India is nowhere close to building frontier AI models and everyone knows it: Wingify's Paras Chopra

India is nowhere close to building frontier AI models and we need to invest in deep, foundational research if it hopes to leapfrog in artificial intelligence, said Wingify founder Paras Chopra at ET AI Awards 2025 on Thursday.

“We are nowhere near frontier models and everyone knows it. The compute gap alone is orders of magnitude,” Chopra said, pointing to the vast difference in GPU capacity between the US and India.

Chopra was speaking during a panel discussion on ‘Build, Borrow, or Burn: The Truth About Models’ with Peeyush Ranjan, founder of Meraki Labs, moderated by ET’s Samidha Sharma.


“Models are lagging indicators of years of research that looked like it was going nowhere,” Chopra said, referencing how companies like OpenAI evolved after researching for years ‘before they stumbled upon something’.

After Wingify, Chopra founded Lossfunk, a lab based out of Bengaluru supporting foundational research on AI.

According to Chopra, India’s biggest constraint is mindset. “We need to shift from a business mindset to a builder’s mindset — from chasing near-term profit to building the future,” he said.

For Ranjan of Meraki Labs, the opportunity lies in leapfrogging rather than competing head-on. “You don’t become great on day one. You go down the learning curve by building things deeply relevant to you,” he said.

Ranjan, a former Flipkart CTO and Google veteran, launched a global AI tutor-led edtech startup, Fermi.ai in January this year.

“AI is a road roller flattening advantages. The question now is: who can solve messy problems most creatively and move fastest?, he said, suggesting that India should build globally competitive AI products in sectors such as education and healthcare, where it has scale and domain strength.

“The right approach isn’t building for India,” Chopra said. “It’s building world-class products from India.”

On building startups from India, Chopra said that companies and founders should reinvest in level zero research with exits, instead of distributing dividends. “The only way to counter that is to invest more in fundamental R&D. That level zero work is what changes things ten years down the line,” he said.