India To Develop Own GPUs In Three To Five Years, Launch AI Foundational Model In Ten Months: Ashwini Vaishnaw

Union Minister Ashwini Vaishnaw announced on Tuesday that India is likely to develop its own advanced computing chipsets, or GPUs, within the next three to five years while a local foundational AI platform expected in the coming ten months, news agency PTI reported.

At the Budget Roundtable 2025 organised by India Today and Business Today, Vaishnaw said that the government would soon provide 18,000 high-end GPU-based compute facilities for AI development to entities across the country and projected the launch of an indigenous AI platform within 10 months.

"We are working on multiple, actually three options, where we take a chipset which is at some reasonable level available in open source or available as a licensed thing, and then build upon that to build our own GPU. That's the approach the entire world has followed and that approach will be able to give us India's own GPU in the time frame of three to five years," Vaishnaw said.

Originally, GPUs (graphics processing units) were utilised for multimedia content processing, which demanded substantial computing power, such as in gaming and video processing.

The demand for GPUs has surged globally due to increased AI applications, with the US company Nvidia holding over 80 per cent of the market share.

The minister noted that numerous startups have efficiently developed AI models, albeit smaller than ChatGPT.

He said that Indian Railways improved confirmed ticket rate by 27 per cent using AI models.

"We already embarked 18,000 GPUs, very high-end GPUs, and out of that, 10,000 are already available. So this 18,000 compute power will be rolled out in a couple of days. The tender process got completed last week, and in another couple of days, 3-4 days, this will be rolled out," Vaishnaw said.

He emphasised that while high-end computing infrastructure, which is a basic requirement for developing AI models, is typically costly, the government has established a mechanism to provide affordable access to this critical infrastructure.

"The researchers, startups, academicians, colleges, IITs, all of them can have access to this compute power, and they can start foundational models," Vaishnaw said.

Responding to queries about the timeline for India's AI foundational model, Vaishnaw said "10 months is the outer limit."

He noted that numerous research papers and mathematical algorithms, like those used by Chinese AI company DeepSeek, have significantly streamlined the process.

Also Read: After Olympic Bid, India Eyes Hosting 2030 Commonwealth Games: Report