This instinct to leapfrog, to diffuse technology in ways no one expects, is in India’s DNA, and it is this instinct that informed the Prime Minister’s speech and the various technology missions, from Quantum to Artificial Intelligence. With the AI revolution reshaping economies and redrawing the map of power, the question for India is not whether we will build AI but whether we will build it differently.
The IndiaAI Mission’s announcement of BharatGen out of the Indian Institute of Technology Bombay as the sovereign AI initiative, backed by more than ₹900 crore in funding, is more than a declaration of intent. Bharat will not simply depend on imported intelligence; it will harness the talents of its best to build it on its own terms. The challenge is not merely about catching up with the state-of-the-art models or creating India’s DeepSeek moment. It is about diffusion, for true sovereignty cannot be achieved in the face of aggressive pricing and bundled offers from international platforms looking to lock up the Indian consumer. India’s sovereign AI will have to chart a viral path that spreads the way prepaid recharge cards once did, like UPI QR codes, like Aadhaar numbers.
To become India’s AI Manhattan Project, BharatGen has to be more than a model. It must become an ecosystem anchor. "It must evolve into a stack, much like Aadhaar or UPI did, with APIs, developer toolkits, open deployment frameworks, as the next generation of digital public infrastructure. The goal of diffusion must be to seed clusters of innovation across state universities, regional startup hubs, and even grassroots innovators, spreading the capacity to build, test, and deploy AI across the nation. Diffusion must also mean redefining the role of citizens. They are not just end users, but co-creators. Farmers supplying data about soils, teachers refining prompts for classrooms, health workers validating AI responses — all of them contribute to the system’s quality. If Aadhaar was a unique identity built on citizen participation, BharatGen can be a unique intelligence shaped by citizen input.
For this to succeed, AI must live where people live. And in India, that means the edge. Reliable broadband and infinite cloud credits are luxuries. Smartphones costing ₹5,000, low-cost tablets or laptops in government schools, and even feature phones enhanced by voice AI — that is where intelligence must reside. Imagine a farmer in Bihar asking about soil health in Bhojpuri and receiving instant offline answers. That is the kind of edge AI must aspire for to leverage adjacent deep-tech innovations such as Direct-to-Mobile (D2M) Broadcasting.
But diffusion does not happen by accident. It must be designed. Hardware has to be an integral element of sovereignty. Without domestic chip design for low-power accelerators and without affordable AI-ready devices, diffusion will stall. Sovereign AI must, therefore, mean sovereign compute and uniquely Indian means of pushing data and models to the edge.
The current haste in India to adopt western models without guardrails carries the risk of AI diffusion amplifying biases against Bharat, while spreading misinformation in dialects spoken by vulnerable communities. But retreat is not an option. India’s sovereign AI must also be a participatory AI that harnesses the power of communities to give the disproportionately funded western models a run for their money. Hence, the challenge for BharatGen is to emerge as the UPI of AI across the country. India has in the past shown how to build digital public goods that the world adopts. UPI is now being exported. Aadhaar has inspired similar identity systems abroad. BharatGen can be the next chapter. The Global North may dominate the AI leaderboard in parameter counts. But India can dominate in something more fundamental — in making AI usable, accessible, and relevant for a billion people. National repositories of edge-ready models and open toolkits can provide a backbone to small businesses. AI literacy campaigns in schools and villages can ensure citizens not only use AI but also help shape it. Diffusion, in other words, can become democratisation.
India stands at the threshold of its AI moment. BharatGen, with its substantial public funding, can be the flagship, but the fleet must be the ecosystem that surrounds it. Success of this 21st century Manhattan Project will not be measured in benchmarks alone, but in whether AI helps a farmer plant better, a nurse diagnose faster, a student learn smarter, or a weaver sell wider. While Silicon Valley is focused on the race to build super intelligence to empower few, Bharat must build everyday intelligence to empower the many billions — not just in India but across the Global South.
The author is cofounder, DeepTech for Bharat (AI4India.org), and former CEO of Prasar Bharati
 "artificial intelligence, Technology, Manhattan")
