI, high-performance computing and media-driven use cases are pushing requirements into much higher-density territory, with deployments reaching 120-140 kW per rack in some facilities. Adoption is not uniform across the ecosystem. Enterprises continue to take a measured approach, preferring gradual increases in density and hybrid models instead of immediate transitions to extreme-density racks. In contrast, hyperscalers, large media workloads and research-driven environments remain the primary adopters of high-density configurations, consistently requiring concentrated GPU clusters and higher sustained power delivery.
Cost remains a central variable shaping market behaviour. GPU acquisition continues to be expensive for most end-users, creating a strong preference for GPU-as-a-service, bare-metal GPU offerings and shared GPU pools delivered by operators. Government programmes are also contributing pooled GPU capacity, numbering in the tens of thousands of units, to support departmental and public sector workloads, though these pools remain considerably smaller than those in larger global markets.
Market expansion indicators reinforce this momentum. There has been a fivefold growth in India’s data centre capacity over the past three years, with industry projections suggesting an eightfold expansion over the next five years. Over the next 12-18 months, committed hyperscale investments and new cloud announcements are projected to accelerate the number of visible high-density deployments across new campuses.
Workload shifts
The concentration of hyperscale activity in cities such as Mumbai, Chennai, Hyderabad and Delhi has generated operational pressures like increased localised power demand, urban heat-island effects and transmission losses from long-distance power transfer. Therefore, India’s data centre footprint is expanding beyond traditional metropolitan concentrations into a more distributed national landscape.
Operators are deploying facilities in Tier 2 and Tier 3 cities to meet growing local digital demand and latency-sensitive transaction requirements. This regional expansion includes conventional edge sites. Edge facilities function as regional hubs rather than smaller versions of core campuses. Their role is to process latency-sensitive transactions locally while remaining tightly linked to larger facilities for heavier compute or storage tasks. By bringing processing closer to users in B-class cities, operators aim to reduce delays or failures caused by fibre latency and improve the responsiveness of digital services. New and upcoming deployments in locations such as Lucknow, Patna, Guwahati, Jaipur, Ahmedabad, Ludhiana and Chandigarh reflect this strategic diversification. However, operating these distributed sites comes with its own set of requirements. Smaller edge facilities need different staffing patterns, remote management capabilities and standardised operational playbooks to maintain reliability. They also rely on tailored power and cooling set-ups, typically lower-density, modular configurations that differ from those in hyperscale campuses.
The readiness of India’s data centre infrastructure varies significantly, shaped by the generational divide between legacy facilities and modern, purpose-built campuses. Older data centres, many originally designed for low-density IT loads, often face structural and mechanical constraints that limit safe operation to around 6-10 kW per rack. Handling high density levels is not just a matter of putting in bigger racks; it changes the way the entire hall has to be prepared. This means upgrading coolant distribution units, reinforcing floors to take higher loads, reworking mechanical pathways and putting in new power delivery frameworks that can carry sustained heavy draw.
Infrastructure and engineering readiness
Liquid cooling has emerged as essential for facilitating high density levels. Even advanced deployments remain hybrid in practice. For instance, practical liquid coverage reaches approximately 80 per cent of a hall, with the remaining reliant on traditional air-cooling approaches. Several operators already deploy direct-to-chip (D2C) liquid-cooling systems in production settings, using largely India-manufactured mechanical and cabling components. Modern facilities designed for hyperscale clusters report per-rack capacities of up to 120 kW, with sustained operating averages exceeding 80 kW where required.
Further, reduced dependence on raised floors and large air-handling volumes allows for lower column heights and more compact vertical layouts. This, in turn, enables additional usable floors within the same building envelope, improving spatial efficiency and creating up to 20 per cent more deployable area in comparison with traditional air-cooled configurations. Such planning efficiencies directly influence capex outcomes, particularly in high-density environments where thermal design and structural load considerations translate into measurable economic gains.
India’s construction economics further strengthen its position as a competitive data centre destination. Delivered cost per MW, inclusive of land, substations, mechanical, electrical and plumbing (MEP) infrastructure and low-side systems, remains significantly below many global benchmarks, approximately $5 million-$6 million per MW compared to global averages closer to $10 million. Greenfield development timelines also benefit from the ability to pre-position land and power infrastructure – operators with substantial land banks in key corridors can deliver new campuses within 24-30 months, provided fibre routes and transmission access are already accessible.
