India’s data centre growth has brought foundational questions into focus. The debate has moved well beyond capacity targets, focusing instead on where new facilities will be built, how they will handle the thermal demands of artificial intelligence (AI) workloads and how they will be sustainably powered. Demand from hyperscalers, cloud providers and AI-first enterprises is rewriting capacity road maps, while energy availability, liquid cooling at scale, data sovereignty mandates and the imperative of renewable-led operations are reshaping how facilities are designed, located and financed. In this forum, sector leaders share their assessments of where the industry stands today and what will define its next growth phase. The following perspectives are drawn from recent articles featured in tele.net, a sister publication of Indian Infrastructure. Edited excerpts…
Sharad Agarwal
Sify Infinit Spaces has an operational capacity of roughly 200 MW across 15 data centres in seven cities of India. It focuses on three distinct market tiers – hyperscale AI campus deployments, enterprise and edge – and currently holds a 15 per cent market share. The intent is to cater to the full spectrum of demand as it evolves.
There is a coexistence between what has already been built and what is coming. In an environment where capacity doubles every year or two, what exists today begins to look increasingly small against what will arrive. The real question is whether the infrastructure being built right now is suited for the next generation of graphics processing units (GPUs), and whether what we plan today will remain relevant for 10-20 years. Those are open questions we are actively working through. For older facilities built 25 years ago, we are evaluating whether to continue with them, or exit or refurbish them to handle high-density cloud workloads, if not AI workloads directly. Liquid cooling, theoretically, can be retrofitted into older facilities, though they were never designed for it.
Mumbai has a sanctioned data centre demand of 5.6 GW from Maharashtra State Electricity Distribution Company Limited alone. This is far too large to be served from one metropolitan area. NVIDIA’s Vera Rubin reference architecture requires 250 acres of land for 150-200 MW of capacity. This is not something that can be deployed in Mumbai or any dense metro. It must go inland. Compare this with the US, where data centre capacity was historically concentrated in northern Virginia and Washington State, but is now spreading to Ohio, Indiana, Wisconsin, Arizona, Nevada, Utah and Wyoming. India has four times the US population and its own demand curve to serve. The constraint will be energy. Bringing sufficient, reliable energy to these inland locations is the central problem to solve over the next decade and beyond.
Almost all hyperscalers we host in India are US-headquartered. They self-build in the US. Yet, in recent quarters, 7 GW worth of orders have been placed on co-location operators. The ability to execute, to actually bring infrastructure to the market during this transformational period, is what differentiates participants.
At present, there are somewhere between 20 and 25 significant data centre operators in India. As the industry moves along the maturity curve, consolidation tends to follow. This is a capital-heavy business, and capital dynamics drive consolidation. It will happen, likely more towards the later stages of the maturity curve.
The data centre industry has adapted meaningfully towards sustainability. The lowest-hanging fruit was renewable energy, and I would say nearly every operator has begun pursuing power purchase agreements (PPAs) for solar and wind. We have more than 300 MW of PPAs in place, of which approximately 128 MW is already wheeling in. As loads increase, we are ensuring that 60 per cent of our energy comes from renewable sources. Customers, particularly hyperscalers, want maximum renewable content from us, because it reduces what they must offset through virtual PPAs.
On water usage, the industry decided long ago to avoid water-based chillers and evaporative cooling. As a result, water usage effectiveness is not even a measurement parameter in India. We are also working with the Indian Green Building Council (IGBC) to ensure all our facilities achieve Gold or Platinum certification on energy consumption. We received an ESG rating of 79 out of 100 from a rating agency and are working to improve that score meaningfully. A 100 per cent green data centre is possible, but it carries a premium. The moment for it will arrive as hyperscaler and customer expectations continue to rise.
“A 100 per cent green data centre is possible, but it carries a premium. The moment for it will arrive as hyperscaler and customer expectations continue to rise.” Sharad Agarwal
Ashish Arora
Nxtra is the digital infrastructure partner for India’s AI journey, combining AI‑ready, sustainable data centre capacity with the power of Airtel’s pan‑India networks, global subsea systems, and cloud and cybersecurity capabilities.
