Leading technology companies, including Nvidia, Intel, AMD, Qualcomm, Microsoft Azure, Amazon Web Services (AWS), Google Cloud, and Palo Alto Networks, have shown keen interest in the Indian government's  Rs 100 billion tender for acquiring graphics processing units (GPUs). This tender is a pivotal part of India’s ambitious Artificial Intelligence (AI) Mission.  During a pre-bid meeting, the Ministry of Electronics and Information Technology (MeitY) described the response as "exceptional."

Due to the overwhelming interest and numerous industry suggestions, the government has decided to extend the bid submission deadline by two weeks after issuing clarifications to the bidders.  

Abhishek Singh, Additional Secretary, MeitY, mentioned that around 500 suggestions were made regarding the tender during the meeting. Representatives from other major companies, including Yotta, L&T, E2E, Tata Communications, and Vigyanlabs, were also in attendance.  To qualify for bidding under the Request for Empanelment (RFE) guidelines, companies must either present a purchase order for 1,000 GPUs or prove they have a stock of 1,000 GPUs. Concerns raised by several international GPU manufacturers, cloud service providers, and data centre operators during the meeting prompted MeitY to consider extending the initial September 6 deadline to address these issues.  

Singh highlighted the enthusiasm for the India AI Mission, noting the significant response to the RFE. He explained that, following the release of clarifications, companies would have two weeks to submit their bids, with the deadline for suggestions set for August 31, at 5 PM.  One of the primary concerns discussed was the RFE's perceived favouritism towards larger corporations, particularly regarding the teraFLOPS computing power specifications (one teraFLOPS equals one trillion floating-point operations per second). Singh assured that the RFE is designed to provide equal opportunities for both large and small players.  

Suggestions were also made to create a separate category for startups and micro, small, and medium enterprises (MSMEs) without prior experience, allowing them to participate in the tender. Participants further recommended the inclusion of low-range GPUs, suitable for smaller workloads, under the India AI Mission to address diverse needs.  According to the RFE’s scoring criteria, companies with an average annual turnover of Rs 1 billion to Rs 1.5 billion over the last three years will receive five marks. Those with a turnover of Rs 1.5 billion to Rs 2 billion will receive 15 marks, while companies with a turnover of Rs 2 billion or more will receive 25 marks. Some attendees suggested that smaller companies might need to collaborate to meet these financial requirements.  

Singh explained that these financial requirements are in line with General Financial Rules, ensuring that only companies with adequate financial strength and capability can undertake such complex projects.  

Sunil Gupta, CEO, Yotta Data Services, praised the RFE, describing it as a visionary move for AI/GPU cloud service providers.

He highlighted Yotta’s recent purchase of 16,000 Nvidia H100 GPUs, with 4,000 already operational, as an example of the significant investment required to build India’s AI ecosystem.  Gupta also emphasised that the empanelment framework is designed to foster collaboration among various stakeholders in the AI ecosystem, including data centres, cloud providers, managed service providers, and system integrators. "This will help grow the AI ecosystem in India by making GPU capacity available in a cost-effective manner," Gupta added.  (BS &ET)