BharatGen, India’s government-backed AI consortium led by IIT Bombay, is set to be selected for the second phase of the IndiaAI Mission alongside seven other companies, marking a major milestone in India’s push for AI sovereignty. The announcement will expand the Mission’s foundation model initiative to 12 companies total, significantly boosting efforts to build open-source AI models tailored for Indian languages and contexts.
Eight New Companies Join India’s AI Revolution
Union IT Minister Ashwini Vaishnaw is expected to announce the second-phase selections next week in New Delhi, adding eight new firms to the existing four beneficiaries under the ₹1,500 crore IndiaAI Mission incentive program.
The new beneficiaries joining BharatGen include Fractal, Tech Mahindra, Avataar.ai, ZenteiQ.ai, Genloop, Intellihealth, and ShodhAI. These companies will join the original four selected firms: Soket AI Labs, gnani.ai, Gan.AI, and Sarvam.
BharatGen’s inclusion represents a significant government endorsement of the consortium’s work, which spans multiple IIT campuses, including IIT Bombay, IIT Kanpur, IIT Mandi, IIT Madras, IIT Hyderabad, IIIT Hyderabad, and IIM Indore.
BharatGen’s Breakthrough AI Model
In May 2025, BharatGen launched Param-1, India’s first foundational large language model built entirely from scratch with 2.9 billion parameters. The bilingual model features 25% Indic language data – dramatically higher than global competitors like Meta’s Llama, which contains just 0.01% Indian language content.
Professor Ganesh Ramakrishnan at IIT Bombay, who leads the BharatGen team, also released 20 speech models across 19 Indian languages on the AIKosha platform, specifically targeting voice-first interfaces for India’s diverse linguistic landscape.
The open-source model was pre-trained on five trillion tokens in English and Hindi, demonstrating India’s capability to develop sovereign AI systems that understand local contexts and cultural nuances better than foreign alternatives.
Also Read: US-India Trade Talks Advance with AI and Tech Focus as Trump Plans Modi Meeting
Fractal’s Mathematical Reasoning Breakthrough
Among the other selected companies, Mumbai-based Fractal made headlines with its open-source model Fathom-R1-14B launched in May. The company claims this model delivers mathematical reasoning performance that surpasses OpenAI’s o1-mini and o3-mini models, approaching o4-mini levels at a remarkably low post-training cost of just $499.
Fractal’s achievement is particularly significant as the company is slated for an IPO later this year, making it one of the most commercially advanced AI firms in India’s sovereign AI ecosystem.
Massive Response to AI Sovereignty Call
The IndiaAI Mission has generated unprecedented interest from India’s tech community. Starting with 67 applications by February 2025, the program received 120 additional applications the following month, with the total now reaching over 500 applications from Indian and global startups and researchers.
This overwhelming response demonstrates the strategic importance Indian companies place on building indigenous AI capabilities rather than relying on foreign models and infrastructure.
GPU Supply Challenges Emerge
Despite the program’s success, GPU availability has become a critical bottleneck. Out of the planned 34,333 GPUs under the IndiaAI Mission, only 17,374 GPUs are currently installed and operational, creating supply constraints as more companies seek subsidized compute access.
A data center executive told Economic Times: “We are facing a chicken-and-egg story. Until recently, Indian startups had no access to affordable GPU compute. The government stepped in to address that, and suddenly we saw a surge of interest from startups building AI models. But now we’re back to a supply crunch, since many GPU purchase orders exist only on paper”.
The government is now considering prioritizing inference-efficient hardware alongside training capacity in future GPU procurements to address this challenge.
Vision for Global Leadership
Minister Vaishnaw has repeatedly emphasized that the goal for each selected team is to become a “global top-five player” in their chosen AI sector, whether in multilingual foundation models, speech AI, or multimodal applications.
BharatGen is particularly well-positioned for this goal, with its multimodal capabilities spanning text, speech, and vision across 22 Indian languages. The consortium has already developed applications in agriculture, governance, and defense sectors, with plans to deploy these solutions across all states and districts once fully operational.
The selection of these eight additional companies under the IndiaAI Mission represents India’s accelerating commitment to building a comprehensive AI ecosystem that serves local needs while competing globally. With computing infrastructure, talent development, and startup financing all aligned under the Mission, India is positioning itself as a major force in the global AI landscape.