As India positions itself as a global hub for data centres and AI infrastructure, our ambition is to lean in and scale with discipline. We are investing to move from roughly 300 MW today to 1 GW over the next few years, and because of Airtel’s strengths in connectivity, subsea capacity and security, we can do this in a way that is highly differentiated, not just in racks and MWs, but as an integrated platform that helps Indian and global customers build and run AI workloads reliably, securely and at scale. Sustainability is central to our growth strategy. With our net zero by 2031 commitment, RE100 membership, SBTi-aligned targets and the deployment of AI to optimise energy consumption across our facilities, we are ensuring that as we scale to power India’s AI future, we do so responsibly. What really differentiates us is the integration. Nxtra is deeply woven into Airtel’s fibre, 5G and edge footprint, subsea cables and cable landing stations, and our cloud and cybersecurity platforms. That means we can offer customers a single, seamless fabric, from the device at the edge and the data centre core to the cloud, with low‑latency connectivity, secure interconnects and end‑to‑end visibility.
At Nxtra, we have built one of India’s most advanced and sustainable data centre networks, designed to meet the evolving needs of enterprises, hyperscalers and government as they embrace AI at scale. We are deeply focused on creating a future‑ready, AI‑optimised infrastructure that is intelligent through innovative design, sustainable by deliberate choice and deeply integrated with Airtel’s industry-leading digital networks. On AI‑ready infrastructure and power density, we are significantly stepping up investments to scale. Our new campuses are being purpose-built as hyperscale, high‑density facilities with robust power and cooling infrastructure to support AI and machine learning (ML) workloads, large model training and inference at scale. All our upcoming hyperscale facilities are being built to a minimum IGBC or LEED Gold green building standard, because energy and resource efficiency should be baked into design from day one.
We are also the first data centre company in India to deploy AI within our operations, enabling predictive maintenance, energy efficiency automation and optimised capital utilisation, initially at our Chennai facility and now being rolled out across all core facilities.
What really differentiates us is our connectivity strength. Nxtra is tightly woven into Airtel’s pan‑India fibre, 5G and edge network, as well as its global submarine cable systems, giving customers low‑latency access from edge to core to the cloud. This allows us to offer AI‑ready data centres that are not just high‑density boxes but part of a unified digital infrastructure fabric, from over 120 edge sites to large core campuses, that can move and process data securely and at scale for AI workloads across India and beyond.
Finally, we are doing all this in a way that is responsible and future‑proof. We have committed to becoming net zero by 2031 and are significantly increasing our share of renewable energy through long‑term green power agreements and captive projects. In FY 2024-25, we sourced 49 per cent of the total electricity consumption across our core data centres from renewable sources and became the first data centre company in India to join the global RE100 initiative, committing to 100 per cent renewable electricity across all our operations.
Energy availability is one of the most decisive factors for data centres today. The next wave of AI and cloud growth will be constrained not by demand, but by access to reliable, low-carbon power. That is why we are investing ahead of the curve in firm, round-the-clock renewable energy through long-term partnerships and collaborations with key industry experts to secure scalable green power for our facilities. Coupled with our net zero by 2031 commitment and green-by-design approach, this gives our customers confidence that their growth in India is aligned with their global ESG and decarbonisation goals.
From a data sovereignty perspective, India is already moving towards a more localisation-driven regime, especially for critical and sensitive workloads. We see this as a structural tailwind for India’s data centre ecosystem, because it creates a strong case for building capacity closer to the end-user, backed by domestic networks and cloud connectivity. At Nxtra, our footprint of large core data centres and distributed edge facilities across multiple Indian cities is designed to support this, enabling global and domestic customers to host, process and analyse data within the jurisdiction, while still integrating seamlessly into their global architectures.
Further, selecting a data centre’s location entails checking multiple aspects, but primarily revolves around three major factors: the availability of quality power, rich connectivity and long‑term resilience. These are exactly what Nxtra is built around. Our data centres sit on very stable grids with large, scalable power, built‑in redundancy and growing renewable mix, engineered for high‑density, energy‑efficient, sustainable operations. They are plugged into multiple fibre routes with direct access to subsea landings, national long‑distance networks and key urban aggregation points so that customers get ultra‑low‑latency, highly resilient connectivity.
