The AI Smoke and Mirrors Fuelling India’s Data Centre IPOs

India's data centre ecosystem is preparing for a defining shift. The companies fuelling India’s AI pivot and establishing facilities designed to handle the massive AI compute demand with high-density GPUs, advanced cooling systems, and server racks are now turning to public markets, signalling that data centres are being repositioned as the backbone of India's AI economy.

Earlier this year, Sify Infinit Spaces filed for what was expected to be India's first pure-play data centre IPO, while Yotta is actively working towards a listing. Operators including STT GDC India, CtrlS, and Nxtra by Airtel are reportedly evaluating public-market routes, hoping to convince institutional and retail investors that data centres are not merely real estate assets housing servers but strategic AI infrastructure.

Yet beneath the enthusiasm lies a fundamental question: are these companies raising capital to meet proven demand, or building infrastructure for an AI boom that has yet to fully materialise?

"I believe both forces are in play, but it's driven more by future expectations," Arindam Goswami, Staff Research Analyst at the Takshashila Institution, tells AIM. "India's data centre market is presently expanding on the foundations of demand from before the AI story: cloud migration, enterprise IT, BFSI, and localisation rules from the RBI, SEBI, and the DPDP Act that keep sensitive data onshore."

He argues that while these structural drivers remain important, "this demand is, however, still thin, and probably not commensurate with IPO expectations." Instead, he sees public listings primarily as "a financing mechanism to build ahead of the AI curve", allowing data centre operators to fund AI-ready capacity expansion while positioning themselves as AI infrastructure platforms rather than conventional colocation providers.

That distinction matters because enterprise AI adoption may not accelerate as quickly as market valuations suggest.

Goswami points to what he calls the "Productivity J-curve", wherein organisations spend years restructuring workflows before AI meaningfully improves productivity. Falling inference costs alone may not guarantee widespread enterprise adoption, particularly if customer pricing remains high. "If capability keeps improving without matching revenue, some of this capacity risks sitting idle," he cautions.

Some data centre companies are now rethinking their IPO strategies. As per a Bloomberg report, Sify Infinit Spaces has put its $391 million IPO offering on hold amid a weaker stock market.

Yotta's Global AI Infra Play

Yotta's planned public listing reflects the changing economics of building AI infrastructure in India. Unlike traditional colocation facilities, AI-ready data centres require billions of dollars in upfront investment, pushing operators to look beyond conventional debt and private capital.

Speaking to AIM, Sunil Gupta, Co-founder, Managing Director and CEO at Yotta Data Services, confirms that the company's IPO plans remain on track and dismisses recent reports which claimed that it was raising up to $600 million at a $4 billion valuation.

Yotta recently secured $150 million at a valuation of $3.9 billion and has already begun its pre-IPO fundraising. The company has attracted more than 150 investors, including family offices and high-net-worth individuals, before reaching out to large global institutional investors.

"We are actively in touch with them," he says, adding that Yotta expects to file its draft red herring prospectus in the coming weeks. Before going public, Yotta wants to shore up capital and aims to raise between $1.25 billion and $1.5 billion.

Gupta says the fundraising is being driven by the scale of investment required to build AI infrastructure. "Our capex requirements are huge," he says, noting that while Yotta is also exploring innovative financing models, including revenue-sharing structures and partnerships with global funds investing directly in GPU infrastructure, "even above that, we have to do our own fundraising with IPOs."

He also stresses that the listing is intended to fund expansion rather than provide an exit for existing shareholders. "No promoter is exiting, no promoter is selling shares," Gupta says. "The entire funding... will go to the company for further capex investments."

Governance Questions

While AI optimism dominates investor conversations, governance experts argue that another set of risks remains largely invisible.

"Honestly, right now it's mostly growth," observes Vidhi Sharma, Head of Responsible AI Governance at think-tank Future Shift Labs. "I don't think that investors are being careless; it's just that they price what's actually disclosed to them. Power consumption, water use, the whole societal footprint of running these facilities—none of that shows up anywhere in a prospectus today."

She believes public listings should become the moment these questions are formally addressed.

Drawing parallels with previous technological transitions, Sharma argues that markets are currently rewarding construction speed rather than understanding long-term costs. "Valuations are tracking how fast things are being built, not what they'll actually cost down the line," she says, comparing today's blind spots around AI infrastructure with earlier periods where systemic risks remained hidden until much later.

These include the rapidly evolving hardware cycles. Data centre companies need to replace servers, networking gear, and cooling systems as computing hardware depreciates and higher-performance chips enter the market.

For Sharma, mandatory disclosures should extend beyond traditional financial reporting.

"I believe we're at the concrete-pouring stage with data centres," she says. "Energy and water intensity are the obvious metrics, but workload type matters just as much." She argues that facilities training frontier AI models face very different resource requirements from those primarily handling inference workloads, yet investors currently have little visibility into those distinctions.

Her concern extends beyond environmental issues. Drawing on her own research into AI supply chains, Sharma says the same disclosure gaps that obscure energy consumption also hide labour conditions affecting content moderators and data annotators. "Nobody's counting these costs, but they don't just vanish; they get built into the foundation and get more expensive to dig out the longer you wait."

Can Infra Outrun Demand?

Not everyone believes India's AI infrastructure expansion is speculative.

Rohit Kapoor, Co-founder of sovereign cloud platform Cloud4India, argues that data localisation requirements are creating genuine structural demand rather than a purely AI-driven bubble. However, he notes that execution remains significantly behind, with only a fraction of announced capacity currently operational. 

He also warns that data centre concentration in Mumbai and Chennai, power constraints, and the rapid evolution of AI hardware present underappreciated risks for investors.

At the same time, Kapoor contends that India's capital base remains overwhelmingly institutional, suggesting the industry is not turning to IPOs because private capital has dried up. Rather, it’s a sovereign AI necessity.

“L&T's sovereign AI factory tie-up with NVIDIA is a solid proof point for India's 'sovereign infrastructure' narrative. The 20-year tax holiday, through 2047, also signals long-term government confidence in demand rather than capital scarcity,” Kapoor notes.

For policymakers, however, financing is only one part of the equation.

Goswami argues that India has become increasingly effective at encouraging data centre construction but less successful at governing its long-term consequences. While incentives around land, approvals, and electricity have accelerated investment, there are broader questions surrounding grid reliability, water stress, energy planning, and digital sovereignty.

"Policy thinking in India increasingly recognises data centres as strategic digital infrastructure rather than ordinary real estate, which is the right conceptual shift," he says. But he adds that true digital sovereignty requires more than storing data within India's borders. "More than just 'keeping data in India', India needs to orient towards 'keeping critical compute, control, and resilience in India'."

That shift may ultimately define whether India's data centre IPO wave marks the emergence of a durable infrastructure asset class or simply the latest chapter in AI exuberance.