TCS' $6.5 billion gamble: Building the data factories that could define India’s AI

So when Krithivasan—himself a quiet engineer who inherited the corner office after Rajesh Gopinathan’s abrupt exit in 2023—outlined this unprecedented investment, it was more than a capex announcement. It signalled a philosophical reversal within the Tata ecosystem, a return to building the physical foundations of digital India.

“The Tata Group companies are in power, real estate, project management businesses…So, there is a different kind of synergy we can build in here," Krithivasan said, during a call with analysts on 9 October.

The timing, though, came against an uneasy backdrop. Ahead of TCS’s latest earnings, on 9 October, its stock was languishing near a 52-week low. Growth was tapering. Analysts at Motilal Oswal and Nuvama projected a full-year revenue decline—the first since TCS’s 2004 listing. The mood across India’s $283-billion IT industry was sombre. Global clients were trimming technology budgets, US visa restrictions had upended the traditional onsite-offshore model, and TCS had just confirmed the layoff of over 6,000 employees in the September quarter.

In this subdued climate, the $6.5 billion commitment stood out as an act of economic courage, an attempt to reimagine TCS’s future rather than merely defend its margins.

TCS did not respond to Mint’s detailed questionnaire seeking comments.

April at the Taj

The idea took shape earlier this year.

In April, Chandrasekaran, who has been steadily integrating the Tata empire into a digital-industrial group, asked senior TCS executives to explore the data centre opportunity. It was a directive rooted in strategic timing as much as necessity.

The same month, at a high-profile event in Delhi’s Taj Mansingh Hotel, TCS unveiled its ‘sovereign cloud network’, a service designed to store and process data entirely within Indian borders in compliance with new data-localization rules. It was a logical extension of TCS’s IT services portfolio: the company already provided engineers, managed IT infrastructure, and offered private cloud services. What it lacked was real estate—the physical layer of computing power that anchored the digital world.

View Full Image

That gap is precisely what the new investment aims to close. Data centres are the factories of the AI age—vast, secure facilities that house the servers, networking equipment and cooling systems needed to process and store the petabytes of data generated every day. India’s regulatory mandates and digital explosion have created a perfect storm of demand. The Reserve Bank of India, the central bank, now requires payment data to be stored domestically; the Securities and Exchange Board of India, the markets regulator, has similar stipulations for market participants. As 5G networks roll out and connected devices multiply, the volume of data produced by consumers, vehicles, and industrial systems is surging. Add the rise of generative AI, and the need for high-speed computing infrastructure becomes existential.

TCS believes it can monetize this infrastructure in two ways: leasing capacity to hyperscalers such as Amazon Web Services or Microsoft Azure, and using it for its own contracts with governments and enterprises. The company’s first 150 MW facility, to be built at a cost of $1 billion, is expected to go live within 18 months.

Different from Accenture

The move also marks a clear break from how TCS’s global peers have evolved. Accenture, the $70-billion consulting giant, has doubled down on advisory-led AI and analytics rather than infrastructure. Its bet is that the profits lie in algorithms, not assets. Indian peers like Infosys and HCL Tech have largely taken the same path, positioning themselves as service orchestrators on top of third-party cloud platforms. On 13 October, HCL Tech announced that it had won $100 million in revenue from AI operations in the September quarter, making the Noida-based company the first among Big Five to quantify revenue from AI technologies.

TCS, by contrast, is wading into the physical side of that digital stack. It is a high-stakes pivot that could recast its risk profile and profit structure. “The beauty of the data centre business is its recurring stability," said one executive familiar with the plan. “But it will test patience before delivering cash returns."

The executive did not want to be identified.

Dividend reset

The patience required may be considerable. Unlike software projects, data centres are capital-intensive, slow to scale and operationally demanding. TCS’s decision to build rather than rent puts it in competition with hyperscalers and domestic real-estate developers alike. The company is entering a field already crowded with global giants—Microsoft, Google, Meta and Amazon—as well as Indian powerhouses such as Reliance, AdaniConnex and Yotta.

The irony is that, given its scale and client relationships, TCS should have been a natural candidate to own data centres years ago. Instead, it spent two decades returning wealth to shareholders— ₹3.66 trillion through dividends and buybacks since its listing. That generosity may now come at a cost. With the new investment cycle, margins could tighten, and dividend payouts to parent Tata Sons may moderate. Last year, TCS contributed 83% of Tata Sons’ income.

Data centres can be bracketed into four types. First, there are the enterprise data centres, which are facilities owned and run by companies and houses their computing networks. A second category is hyperscale data centres, also known as Cloud data centres. Amazon and Microsoft own and manage these facilities. Edge data centres are the third category. Here, smaller facilities are set up closer to an organization, thereby improving speed, and are essential for real-time data processing tasks. Finally, there are the colocation data centres, a space TCS is venturing into. In this category, the data centre owner sets up the physical infrastructure, manages the facility, oversees the building, power, and cooling for the servers. The customer supplies and manages their equipment.

View Full Image

McKinsey, a consulting firm, states that colocation data centres operate at lower profitability than cloud data centres. This is because the hyperscalers, who are the primary customers of colocation data centres, can negotiate better prices since they also set up and have their own facilities.

The operating margins of Equinix and Sify, who run colocation data centres, are in the range of 22% and 19%, respectively. In comparison, the operating margins enjoyed by AWS and Azure are significantly higher, at 35-37%.

Executives say TCS’s balance sheet can absorb the shock. The company generated ₹46,449 crore in free cash flow last year, more than enough to underwrite the initial phases. Yet investors accustomed to industry-leading operating margins (TCS ended with 24.3% in operating margin last year) may need to reset expectations. The stock’s 3.5% dip in three days following the announcement suggests the market is already doing so.