Our campuses are also designed to keep expanding safely over time, with cooling designed for local climate and responsible water use, so that our facilities can support the next wave of AI and cloud growth for many years, not just the next few quarters. In short, Nxtra locations are picked to deliver intensity, reliability and sustainability at scale.
In terms of cooling, we use a combination of modern air‑based and liquid‑assisted cooling systems, depending on the workload and the climate. The idea is to remove heat efficiently and keep equipment at safe and stable temperatures while ensuring the lowest energy and water footprint. For this, we ensure separate containment of hot and cold aisles, efficient chillers and next-gen advanced cooling close to high‑density servers. We ensure to keep our infrastructure cool and reliable, while steadily improving our energy and water efficiency in line with global best practices.
Aligned with strategic growth plans, we are supporting both high-density training and distributed inference through a mix of hyperscale campuses and edge data centres. We are focusing on developing purpose-built, AI-ready data centres capable of handling 70-80 kW per rack, utilising liquid cooling to manage high-density workloads. We have also partnered with Google to establish a 1 GW AI-focused hub in Visakhapatnam, which is specifically designed to handle the massive, high-density requirements of training.
India is incredibly well positioned in the global data centre ecosystem as one of the world’s largest and fastest growing digital economies, which generates a disproportionate share of global data. We have a very strong policy push around Digital India, data localisation and now the IndiaAI Mission. All of this is creating a structural demand for high quality, AI‑ready data centre capacity here at home. At the same time, global hyperscalers and investors are looking at India as a long‑term, strategic market, not just to serve Indian users, but increasingly as part of their global footprint. This is being seen in the pace of investments in new campuses coming up at hubs like Mumbai, Chennai, Visakhapatnam, Hyderabad and Noida, and the growing depth of subsea connectivity into the country.
Policy support, such as tax benefits up to 2047 for infrastructure serving global demand, introduced in the Union Budget 2026, is further strengthening India’s proposition as a global data hub and encouraging sustained investor interest. India brings together both scale and skill. With a deep talent pool to build and operate next-gen infrastructure, this combination of talent and policy support creates a strong foundation for growth, with continued investor participation key to accelerating capacity build-out.
Despite this, the current share of global data centre capacity remains relatively small compared to the data generated, indicating a long runway for growth. By effectively executing on our power, sustainability and talent strategies, I believe India can move from being an underserved market to being one of the key global nodes for digital and AI infrastructure over the next decade. Our ambition at Nxtra is to lean into this opportunity and build world‑class, sustainable infrastructure that not only serves Indian demand but also integrates India more deeply into the global digital fabric.
“Energy availability is one of the most decisive factors for data centres today. The next wave of AI and cloud growth will be constrained not by demand, but by access to reliable, low-carbon power.” Ashish Arora
Surajit Chatterjee
CapitaLand Data Centre is now in its sixth year in India, building 245 MW of capacity across Mumbai, Chennai, Hyderabad and Bengaluru. Mumbai is already completely leased out. Our aim is to reach 600-700 MW over the next three to five years, and we have chosen our markets carefully. From a go-to-market standpoint, we focus approximately 70 per cent on hyperscalers and 30 per cent on enterprise, with specific emphasis on banking, insurance and e-commerce. In the hyperscale space, we are increasingly focused on Build-to-Adapt – our approach to built-to-suit – and on liquid cooling as a technology imperative.
Design is one of the industry’s most dynamic challenges. Customer requirements evolve rapidly, and the critical question for every operator is whether they have the ground-level capability to adapt as those requirements shift. Power approvals take time, build timelines are long and both are running in parallel. What is encouraging is that original equipment manufacturers (OEMs), operators, design consultants and the government are increasingly at the same table, working collectively to make delivery faster and more viable.
There was a period of concern that if hyperscaler self-builds continued to grow rapidly, co-location growth would suffer. However, these two are running in parallel. Self-builds serve a different purpose and require longer lead times to go live. Co-location is faster to execute. Workloads are being divided accordingly. As a global data centre player with presence in Singapore, China and Europe, we have existing hyperscaler relationships that form our India strategy. But the enterprise space, particularly banking, insurance, e-commerce and IT, given the volume of AI deployment and content creation happening behind the scenes, is a segment that has real potential to scale.
