SoftBank Billionaire Invests $230M in Indian AI Startup Krutrim

Ola Founder Invests Heavily in AI Startup Krutrim

Bhavish Aggarwal, the founder of Ola, is committing $230 million to Krutrim, an artificial intelligence startup he established. This substantial investment arrives as India strives to become a significant player in the AI sector, currently led by firms from the U.S. and China.

Funding Details and Future Plans

The funding for Krutrim, which focuses on developing large language models (LLMs) specifically for Indian languages, is primarily coming from Aggarwal’s family office, according to a source familiar with the situation. Aggarwal announced on X that Krutrim aims to secure $1.15 billion in investment by the following year.

The remaining capital will be sought from external investors, the source added.

Open Source Models and Supercomputer Development

This announcement coincides with Krutrim’s decision to make its AI models open source. Furthermore, the company is planning to construct India’s largest supercomputer in collaboration with Nvidia.

Krutrim-2 Performance

The lab recently released Krutrim-2, a 12-billion-parameter language model demonstrating notable capabilities in processing Indian languages. In sentiment analysis evaluations shared by Krutrim, it achieved a score of 0.95, surpassing the 0.70 score of competitor models.

The model also exhibited an 80% success rate in code-generation tasks.

Specialized Models for Indian Languages

Several specialized models have been released as open source, including systems designed for image processing, speech translation, and text search. All of these are optimized for use with Indian languages.

Aggarwal’s Vision for an Indian AI Ecosystem

“We’re nowhere near global standards yet, but significant progress has been made within a year,” Aggarwal stated on X. “By open sourcing our models, we hope to foster collaboration within the Indian AI community to build a world-class Indian AI ecosystem.”

India’s Push in the AI Landscape

This initiative is part of India’s broader effort to establish itself in the artificial intelligence arena, which is currently dominated by companies based in the U.S. and China. The recent emergence of DeepSeek’s R1 “reasoning” model, developed with a relatively modest budget, has generated considerable attention within the technology industry.

Support for DeepSeek and Domestic Hosting

India recently acknowledged the advancements made by DeepSeek and announced that the country will host DeepSeek’s LLMs on its own servers. Krutrim’s cloud services began offering DeepSeek on Indian servers last week.

BharatBench: A New Evaluation Framework

Krutrim has also created its own evaluation framework, BharatBench, to assess the proficiency of AI models in Indian languages. This addresses a current deficiency in existing benchmarks, which predominantly focus on English and Chinese.

Technical Specifications and Performance

The lab’s technical strategy incorporates a 128,000-token context window, enabling its systems to process longer texts and engage in more intricate conversations. Published performance data indicates that Krutrim-2 achieved high scores in grammar correction (0.98) and multi-turn conversations (0.91).

From Krutrim-1 to Supercomputer Deployment

This investment follows the launch of Krutrim-1 in January, a 7-billion-parameter system that represented India’s first large language model. The deployment of the supercomputer, in partnership with Nvidia, is scheduled for March, with further expansion planned throughout the year.