ET Soonicorns Summit 2025 curtain raiser: AI for Bharat and the race to build localised data centres

With the ET Soonicorns Summit 2025 set to take place on 22 August in Bengaluru, EconomicTimes.com kicked off the conversation with a high-stakes virtual curtain raiser livestream. Among the three panels diving into different facets of artificial intelligence (AI), the opening session—AI for Bharat: How Localised Data Centres Can Bridge the Digital Divide—set the tone for the scale of ambition and urgency the Summit promises to deliver.

Moderated by Akhil George, the discussion brought together four industry leaders: Jai Asundi, Executive Director, Centre for Study of Science, Technology and Policy (CSTEP); Adarsh Natarajan, Founder and CEO, Aindra; Abhinav Aggarwal, Co-Founder and CEO, Fluid AI; and Ankit Bose, Head, Nasscom AI.

The central question: How can India harness localised AI-ready data centres to power inclusive innovation, protect sovereignty, and build for Bharat, not just for the technically savvy but for every citizen?


AI for Bharat: A mission beyond the tech elite
For Jai Asundi, Executive Director, Centre for Study of Science, Technology and Policy (CSTEP), the essence of AI for Bharat is about solving “problems for India” across its diverse linguistic, social, and economic landscape. “It cannot be only for all the technically savvy folks like us,” he stressed. “But also for… the farmer, the person in the tier two, tier three town… or even the poor person in the city.”

In his view, localised data centres are critical because “the data that we generate… is valid for Indians” and must be stored and modelled in context, rather than on non-contextual, foreign-built systems.

Adarsh Natarajan, Founder and CEO, Aindra, brought this home from a healthcare perspective. At Aindra, which builds AI models for cervical cancer detection, “patient data is absolutely sacrosanct,” he emphasised. “Where do you store this data? Where is it available for us to actually take and model around?… This is absolutely an era where we really need to think long and hard about where we are storing our data.”
Trust, narrative, and sovereignty
Abhinav Aggarwal, Co-Founder and CEO, Fluid AI, which offers conversational AI solutions and AI-based analytics, argued that sovereignty in AI is as much about trust as it is about infrastructure. “You can’t have [critical data] moving to parties where you have no control,” he warned. The risk goes beyond privacy; it's about controlling the narrative itself.

“I might ask an opinion from ChatGPT about Donald Trump, which will be very different if I ask a DeepSeek… And if India doesn’t have those models, it’s very easy for an external party… to control India’s narrative very unknowingly.”

For Ankit Bose, Head of Nasscom AI, AI for Bharat means “Indianised models, having Indian languages, Indian culture, context, and for Indians… at the cost we can have.” Most existing models are rooted in Western cultural assumptions, he noted, while India’s 22 official languages and countless dialects require unique AI training. “AI for Bharat is… the biggest thing which can give power to the deprived communities,” he said, from rural healthcare to education and policy access.

The cost challenge and public–private models
High-performance AI is expensive. As Asundi pointed out, “getting GPUs is very expensive, running data centres is very expensive… The capex… is going to be very high.” His suggestion: let the government invest in infrastructure and allow startups and smaller firms to “bear the marginal costs as opposed to the average cost.”

This mirrors the emerging IndiaAI Mission’s approach, which seeks to pool and allocate AI resources, including GPUs, more efficiently.

Natarajan agreed, noting that healthcare AI, in particular, suffers from both infrastructure and data bottlenecks. Without policy intervention, “it will never achieve the kind of scale that… a policy can provide.” He called for anonymised healthcare data sharing under strict privacy safeguards, enabling “grassroots-level innovations happening across the country in multifold.”

India’s AI infrastructure gap
Bose laid out the stark reality: India generates “20% of the global data” but has only “approximately 3% of global data capacity.” For AI-ready facilities, the gap is worse—current GPU availability is “not beyond 20,000 to 25,000 GPUs” for the entire country, compared to single companies abroad holding multiples of that.

The government’s pledge to allocate 20,000+ GPUs to startups and research is a start, but Bose stressed: “Is that enough? No… More capacity needs to be pumped in so that we can stay ahead in the global race.”

Enterprise implications: Cost, cultural fit, compliance
From an enterprise standpoint, Aggarwal identified three key drivers for developing local AI infrastructure. The first is economic retention—ensuring that gross domestic product (GDP) and foreign exchange remain within India instead of flowing abroad when companies procure AI services from overseas providers. The second is cultural adaptation, which he illustrated with the example of a bank wanting its AI to “sound Indian” rather than like an American voice attempting an Indian accent. The third is data compliance, with local infrastructure helping enterprises avoid regulatory and privacy risks by keeping sensitive data within national boundaries. Beyond these immediate drivers, Aggarwal emphasised India’s need for “patient capital” to support long-term innovation, and urged a shift in large technology firms’ return-on-investment (ROI) models towards building intellectual property-heavy, foundational AI platforms rather than relying solely on service-driven offerings.

Global lessons—or a unique Indian model?
On whether India could adapt decentralised strategies from other nations, Asundi was clear: “We may need to create our own models… We do fall in a fairly unique place,” given India’s scale of data generation and the stakes for sovereignty.

Natarajan expanded on how this could fuel non-metro innovation: India’s vast patient volumes generate unparalleled datasets. The challenge is making that data “usable” and ensuring innovators have “the right kind of ingredients” to work with—quality infrastructure, quality data, and localised processing power.

From curtain raiser to main stage
The first panel of the curtain raiser left no doubt that the AI for Bharat mission is not just about technology, but about ownership, context, and empowerment.

As Deepak Ajwani, Editor, EconomicTimes.com, said while opening the day’s livestream, AI is “beautiful, beneficial, purposeful and profitable” but also brings risks—from financial fraud to ethical challenges—topics covered in the subsequent panels.

If this first discussion is any indicator, the upcoming ET Soonicorns Summit 2025 will not just showcase startups and technology; it will spotlight the policies, partnerships, and infrastructure India needs to lead in AI on its own terms and other sessions on capital efficiency, defensible moats, and scaling strategies. The stakes are high, the ambitions higher, and the conversations have only just begun.

The ET Soonicorns Summit 2025 returns to Bengaluru on 22 August for its fourth edition. Registration for the ET Soonicorns Summit 2025 is now open.

360 ONE is the Presenting Partner of the ET Soonicorns Summit 2025, with Raymond as the Wardrobe Partner and Shiv Nadar University as the Ecosystem Partner.