Every operator will eventually encounter the natural limits of metropolitan infrastructure. The Visakhapatnam announcement is an early signal of how new geographies are opening up, and we expect this trend to continue. A greenfield project in India today, from land acquisition to power sourcing, takes 26-30 months. Greater coordination between operators, utilities and the government is beginning to push that in the right direction. Operators who will define the next phase are those who plan early, move with speed and adopt the right technologies ahead of demand.
Consolidation will happen, but it will not have a single flavour. There are two dimensions: how investors choose to consolidate – spreading across markets or going deeper in fewer locations – and where operators elect to expand next. With roughly 22-25 significant players today, each working within a defined scope, the first meaningful phase of consolidation is likely to become visible around 2030.
Our approach is twofold: fulfilling client sustainability requirements through PPAs and captive set-ups, while investing directly in the renewables sector to generate a revenue model from it. Two years ago, we commissioned our first solar power plant in Chennai with 25 MW, serving our office and commercial assets. For us, sustainability is also not only about serving the data centre industry. It is about creating an infrastructure layer that adds value across our multiple asset classes in India, spanning data centres, the commercial space and logistics.
We are here to help shape the infrastructure foundation that India’s digital economy will run on; through AI-ready design, liquid cooling at scale, and a Build-to-Adapt approach that keeps pace with how technology and customer requirements continue to evolve. The opportunity ahead is significant. So is the responsibility that comes with it.
“For us, sustainability is not only about serving the data centre industry. It is also about creating an infrastructure layer that adds value across our multiple asset classes in India, spanning data centres, the commercial space and logistics.” Surajit Chatterjee
Manoj Paul
Megawatts, long been used to describe data centre capacity in India, measure power, not business value. The focus needs to shift away from megawatts to what you can actually do inside a data centre. For enterprises today, the metrics that matter are interconnection density, ecosystem depth, cloud on-ramp availability, latency performance and uptime reliability. The reason is straightforward – power is an incomplete metric for diverse, interconnection-driven workloads. Equinix’s core business is to serve our 10,500 customers all over the world, and to support digital services that matter to everyone’s daily lives. Measuring data centre capacity solely on power introduces a structural bias that overlooks where value is actually created.
What enterprises actually need is proximity to the right partners: cloud providers, network service providers, internet exchanges and industry peers, all within the same facility and, most importantly, flexibility to grow and reduce their IT infrastructure and interconnection to align with changing business needs, with minimal entry and exit barriers along with cost optimisation.
AI does not just need compute, it needs connectivity. As AI moves from training to inference at scale, workloads need to run close to users, close to data, and with access to the right ecosystem partners. Interconnection is the nervous system that makes it possible. India leads globally in multicloud adoption, with 83 per cent of enterprises using multicloud tools and connecting to an average of 37 distinct cloud and software as service providers. As these enterprises deploy AI inference in production, the demand for low-latency, private, carrier-neutral interconnection will intensify significantly.
Equinix’s leverage lies in its ecosystem model. Our Mumbai campus hosts the largest concentration of cloud on-ramps, telecom providers, over-the-top platforms, internet exchanges and financial ecosystem participants in India. Through Equinix Fabric, a software-defined interconnection service, customers can dynamically connect to multiple clouds and partners on demand via software-defined paths. In India, we are now extending this ecosystem from Mumbai to Chennai, creating the foundation for a national digital backbone.
MB3, our fourth IBX in Mumbai, is one of the largest retail data centres in the city, with an initial investment of over $95 million, more than 1,370 cabinets scaling to 5,475 across three phases, and advanced liquid cooling capability. CN1 in Chennai is our first AI-ready facility in southern India, built on a 6-acre site in Siruseri with an initial investment of $69 million and 800 cabinets scaling to 4,250. Today, our Mumbai (MB1, MB2, MB3 and MB4) and Chennai (CN1) campuses represent over $365 million in investment and a combined capacity of more than 12,000 cabinets at full build-out. Both facilities go well beyond traditional co-location. They are engineered for high-density AI workloads from day one (not retrofitted) with liquid cooling, fault-tolerant 99.9999 per cent uptime architecture and deep interconnection built in as standard.
