No need to spend billions to build LLM, barrier to entry for AI models coming down: Nandan Nilekani

Nandan Nilekani, co-founder and chairman of Infosys in conversation with Nalin Mehta, Managing Editor, Moneycontrol

Infosys co-founder and chairman Nandan Nilekani on Friday discussed India’s artificial intelligence (AI)-driven future and the challenges of digital transformation.

Speaking about DeepSeek, the new Chinese AI startup that has made headlines worldwide after it topped app download charts, Nilekani said, “What DeepSeek has done is that they have shown that you don’t need billions of dollars to create a reasonably good large language model. That’s a big breakthrough.”

He was in conversation with Nalin Mehta, Managing Editor, Moneycontrol during a session at AIMA.

The barrier to entry of creating AI models has significantly come down, he said pointing towards the latest developments in the sector. “The government is focused on the India AI Mission and I think we will see in the next one year, some very significant models coming from India,” Nilekani stated.

On being asked if India must spend on foundational models, Nilekani stated, “It’s really all about the price point. Would I spend a billion dollars to build a large language model (LLM)? No. But if I can deliver a LLM in $50 million, sure. The technology is moving so quickly that it’s dropping in price and becoming more efficient.”

He also expressed hope that India will be a leader in applying AI on scale. “What I want to see is the application of AI at the population scale. For instance, in sector such as education, agriculture,” he said.

The challenge with AI is to make it work at our population scale, Nilekani said, adding, “that’s the way we have to go”.