Power, sustainability and supply chain ecosystem
India’s modern hyperscale data centre environments increasingly rely on transmission-level power integration to support rising density and uninterrupted operations. Large campuses now draw supply directly from high voltage grid infrastructure, typically through on-campus gas insulated substations (GISs), which eliminates dependency on distribution networks and ensures more stable, large-volume power delivery. Such approaches reduce mechanical stress, preserve system life and maintain operational efficiency during minor grid fluctuations. Operating directly on the grid enables campuses to avoid load shedding entirely and leverage generator synchronisation systems that stabilise minor frequency fluctuations without initiating full transitions to backup power. Some facilities further reduce electrical conversion losses by adopting next-generation data centre power models at the rack level.
Supply chain reliability remains a critical determinant of project timelines and operational readiness. Operators highlighted dependencies on imported items, such as coolant distribution units, GISs and select high voltage switchgear, which can create delivery bottlenecks. To manage this risk, organisations adopt empanelled partner models that include consultants, core-and-shell construction firms, MEP providers, integrators and passive infrastructure suppliers. Early procurement, thorough technical requests for proposal and multivendor strategies are essential to prevent cascading delays during construction.
Energy economics play a central role in shaping long-term operational strategy. Operators consistently report renewable power usage in the mid-20 per cent range, not due to lack of availability but because state regulations cap the allowable share of green energy in individual facilities. India’s favourable solar conditions, amounting to approximately 330 sunny days per year, significantly improve the cost basis for renewable integration. Longer-term renewable penetration, however, depends on the affordability of long-duration energy storage systems, which will eventually enable more extensive reliance on intermittent sources.
Cooling choices also influence power efficiency. Conventional air-cooled facilities operate with design power usage effectiveness (PUEs) in the 1.4-1.5 range, while D2C liquid-cooled environments demonstrate lower design PUEs near 1.3-1.35. This means that as GPU-centric workloads drive higher rack power densities, liquid-cooling adoption becomes directly tied to achieving sustainable efficiency levels.
At the same time, there is a growing domestic manufacturing footprint. Many original equipment manufacturers now produce mechanical, electrical and cabling components in India, supporting liquid cooling deployments with locally manufactured parts, except for a few specialised items. There is a willingness to evaluate emerging Indian suppliers, particularly those offering demonstrable PUE improvements or cost advantages.
Key constraints
India’s data centre regulatory environment has advanced significantly, with multiple states formally identifying data centres as an independent infrastructure class. This recognition has streamlined financing models and provided greater clarity for large-scale campus planning. However, operators continue to navigate constraints that influence build timelines and operational strategy. Renewable energy caps in several states limit the proportion of green power that data centres can consume, even when solar availability is abundant. Approvals tied to high voltage infrastructure, such as transmission interfacing, substation integration and power evacuation routes, remain lengthy and require multi-agency coordination.
Infrastructure gaps extend beyond power. There is an absence of a pan-India, low-latency fibre backbone. Reliance on coastal landing stations and limited inland redundancy increases latency variability and restricts the placement of hyperscale capacity in emerging regions. Broader access to long-haul inland fibre and ring-based fault-tolerant paths is viewed as critical for national-scale workload distribution. There is also a systemic gap in domestic digital platforms. While physical data centre expansion has accelerated, India’s ecosystem still lacks indigenous operating systems, storage platforms and enterprise-grade applications at scale. There is a need to strengthen local application development and create national digital stacks that reduce dependence on foreign cloud and software ecosystems. These capability gaps are seen as structural barriers that must be addressed alongside physical infrastructure growth.
In sum
The long-term outlook for India’s data centre sector reflects sustained growth driven by AI adoption, high-density compute requirements and expanding regional digital ecosystems. It is anticipated that rack densities will continue rising. These expectations already influence current design decisions, prompting the integration of advanced liquid-cooling infrastructure, higher slab-load tolerances and transmission-level power planning into new builds. Future facilities are being engineered with assumptions that next-generation workloads will concentrate even more compute per square foot than today’s GPU clusters.
Geographic redistribution will intensify over the coming years. As digital consumption grows in Tier 2 and Tier 3 cities, latency-sensitive workloads will shift closer to end-users, supported by edge nodes tightly integrated with core hyperscale campuses. This approach reduces overconcentration in metro hubs, alleviates localised grid pressure and improves service responsiveness for financial, consumer and public sector applications. Furthermore, energy strategies are expected to undergo parallel evolution. Once long-duration energy storage becomes cost-effective, renewable energy penetration will rise beyond existing regulatory ceilings, enabling more stable and higher-volume use of green power. This shift will strengthen the operational economics of high-density environments while supporting national sustainability goals.
Domestic manufacturing is also positioned for deeper integration. As more mechanical, electrical and cooling equipment is produced locally, operators expect shorter lead times, reduced exposure to geopolitical risk and tighter alignment with hyperscaler specifications. The broader trajectory points towards a sector in which physical infrastructure growth, domestic technology development and rising AI workloads reinforce one another, shaping a long-term path of sustained national expansion.
