Building an AI model from scratch costs hundreds of millions of dollars. Most countries never attempt it. BharatGen is trying anyway — and the consortium warns that India has no real choice.
“We can only truly understand AI models when we build them from scratch,” said Ganesh Ramakrishnan, the founding director of BharatGen and a chair professor at IIT-Bombay. He compared the process to cooking.
Know the recipe, and you can adjust ingredients for your own needs. Without it, you just eat what you are given. That analogy carries a blunt warning.
Indian enterprises and startups today overwhelmingly rely on foreign foundational models. The country consumes AI built elsewhere.
BharatGen wants to change that, but the clock is running. The consortium received a government grant of ₹1,058.82 crore on 18 September 2025. That money comes from the Centre’s $1.2-billion AI mission.
BharatGen is the largest single beneficiary of that fund. It is also India’s only not-for-profit AI entity. Established in December 2024 by IIT-Bombay, it now operates as a consortium of AI labs spread across nine Indian institutes.
The project is building one of India’s first sovereign large language models. Those are the foundational AI systems that mimic human writing and speech by analyzing vast amounts of data.
The model is being built from scratch — not adapted from an existing foreign one. That distinction matters. Ramakrishnan said that knowing exactly what goes into an AI model — especially its data — is essential for building effective tools.
He warned that an ecosystem built solely around using AI tools will remain at what he called “the consumption layer.” Users, he argued, must move “from the last mile to the first mile.” The risk is concrete. If India stays a consumer of foreign AI, it loses control over data, over model behavior, over the very systems that will shape its digital economy.
The country becomes a market, not a maker. That is the outcome BharatGen exists to prevent. The consortium is now developing its first model.
The work is early. The funding is substantial but finite. The goal is a sovereign AI capability — an Indian system built on Indian data for Indian needs.
Ramakrishnan put it plainly: “For AI adoption to scale, users have to move from the last mile to the first mile. We need people who understand, build and improve.” That is the core of what BharatGen is trying to do.
Not just fund research. Build a generation of builders. The alternative, in the consortium’s own framing, is permanent dependency — a country that uses AI but never truly owns it.
Sources
- India must build foundational AI models or risk becoming a mere…
- BharatGen Warns India Must Build Foundational AI Models Now
