Govt to finalise first set of proposals to build country’s foundational AI model within two weeks

The Centre has received almost 200 proposals for building India’s AI (artificial intelligence) foundational model. The first set of proposals are likely to be finalised within the next two weeks, said Abhishek Singh, the CEO of India AI Mission and additional secretary of the Ministry of Electronics and Information Technology (Meity).

“There was a lot of expectation (in India). China has come up with DeepSeek, when will India have a foundational AI model. So, we have got almost 200 proposals for building India’s AI foundational model. The first set of proposals are likely to be finalised within the next two weeks,” he said.

Singh said that the government is hopeful that these companies (who are working on the foundational model of AI) will soon be working on building a foundational model that will be trained in India.

Listing out the support that the government is providing to the overall AI ecosystem, Singh said that the Centre is providing fellowships to students and researchers for skilling, setting up fund of funds, working with accelerators and incubators for funding support.

“Government is trying to provide support at all levels, for datasets, for compute, for foundational model, and for building trusted AI to ensure that our startups and our researchers come up to the expectations and build something that we can truly call Indian and state of the art,” he said.

Gupta said that due to the government’s support in making graphic processing units (GPUs) accessible to startups and researchers, its cost has come down to Rs 67 per GPU per hour, which is less than a dollar.

“About 14,000 GPU are available at a very low cost, at Rs 150 per GPU per hour. Over and above, the government is giving 40% support. So the net cost of a GPU is less than a dollar, at Rs 67 per GPU per hour. It ensures that entrepreneurs, start-ups, and researchers who are working on AI models and applications can access that,” he said.

AI and deeptech requires patient capital, he said, adding that getting returns on such investments takes time. Hence, the government realises the need to provide compute infrastructure, data set, funding for foundational model and startups who are working in this field, he said.