Three things happened in eleven days.
The US ordered Anthropic to cut off its most powerful models for foreign users.
HCL Tech put $150 million into Sarvam.
And Mukesh Ambani told Reliance shareholders that India cannot keep renting its intelligence.
Everyone agrees India needs sovereign AI. And yet, almost nobody agrees on who should pay for it. Most of the money that is moving into AI goes into data centres, the safest layer, not the models themselves.
Praveen Gopal Krishnan makes the argument that corporate India has to build and fund the country's foundation models because the state can't do so at scale and the market hasn't chosen to. In this episode, he poses this argument to two guests — one a public policy expert, and the other an AI investor, builder and entrepreneur.
Nitin Pai is co-founder and director of Takshashila Institution, India's largest independent public policy school. Manav Garg built Eka Software from India and sold it to a US private equity firm, co-founded Together Fund and AI Bhumi, and is now executive chairman at Emergent — an Indian AI company whose product runs entirely on foundation models it doesn't own.
Both Nitin and Manav push back from two different directions and describe the incentives that keep conglomerates out of anything that doesn't compound quarter over quarter, and why the unit that matters isn't the company but the builder willing to take the risk inside it.