Global giants take AIm at India’s GPU ambition

Top chip companies, cloud giants and cyber security firms have shown interest in the government’s Rs 10,000 crore graphics processing units (GPUs) procurement tender, central to India’s Artificial Intelligence Mission. Nvidia, Intel, Advanced Micro Devices, Qualcomm, Microsoft Azure, Amazon Web Services, Google Cloud and Palo Alto Networks were among the scores of companies that attended a pre-bid meeting on Thursday.

The ministry of electronics and information technology (MeitY) termed the participation “exceptional” and said it will extend the last date for bids by two weeks from the date of publication of clarifications to the tender. It got around 500 suggestions on the tender from those who attended the meeting, said MeitY additional secretary Abhishek Singh.

Also present at the meeting were representatives from Yotta, L&T, E2E, Tata Communications and Vigyanlabs, among others.


As per the terms of the Request for Empanelment (RFE), the companies are either required to show the government a purchase order of 1,000 GPUs or have a stock of 1,000 GPUs in order to be eligible to bid.

Several issues were raised by global GPU manufacturers, cloud service providers and data centre operators and MeitY has told them it will take these points into consideration and issue a clarification, besides extending the deadline from the current September 6, said people aware of the matter.

“There is lots of enthusiasm and interest in the India AI Mission,” Singh said. “We have received a huge response to the RFE. As is the process with any government tender, we will give two weeks from the date of publication of clarification for companies to submit their bids. Companies have till Saturday (August 31), 5 pm, to submit their suggestions.”

One of the key concerns was that RFE favours big companies, especially the teraFLOPS specification, referring to computing power. (One TeraFLOPS is equal to one trillion floating-point operations per second.)

Singh said the RFE is original equipment manufacturer-agnostic and both small and large players have been accorded equal opportunity.

“One of the suggestions given in the meeting was that there should be a category for startups and micro, small, and medium enterprises (MSMEs) who don’t have prior experience to participate in the tender,” he said.

The companies that participated in Thursday’s meeting told ET that if the government wants to facilitate 1,000 GPUs’ time for startups, academia and research organisations under the India AI Mission, then even low-range GPUs that can cater to smaller workloads should be included in the RFE, not just high-end and mid-range GPUs.

The companies said end users should have a choice among various GPUs and should not be restricted to one kind. Singh said that choice will be available.

As per the scoring pattern mentioned in the RFE, if the average annual turnover for the last three years of a company is greater than Rs 100 crore and less than Rs 150 crore, then five marks will be awarded to the bidder. For bidders with turnover greater than or equal to Rs 150 crore and less than Rs 200 crore, 15 marks will be awarded. A turnover greater than or equal to Rs 200 crore will yield 25 marks.

This may require smaller companies to partner with others, some of the attendees said.

Singh said the minimum turnover requirements are as per General Financial Rules. Companies need to have the necessary financial strength and capability to be able to implement projects of such complexity, he said.

“MeitY has released a visionary RFE for AI/GPU cloud services providers,” said Sunil Gupta, chief executive of Yotta Data Services, which has placed a purchase order of 16,000 H100 GPUs from Nvidia and went live with 4,000 H100 GPUs a few months back. This will help grow the AI ecosystem in India by making GPU capacity available in India in a cost-effective manner, he said.

Equally important is the way the empanelment framework has been structured, he said. It will encourage multiple service providers in the AI ecosystem such as data centres, cloud providers, managed service providers and system integrators to join hands, form consortiums, invest in related infrastructure and make AI GPU computing capacity available to the Indian market, thus enlarging the ecosystem, he said.