The government has reduced the revenue threshold for companies bidding on the Rs 10,000 crore GPU tender, addressing a key concern of the industry. This move will enable smaller enterprises to participate, with the revenue threshold now set at Rs 50 crore, down from the previous Rs 100 crore, as clarified by the Ministry of Electronics and Information Technology (MeitY) on Friday. Smaller cloud companies had advocated for this change, as the previous threshold would have required them to form partnerships to qualify.
Tarun Dua, CEO of Delhi-based AI cloud provider E2E Networks, welcomed the change, noting that the relaxed revenue qualifications now allow smaller companies to bid directly. However, he mentioned that some new clauses need further clarification.
The updated rules require bidders or their primary partners to have had an average annual turnover of at least Rs 50 crore over the last three financial years or since their registration if they've been operational for less than three years.
Sunil Gupta, CEO of Yotta Data Services, which recently placed an order for 16,000 H100 GPUs from Nvidia, agreed. He said the government has made participation easier for a diverse range of players, including startups, by lowering turnover thresholds and adding flexibility in AI compute infrastructure requirements.
Sumant Parimal, founding partner of Innogress Ventures, noted that this clause poses a risk that only large bidders with turnover between Rs 200-500 crore may be able to take on. While the relaxed turnover criteria may enable more participants in the AI mission bid, Parimal believes that few startups will qualify.
“This bid is not geared toward SMEs or tech startups; it’s more suited for larger corporations and Big Tech,” he added.
Gupta, however, praised the decision to give extra credit to early adopters of AI infrastructure, saying it rewards entrepreneurship. Companies that took the risk to create large GPU bases will now earn additional marks, helping them to contribute more significantly to India’s AI mission.
This clause awards one mark for every 20 AI compute units above 50, up to a maximum of 15 marks. An additional five marks are awarded if at least half of the compute units have double precision (FP64) capabilities.
Gupta emphasized the importance of the Make in India initiative, which the government has prioritized as part of the tender process. He believes this focus on AI infrastructure and sovereignty will be crucial for India’s tech manufacturing sector.
Major tech firms including Nvidia, Intel, AMD, Qualcomm, Microsoft Azure, AWS, Google Cloud, and Palo Alto Networks have shown interest in the Rs 10,000 crore GPU procurement tender, which is integral to India’s AI Mission. Cloud providers and data center operators such as Yotta Data Services, E2E Networks, Tata Communications, and CtrlS also participated in a pre-bid meeting on August 29.
Executives said the huge interest in the tender stems from the lucrative commercial opportunities in India’s booming AI cloud market. Companies are positioning themselves to take advantage of the anticipated surge in demand for GPUs.
On September 11, ET reported that small and medium-sized companies were rushing to form partnerships with each other and larger enterprises to participate in the tender. Under new guidelines introduced by MeitY, bidders must also comply with the Make in India initiative and adhere to sustainability standards, including maintaining a Power Usage Effectiveness (PUE) ratio of no more than 1.35 in data centers delivering AI cloud services.
An executive from Tata Communications confirmed receipt of MeitY’s corrigendum and said the company is reviewing the new requirements. The deadline for bid submissions has been extended to October 16 from the earlier date of September 25.