“TCS management stated that each 150-megawatt build-out phase of the data centre will require $1 billion capex, for a $6.5 billion total investment. We think this direction could offer new potential revenue streams though could negatively impact margins," Keith Bachman, an analyst with BMO Capital Markets, wrote in a note dated 9 October.

Returns are slow, and could take a decade. In addition, technology cycles move faster than depreciation schedules. “This is an asset-intensive industry where labour-arbitrage advantages don’t translate," said Peter Bendor-Samuel of Everest Group, a research firm. “Investors used to high-margin, low-capex returns will need to recalibrate."

Analysts at JM Financial estimated revenue from 1MW of operational data centres in the first year to be about ₹22 crore. TCS’s target of having a 150 MW data centre operational by April 2027 implies an incremental business of ₹3,300 crore. A successful operational capacity of 1 GW in six years would translate to about ₹22,000 crore, similar to its India revenue of ₹21,958 crore last year.

The 3% problem

The deeper question is why TCS waited until now. For years, Indian conglomerates like Reliance and Adani have been laying the groundwork for data centre dominance, tying these projects to their energy and telecom networks. The global hyperscalers, meanwhile, have been expanding capacity across India as part of their global grids. The Tatas, cautious and conservative, stayed on the sidelines.

Part of that hesitation stemmed from perception. Data centres were once viewed less as cutting-edge technology and more as glorified real-estate plays—capital-heavy, regulatory-ridden, and operationally messy. They required land, reliable electricity, and long-term financing, all of which had tripped up the Tata Group before. For a company built on intellectual capital, the business felt foreign.

What has changed is the scale of the opportunity. India today accounts for roughly 20% of global data generation but only about 3% of storage capacity.

Its total installed capacity—around one GW—is precisely what TCS aims to build by 2030. Analysts at Cushman & Wakefield estimate that India’s data centre capacity will triple to 1,825 megawatt (MW) by 2027, driven by new data-localization rules and the country’s voracious digital appetite.

States are competing fiercely for investment. Maharashtra, Tamil Nadu and Telangana are offering everything from electricity-duty waivers to land subsidies of up to 50%. Mumbai, with undersea cable landings and strong power grids, remains the favoured hub, but smaller cities are emerging as contenders.

For TCS, whose domestic revenue still accounts for less than 10% of its total ($30.18 billion) but grew 62% last year—largely due to a ₹1.5-billion BSNL contract—the timing aligns with a broader strategic shift toward India-focused business. As global clients hesitate, India’s digital infrastructure push offers both stability and scale.

The orchestra

Meanwhile, the data centre gamble will not only test TCS’s financial muscle but execution capability as well. Building and operating data centres demands coordination across engineering, real estate, power management, and cybersecurity—far removed from writing code.

Yet, this is where the Tata Group’s breadth could become a decisive advantage. TCS can draw on group companies such as Tata Power, Tata Communications, Nelco, and Tata Electronics for power supply, network connectivity, satellite links, and chip technology. That’s a web of synergies that rivals like Infosys or Wipro simply cannot replicate.

“TCS has a unique opportunity to serve the broader Tata Group across businesses in telecom, networks, energy and cooling solutions," said Ramkumar Ramamoorthy, partner at Catalincs, a consulting firm. “It can cross leverage these capabilities across the conglomerate structure, something that other large IT companies don’t have."

Still, competition will be fierce. To find its niche, TCS must differentiate not by sheer size but by integration—offering clients a seamless bridge between the data centre and the AI workloads that run on it.

“TCS is not trying to compete with AWS or Azure on a raw scale," said Phil Fersht, chief executive of HFS Research, a consulting firm. “It’s trying to own the intersection of infrastructure and services—both the data centre and the AI models that sit on top. That’s where the value will lie."

Such integration could also help TCS reposition itself in the eyes of investors. Global data centre operators like Equinix and Digital Realty command enterprise-value multiples of 20 to 22 times Ebitda (earnings before interest, taxes, depreciation, and amortization), higher than Indian IT firms’ price-earnings ratios. If TCS can credibly present itself as both a technology and infrastructure play, it could attract a new class of investors seeking exposure to digital infrastructure.

Still, this is not a business for those seeking quick wins. Like we mentioned earlier, data centres are a low margin play. Yet for TCS, this may be less about margins than momentum. The IT industry is at an inflection point: AI threatens to commoditize traditional outsourcing even as clients demand end-to-end digital transformation. Owning infrastructure offers TCS a new strategic lever, a way to control the stack from physical data to cognitive intelligence.

“TCS’ moat won’t be real estate—it will be orchestration," said Fersht. “By owning the infrastructure, it can combine data engineering and AI services under one roof."

Krithivasan’s wager, then, is not for immediate approval but for long-term relevance. As India’s data centre capacity surges and its AI ambitions swell, the company is positioning itself at the intersection of both trends. If it captures even a modest slice of the coming capacity boom, TCS could evolve from an outsourcing icon into a domestic infrastructure powerhouse.

For the Tata Group, too, this is a return to form. From steel and power to software and satellites, its empire has always thrived on long horizons. The data centre bet is merely the digital reincarnation of that legacy.

For Krithivasan, the challenge is deeply personal. He took charge in a moment of flux—after a surprise leadership exit, in a slowing industry, with AI rewriting the rules of competition. Now he must steer a company known for intangible efficiency into a world of tangible assets and operational grit. If he succeeds, TCS will not just sell cloud services—it will own a piece of the cloud itself.