Mumbai: India must urgently build sovereign AI capabilities or risk becoming dependent on foreign technology firms that are using Indian data to strengthen their own models, former CEO of Niti Ayog Amitabh Kant has said.
Speaking at the Mint All About AI Tech4Good Awards, he said the country’s technological trajectory will be defined by how quickly it expands its domestic computer infrastructure, protects sensitive data, and supports startups building indigenous AI systems.
“No other technology in recent times has taken the world by storm like artificial intelligence,” Kant said. He described the moment as a rare convergence of economic opportunity, national security concern, and demographic potential, adding that India’s position as one of the world’s largest generators of digital data gives it a unique opening to shape AI systems rooted in local needs.
Kant warned that while India’s digital public infrastructure has powered inclusion at scale, the country’s lagging compute capacity threatens to slow its progress. He pointed to the recent OpenAI and Nvidia partnership to create 10 gigawatts of GPU capacity, equivalent to roughly five million of the latest high-performance processors. India, by comparison, has a base of about 30,000 GPUs.
“Bridging this gap requires massive private sector investment and foreign direct investment. Today, OpenAI’s ChatGPT in India reportedly has more monthly active users than in any other country. About 33 per cent higher than even the United States,” he said.
He argued that sovereign capability is essential for three reasons. First, self-reliance would encourage a stronger domestic startup ecosystem and attract capital into AI hardware and software. Second, models attuned to India’s languages, cultural context, and public sector needs would make AI applicable to every citizen. Third, a secure domestic infrastructure would be critical for national security, particularly as AI systems become deeply embedded in public services.
Kant said global firms are offering free or low-cost services and using this data to train closed-source models at scale. “Our future may involve having AI services powered by our own data but owned by others and sold back to us,” he said.
He called for a regulatory framework that allows global models to operate in India but requires them to run on infrastructure hosted within the country. No user data, he said, should leave India for applications involving large language models. This approach, he argued, would attract investment into computing capacity and strengthen privacy protections.
He said India’s emerging AI startups, including Sarvam AI, Sokhat AI, Dhani AI, and Gantt AI, are beginning to build foundational models but will need access to large pools of high-quality data and powerful compute to reach global standards. National programmes aimed at attracting top AI researchers, improving access to anonymised datasets, and expanding public compute marketplaces will be essential.
Kant said global examples show that countries arriving late to a technological shift can sometimes leapfrog incumbents. “History shows that in tech, sometimes those who come second can build faster, learn smarter, and avoid earlier pitfalls,” he said, noting how Google overtook early search engines and how later entrants transformed hardware markets.
(IANS)
