For India’s fast-growing data-centre companies, the difference between winning and losing new business comes down to two basic things: how much land they hold and how quickly they can get power to it. That was one of the central points JM Financial made in its recent report, which laid out three key debates shaping the country’s data-centre industry.
The brokerage’s note pointed to TCS as an example of how fast the sector is moving. The company’s HyperVault unit is targeting 1 gigawatt (GW) of capacity over the next five to seven years, at an estimated capex of about USD 1 billion for every 150 megawatts (MW) built. As per the report, this scale of spending reflected how central data centres have become in India’s AI ambitions and how much money is now chasing the space.
Land and power hold the key
According to the report, access to land parcels with reliable power supply has become the single biggest factor separating serious players from the rest. As AI workloads push up rack densities and power needs, companies that already have land banks with grid connectivity, or ones that have executed large data centre projects before, are likely to have an edge. JM Financial said such companies would probably be able to cut deployment timelines meaningfully compared with newer entrants starting from scratch.
The report cited TCS‘s recent land acquisitions as an example of this trend. Media reports cited in the note suggested TCS had acquired two land parcels, one in Vizag and another in Pune, to house infrastructure linked to its partnership with OpenAI. Andhra Pradesh had reportedly allotted 200 acres in the Anakapalli district, while TCS separately bought an 88-acre parcel from Hemisphere Properties for Rs 650 crore, the report noted.
Colocation versus full-stack, and who bears the risk
As per the report, the second theme was whether data centre operators should stick to colocation, which means essentially renting out ready infrastructure, or move up the chain into full-stack offerings that also involve software, compute, and AI models. JM Financial noted that colocation remains a relatively stable and predictable business, supported by long-term contracts and steady infrastructure use. Full-stack offerings, on the other hand, can improve revenue per MW and make customers stickier, but they also bring higher technology risk, greater capital intensity, and more execution complexity, the report said.
This ties into a related question the brokerage raised: whether companies should build capacity themselves (capex) or lease it from colocation providers (opex). JM Financial said hyperscalers are pursuing both routes depending on their needs, while service providers are increasingly leaning toward the opex model to avoid large upfront spending. The capex route, by contrast, gives companies more control over infrastructure that they consider strategically important, the report noted.
TCS and HCLTech are taking different paths
According to the report, TCS’s HyperVault is being built as a passive, AI-ready infrastructure business, with capacity meant to be leased out to hyperscalers, deep-tech firms doing AI training and inferencing in India, or Indian enterprises looking for private cloud options. The unit is backed by a partnership with TPG, under which both firms have committed to invest up to Rs 18,000 crore, with TPG’s share reportedly going up to Rs 8,820 crore for a shareholding of between 27.5% and 49%, JM Financial said.
HCLTech, meanwhile, has taken a different approach. The company announced plans in July to set up a 50 MW full-stack AI data centre, with an initial investment of up to Rs 3,500 crore. As per the report, HCLTech’s stated view was that the bigger opportunity lay in owning the entire stack, including data centres, compute, and the AI models themselves, rather than simply renting out capacity to others.
Sustainability metrics gaining ground
Beyond land, power, and business models, the report said investors are also paying closer attention to sustainability. Power Usage Effectiveness (PUE), which measures how efficiently a data centre uses power for its actual computing needs versus cooling and other overheads, remains a key benchmark. But JM Financial noted that access to renewable energy is fast becoming just as important, with companies increasingly expected to secure clean power through dedicated procurement arrangements.
The report’s overall message was that Indian data centres are still an early-stage business, but one with a lot of room to run. How companies fare from here will come down to a few basics: whether they have land and power lined up, and whether they choose to build, own or lease their way into the market, the report noted.