4 min readNew DelhiUpdated: Feb 16, 2026 08:00 PM IST
India should declare digital infrastructure as the “most essential commodity” to become a leader in AI production and must move from being a massive generator and consumer of data to a sovereign processor of it, at scale, said Sunil Gupta, CEO and co-founder of Yotta Data Services on Monday.
During a discussion titled ‘Inside India’s Frontier Lab and Its Global South Impact’, at India AI Impact Summit, Gupta said that nearly 1 billion Indians carry smartphones connected to the internet, accounting for roughly 20% of global data creation and consumption. However, India processes only about 3% of it domestically, he said, adding, “While this is a problem, it also shows an opportunity.”.
Joseph Joshy of International Financial Services Centres Authority (IFSCA), Rangarajan V of Adani Defence & Aerospace, and Sahil Arora of Qualcomm participated in the discussion.
Gupta underscored that India cannot just rely on imported technologies and must create its own digital infra in terms of computing to define its own models, datasets and use cases.
Over the past 7 years, India’s data centre capacity has grown nearly seven-fold. Yet Gupta cautioned even if infrastructure continues to expand for another decade, the country may still fall short of the compute-scale required to serve both its domestic AI ambitions and emerging global demand.
Hybrid deployment
Sahil Arora said India’s AI future will likely rely on hybrid architecture — with smaller models running on devices and heavier processing handled in the cloud. “You don’t send everything to the cloud. Just send the text to the cloud. It brings down the overall compute or data connectivity requirements and it basically makes your latency down,” he said.
“… if you can run it on a device and then the final, you can say confirmation or final engagement can happen to the cloud, that’s the real India scale deployment scenario in my opinion,” he added.
Story continues below this ad
Localisation a necessity
Rangarajan V stressed the need for indigenisation of AI models to make them relevant to India’s linguistic, cultural and strategic realities. Citing myriad dialects across India and even within states, he argued that localisation is not optional but essential. A model trained elsewhere may achieve partial relevance, he suggested, but the gap in a country of India’s scale translates into hundreds of millions of people.
“You need local models handled by local people in order to be truly relevant for the full population,” Rangarajan added.
Data corridors
Joseph Joshy said India’s new Digital Personal Data Protection (DPDP) Rules, 2025 could unlock strategic “data corridors” between countries, enabling structured and trusted cross-border data flows. If India wants to lead the Global South, he said, it must re-engineer the corridors’ functioning across multiple jurisdictions.
Story continues below this ad
“When you have a bridge between two countries… cross-border data, how does it go? That is a very different perspective than just for our own country,” he said.
Using finance as an example, he pointed to the possibility of portable KYC systems. If identity and compliance data can move seamlessly across borders, MSMEs and individuals could “passport” their business and identity into other jurisdictions, and vice versa.
“…this is going to kind of create this leadership space which you need and in turn is going to feed into all of your AI models here because now your AI is not just restricted to a narrow jurisdiction because of the data restriction,” he said.
With the DPDP implementation timeline in place, he expects a “handshake” between stronger data governance and AI development, creating, what he described, as the right environment for scale.
.