India will soon host an open-source artificial intelligence (AI) model similar to China's DeepSeek on domestic servers, the country’s Information and Broadcasting Minister Ashwini Vaishnaw was quoted as saying by Indian state owned DD News.

The initiative is part of the country’s broader push to expand AI capabilities and provide local researchers and startups with greater access to computing power. The Indian government has already finalised the server infrastructure and required capacity to deploy these models, Vaishnaw stated. The move follows the emergence of Chinese startup DeepSeek’s reasoning model R1, which challenges the prevailing notion that building high-performance AI models requires vast quantities of GPUs.

To support AI development, the Indian government is setting up a portal that will enable startups and researchers to access GPUs at significantly reduced costs. Additionally, students using common compute facilities will receive a 40% subsidy, bringing their costs down to less than INR100 ($1.20) per hour. India currently has around 10,000 GPUs operational of 15,000 high-end units integrated into its AI infrastructure - out of a total of 18,693 nationwide.

Vaishnaw highlighted that DeepSeek was trained using over 2,000 GPUs, while OpenAI’s ChatGPT was built using approximately 25,000. He added that technical partners have already begun collaborating and investing in India's AI initiative.

The government’s AI strategy aims to leverage the technology for large-scale applications, including healthcare, education, agriculture, logistics, and weather forecasting. “Our focus is on using AI to address population-scale challenges,” Vaishnaw said, noting that AI-powered tools for flood and glacier measurement are among the key applications under development. India’s central government will provide AI compute power subsidies for the next four years. Vaishnaw said the subsidy scheme is designed to be sustainable, with expectations that increased availability will drive prices down further.