Our data centres in India also achieve 100 per cent renewables coverage, reflecting sustainability as a design priority, not an afterthought. Our group captive solar project in Yavatmal, Maharashtra, commissioned in November 2025 under our long-term PPA with CleanMax, delivers 26.4 MWp of solar capacity, helping reduce our carbon footprint and marking a significant milestone in our sustainability journey.
AI inference and distributed workloads are pushing data centre deployment away from purely centralised models towards distributed, interconnected infrastructure. Compute is moving closer to where end-users and data sources are likely to be found. High-performance data centres are being located near population centres, enabling the low-latency connectivity inference workloads demand while simultaneously serving as dense interconnection hubs. Vendor neutrality is also becoming a strategic requirement. Deploying on a carrier-neutral, ecosystem-rich platform lets enterprises compose the right mix of compute, network and cloud for each workload.
Similarly, as AI workloads intensify, power density is rising from 5-10 kW per rack to 100+ kW per rack, requiring liquid cooling systems to manage heat from GPUs that can draw up to 1.2 kW each. Air cooling cannot manage that level of heat dissipation at scale. Liquid cooling is no longer a niche capability; instead, it is becoming a baseline requirement for any genuinely AI-ready facility. And enterprises have a mixed requirement – a few liquid-cooled cabinets requiring 100+ kW per cabinet, some central processing unit-based compute requiring 6-10 kW racks, and, in fact, these form a large part of their set-up, and then a few network cabinets, which need 3-4 kW per rack. The challenge is to meet this varied requirement in the same co-location hall.
Equinix’s approach in India reflects this new evolving requirement of customers. CN1 in Chennai is purpose-built to accommodate liquid cooling for high-density, compute-intensive workloads. MB3 in Mumbai is designed from the ground up with advanced liquid-cooling capabilities. Both facilities are engineered for the full thermal stack, not only supporting liquid cooling but also other methods like in-row cooling to support high power requiring cabinets, which do not support liquid cooling, because even the densest AI deployments include an air-cooled component. GPUs, storage and networking have different cooling requirements.
On grid infrastructure, Equinix is pairing capacity growth with renewable energy investment. Our captive solar plant in Yavatmal, Maharashtra (commissioned in November 2025 under our PPA with CleanMax), generates approximately 41.4 million kWh of clean energy annually, reducing carbon emissions by over 30,000 tonnes of CO2 per year. MB3 operates with 100 per cent renewable energy coverage, consistent with more than 250 Equinix data centres worldwide.
Meanwhile, sustainability has moved from a reporting obligation to a procurement criterion, especially in AI infrastructure decisions. We have seen customers deploying high-performance AI training environments make 100 per cent certified green energy a hard requirement, explicitly citing their corporate sustainability frameworks. In India, our sustainability investments are shaping the physical design of our infrastructure. Advanced water-optimised cooling systems and air-cooled chillers are deployed across our IBX facilities. We also provide customers with granular environmental transparency through our Customer Water Report, giving enterprises allocated water withdrawal metrics for their specific deployments.
Union Budget 2026’s proposed tax holiday for global cloud providers, Digital Personal Data Protection Act data localisation requirements, and the IndiaAI Mission’s deployment of over 34,000 GPUs are creating one of Asia-Pacific’s most supportive investment environments. India now ranks second in the region by operational capacity at 1.6 GW and third by development pipeline, with 3.1 GW under construction or planned.
The geography of growth is expanding beyond Mumbai. Chennai is emerging as India’s AI and connectivity hub, anchored by its proximity to subsea cable landing stations linking the Asia-Pacific with Europe, the Middle East and Africa. Hyderabad, Delhi-NCR and Pune are attracting serious investment. Mumbai will surpass 1 GW of operational capacity by the end of 2026, but the more significant story is India’s emergence as a genuinely national digital infrastructure market. The defining trends of the next two to three years are expected to be: AI inference migrating from centralised cloud to distributed, interconnection-rich facilities; continued enterprise shift from in-house data centres to premium co-location; and sustainability-linked procurement becoming standard in enterprise requests for proposal.
“Sustainability has moved from a reporting obligation to a procurement criterion, especially in AI infrastructure decisions.” Manoj Paul
