For nearly three decades, HCLTech built its business on a simple premise: sell services, not steel and concrete. The Noida-headquartered IT major managed data centres for global clients but steered clear of owning them, keeping its balance sheet light and its return ratios among the best in the Indian IT industry.
That changed this July, when HCLTech announced it would invest up to ₹3,500 crore to build AI-optimised data centres with capacity scaling to 50 megawatts—its first move into owning physical infrastructure.
The plan has since crystallised into a ₹14,257 crore first project: an AI data centre in Bhubaneswar, Odisha, developed in a tripartite partnership with AI startup Sarvam and the Odisha government, alongside a 5,000-seat global delivery centre in the capital.
In an interview with AIM, Sriram Hariharan, Corporate Vice President and Global Head of Strategy and Corporate Development at HCLTech, lays out the thinking behind the pivot, and pushes back on the idea that the buildout is a result of FOMO as competitors including Tata Consultancy Services (TCS) have already staked out their own gigawatt-scale ambitions.
From Managed Services to Owning the Asset
On what changed after decades of playing "the high end of the stack," Hariharan points to a shift in economics driven by AI.
A traditional colocation facility, he says, might generate roughly one unit of revenue per square foot or megawatt. A full-stack AI offering—infrastructure bundled with software and services—can generate up to 10–15x that, by his estimate.
According to him, clients no longer want to stitch together infrastructure, cloud, and application vendors themselves. With AI systems carrying what he calls "a little bit of a black box" quality, enterprises increasingly want a single provider that can take end-to-end ownership rather than point to one vendor for the data centre, another for the network, and a third for the model layer.
"From both ends... it's a great opportunity for somebody to make that investment and grow that as a full-stack offering," Hariharan says.
The Return-on-Equity Question
The pivot arrives as analysts and industry observers have already begun flagging capital allocation risk.
HCLTech's stock trades at a premium to its historical average, and some brokerages have warned of an unfavourable risk-reward balance given a muted growth outlook and an untested capital-intensive pivot into infrastructure. The company's own guidance for the year of 1– 4% constant-currency revenue growth was described by analysts as among the weakest full-year outlooks the company has issued in at least four years.
According to Hariharan, return on equity, taken in isolation, "doesn't mean much" unless a company can actually redeploy the capital it raises.
He says the core services business, which is still overwhelmingly asset-light, will continue to get the funding it needs, including capital expenditure tied to the Odisha delivery centre, and that shareholder payouts are not being compromised.
The bar for the data centre business is to generate an internal rate of return above HCLTech's cost of equity over the life of the investment, even if the annual ROE print looks lower than the legacy services business.
Not a TCS Clone
TCS has structured its AI data centre unit HyperVault with a $1 billion investment from private equity firm TPG and up to ₹18,000-crore commitment from both sides to build AI-ready data infrastructure. Pressed on whether HCLTech would structure the business the way, Hariharan confirms that HCLTech is also setting up investments through a dedicated subsidiary and step-down subsidiaries, funded for now entirely by the parent company.
But he was careful to distance the move from any direct comparison.
"We are not obviously mimicking another competitor in terms of the structure," he says, adding that no external financial or strategic partner, private equity or hyperscaler, has been finalised.
The financing mix disclosed so far leans on debt raised "at a good market rate," in India or offshore, once client commitments are locked in; equity partnerships, he says, are not yet on the table.
That marks a real divergence from TCS's playbook.
Where TCS has moved to build hyperscale, multi-tenant capacity with an external capital partner already locked in, HCLTech's 50 MW first phase is smaller, self-funded so far, and explicitly not intended as a co-location play.
"It is still a full-stack offering where we are able to add more value," Hariharan says. On whether HCLTech would rent out space to third parties on a standalone basis, he says,, "No, that is not the intent."
The Delivery Centre Pairing
The choice of Bhubaneswar was framed around a bundle of factors—power availability, a "progressive" state government, and, notably, talent.
HCLTech is pairing the data centre with a 5,000-seat global delivery centre in the same city, a decision Hariharan describes as central to the "full-stack" logic.
Demand, Hariharan says, will come from a mix of sources like sovereign and public-sector AI needs, large domestic enterprises with data-localisation requirements, and AI-native digital companies, though he declines to disclose any signed anchor clients, saying commitments would be announced in due course.
Odisha's government is contributing to the project as part of its financial assistance package.
The Competitive Landscape
HCLTech's move places it in the middle of a broader, and increasingly divided, response among India's largest IT services firms to the AI infrastructure boom.
TCS has gone furthest, moving fast to build HyperVault into a gigawatt-scale, externally capitalised platform aimed at hyperscalers and AI-driven organisations.
Infosys, however, has gone the other way entirely. The company has ruled out building its own data centres, choosing instead to prioritise AI services such as model selection, AI agent deployment, and system modernisation, to protect its cash position and margins from the depreciation and power costs that come with owning physical infrastructure.
Hariharan characterises the 50 MW build as "phase one," with no formal decision yet on a phase two, and says the near-term focus is execution given the long lead times involved in data-centre construction.
Also, he confirms HCLTech is still benchmarking chip vendors, including but not limited to AMD, its existing AI hardware partner, rather than committing to a single supplier. He adds that hiring for the data centre business will be gradual, starting with senior leadership and drawing talent from adjacent industries such as engineering, EPC, and even, he says, sectors with comparable "energy intensity" management challenges, citing naval submarine operations as one unusual talent pool the company is examining.
Whether the pivot pays off will hinge on execution over a multi-year build-out, and on whether HCLTech can convince a market that has grown used to judging Indian IT on capital-light return ratios that a full-stack, infrastructure-owning model can still clear the bar on shareholder returns.
Your reaction