India’s ?235 Crore Answer to DeepSeek ????

India isn’t sitting ducks on AI anymore. Amidst the debate about the country being the use case capital of AI, the government has received 67 proposals under the IndiaAI Mission, 20 of which are focused on LLMs. However, one initiative that truly stands out is BharatGen.?

By Mohit Pandey

BharatGen, funded by the Department of Science and Technology (DST), is currently India’s best bet for homegrown AI, especially for developing multilingual and multimodal LLMs tailored to our needs. Leveraging cost-efficient computing and top talent from IITs, the initiative is built for Bharat, by Bharat, at an investment of ?235 crore ($27 million).

With 50-60 researchers and a growing network of student contributors, BharatGen is scaling fast. Discussions with the government about GPU allocations and infrastructure are already underway.

A National Mission

BharatGen is designed with a whole-of-government approach, ensuring seamless inter-ministerial collaboration. “The first goal was to make BharatGen a national mission—done. The next is to ensure all intellectual property remains with India,” Ganesh Ramakrishnan, head of BharatGen, explained.

Union minister Ashwini Vaishnaw recently announced that India will have its own LLMs within 7-8 months—and BharatGen is a key part of this vision.

“India is building its own foundation model—not just Indian companies,” Ramakrishnan affirmed. “We have made significant technical progress,” he added.?

What Makes it Unique?

Unlike private AI startups like Sarvam and Krutrim, BharatGen has a strong academic foundation. “Private players operate in silos with no academic connection. How sustainable is that?” Ramakrishnan questioned. Private AI firms prioritise profit, while BharatGen ensures inclusivity, focusing on Indian dialects, agriculture, and public services.

Unlike DeepSeek, which struggles with scaling and inference costs, BharatGen is funded for the long haul. It has already rolled out Bharat Datasagar, a multilingual AI research repository, and e-VikrAI, a vision-language model tailored for Indic e-commerce.

“This is not about competition—it’s about necessity. India cannot afford to depend on foreign AI models. It’s like nuclear power—we need our own AI infrastructure.”

Click here to read the full story.

Meanwhile, China is Innovating Fast

DeepSeek has kicked off its Open Source Week by releasing FlashMLA, a decoding kernel designed for Hopper GPUs. The Chinese AI startup has a collection of 14 open-source models and repositories on Hugging Face, and more will be coming soon.

Meanwhile, Alibaba has announced its plans to invest over $52 billion in cloud computing and AI over the next three years, besides a new reasoning model that may be coming soon.

The India vs China AI race is clearly on. ??