Bhavish Aggarwal's AI startup Krutrim partners with Lenovo to develop India's largest supercomputer

Bhavish Aggarwal-founded AI unicorn Krutrim has teamed up with Lenovo to develop what they say will be India’s largest supercomputer, along with a new large language model called Krutrim 3, boasting 700 billion parameters.

The announcement comes on the heels of Ola founder Bhavish Aggarwal’s commitment to invest Rs 2,000 crore in Krutrim, with plans to expand that investment to Rs 10,000 crore by next year.

Speaking at Lenovo Tech World India 2025 in Mumbai, Ola Group CIO Navendu Agarwal said, “We are building the largest infrastructure and again very proud to say along with Lenovo, we are building the largest supercomputer of India. That will be powered in our cloud. Then we are also building the foundational models and our own chip.”

Agarwal noted that Ola currently has a 700-member team dedicated to creating full-stack AI solutions, spanning from application layers and contact center AI to manufacturing AI and an “agentic platform” that both startups and the broader ecosystem can utilize to build cost-effective, India-focused applications.

Krutrim’s progress in large language models has been rapid. In late 2023, the company launched Krutrim 1 with 7 billion parameters, followed by Krutrim 2 at 12 billion parameters—both were released as open-source projects.

Now, Krutrim 3 aims to scale up dramatically to 700 billion parameters, a figure that the company believes will position India as a major contender in the global race for advanced AI development.

“We are very proud to say that we are working on a bigger model, Krutrim 3, which will be a 700-billion parameter model, which will be an answer from India to show that we can build the best,” Agarwal said. “And we are very proud to partner with Lenovo in this endeavour.”

The push for sovereign AI infrastructure has grown increasingly urgent as countries compete to reduce their reliance on foreign platforms and cloud providers.

Agarwal highlighted India’s lack of hyperscale data centers and dedicated AI research clouds, painting Krutrim’s efforts as crucial for the country’s move toward building and hosting its own large-scale AI engines. 

Earlier this year, Krutrim announced it would open-source Krutrim 2 and other foundational models, including Chitrarth 1 (a vision-language model) and Dhwani 1 (a speech-language model).

Bhavish Aggarwal had also revealed partnerships with American chip giant Nvidia for deploying GB200 “superchips,” and said the company had tested advanced AI models from Chinese startup DeepSeek on Krutrim’s domestic servers to bolster data privacy.

In a post on X (formerly Twitter), Aggarwal wrote, “Building India’s largest supercomputer in @Krutrim with @Lenovo! Need much more to bring in the AI revolution to India at scale! All in to make this happen soon,” soliciting suggestions from the public for naming the supercomputer.