Not chasing AGI, betting on scale: IndiaAI chief outlines roadmap ahead of summit

AS countries contend to establish ever so larger AI models, bharat is choosing a different track. “We are not…into the race for artificial general intelligence (AGI),” said the chief executive officer of the IndiaAI Mission in an interview with HT, laying out a pragmatic vision for the country’s AI strategy ahead of the AI Impact Summit.AGI refers to hypothetical AI systems which are capable of performing any intellectual task a human can, rather than being limited to specific functions. Globally, all major AI labs like OpenAI, Google DeepMind, Anthropic, Meta AI (FAIR), Microsoft AI, and xAI are racing to build increasingly powerful systems, pouring billions into larger models and computing infrastructure in pursuit of AGI.Instead of chasing trillion-parameter models or headline-grabbing breakthroughs, Abhishek Singh, CEO of India AI Mission and additional secretary at Ministry of Electronics and Information Technology (MeitY), said India’s priority is simpler, to build systems that work at population scale.The recent Economic Survey argued for smaller, task-specific models that can run on limited hardware and decentralised compute networks, rather than the large, resource-intensive systems being pursued by Big Tech and countries such as the United States and China. Singh echoed this view.“Why not?” he said, “It’s not about big or small. Models should not be measured in terms of number of parameters. Models should be measured on what problem… they are solving,” pointing to use cases across languages, healthcare, education, agriculture and manufacturing. “These models need not be a trillion parameter model,” he added.The first phase of the IndiaAI Mission, with a funding of ₹10,372 crore, is funding a mix of systems, eight large language models and four smaller ones by providing them with subsidised compute. The 12 startups will be present at the AI Impact Summit, where they are expected to showcase either full-scale models or early versions of their systems. The objective, Singh said, is sovereignty.“Our priority is to build something that is sovereign, that is built in India, hosted in India, and designed for our challenges.”If those models later scale globally, “we are not averse to that,” he added, “but the goal is not that.”India: Inference capital of the worldWhere many governments talk about supercomputers to train frontier models, Singh emphasised a different challenge, which is running AI continuously for millions of users.“India has the potential to become the inferencing capital of the world,” he said, adding that India could emerge as a major hub for AI inference workloads. Training builds models, while inference runs them at scale.However, the scale of computing required to run AI services at population level is enormous. Under the IndiaAI Mission, which Singh heads, India currently has just over 38,000 GPUs that have been made available at subsidised rates to startups, researchers and students. That, he suggested, is nowhere near enough.Would close to 40,000 GPUs be enough? “For a country like ours, if we do inferencing, I may require at least 100,000, 200,000,” he said of the capacity needed as usage grows. “For a country of our size, even a million will not be sufficient, if everybody starts using it.”Electronics and IT Minister Ashwini Vaishnaw recently said that the second phase of the India AI Mission will be announced in about five months, adding that the government will also be announcing an addition to the current 38,000 GPUs at the Summit.Singh said recent policy moves, including tax incentives for data centres, should encourage more global investments into India.Presenting Budget 2026-27 on February 1, Finance Minister Nirmala Sitharaman said foreign companies serving global cloud customers through Indian data centres would get a tax holiday until 2047, allowing hyperscalers such as Google, Amazon and Microsoft to route overseas business from India without their global income being taxed here.At a post-Budget briefing, IT minister Ashwini Vaishnaw said India has already secured about $90 billion in AI and data centre investments, with $70 billion underway, and projected the total could rise to $200 billion as more commitments come in.Singh agreed that the tax holiday incentive would be a stimulus for data centre investments. “The global industries are already looking at India as a big data centre place…given the workforce that we have, given the talent that we have, a lot of AI workloads will come to… Indian companies, Indian startups,” said Singh.India targets more signatories than French SummitIndia is hosting the Summit from February 16 to 20 at Bharat Mandapam in New Delhi, with an expected attendance of over 100,000 people. About 20 heads of state, 100 government representatives and over 100 global AI leaders are expected to attend.He described the event as larger than previous global AI summits in scale and participation, with the agenda shaped through consultations with industry, academia, governments and civil society. Even during the HT interview, staffers stepped in with urgent matters, while a line of visitors waited outside his room, a reminder that with days to go, summit preparations were still in full swing.While Singh declined to preview the summit’s outcomes, repeatedly insisting that announcements would be saved for the event itself, there are seven international working groups finalising “very tangible outcomes,” he said.India is aiming for its broadest backing yet on the summit declaration, hoping to surpass last year’s French meet, where 58 countries signed, even as the US and UK stayed out. By comparison, only 11 countries endorsed the Seoul statement and 28 signed at Bletchley Park. Negotiations on the joint declaration are still underway.“We’re hoping that everyone signs,” he said, adding the goal is “more than anywhere else.”Asked whether missing signatures from the US and the UK could weaken the India summit’s credibility, Singh dismissed the concern and struck an optimistic note. “Glass is half full or half empty? Why should we be pessimistic?” he said, adding that his focus was not on who might stay out but on “working with the various countries” and addressing their concerns to arrive at a declaration “acceptable to most countries.”Singh acknowledged that in the global AI race, the US and China remain ahead, with India a distant third. India would close that gap, Singh said, “...by providing more investments in R&D, by investing more, by providing GPUs, by building the dataset platform, by funding foundation models, by providing support for AI projects. We have the capacity to not only catch up, but leap frog.”