Hiranandani-backed Yotta, Tata Communications, Jio, CtrlS and E2E Networks were among 13 companies that presented their technical presentations before the ministry of electronics and information technology (MeitY) on Monday, to make available the AI compute infrastructure in the country, sources said.
These companies have submitted their bids under the Rs 10,000 crore IndiaAI mission to provide graphic processing units (GPUs) and cloud services to startups, academia and researchers.
The 13 shortlisted companies were selected from 19 bidders. Each was given 30 minutes to showcase their readiness in several key areas, including GPU capacity, self-service portal to manage applications of beneficiaries, the Make in India component in servers, and how soon they can provide the minimum GPU units.
An official involved in the process stated that technical scores were assigned based on the bidders’ readiness, proposed GPU models, and ability to deploy compute units immediately. MeitY plans to finalize the empanelment process within a week.
As part of the requirements, each shortlisted company has been instructed to create a dedicated portal where GPUs will be listed. Startups and researchers will be able to submit applications through this platform to utilise the AI infrastructure.
Post the technical round, the shortlisted companies will be evaluated on the commercials or the rate for compute units submitted by them. Based on that the lowest bidder will be selected.
Sources said there can be multiple lowest bidders also based on the models of GPUs proposed by them. For example, if a company has bid the lowest to make available the Nvidia H100 GPUs and other has bid the lowest to provide Nvidia L40 GPU or an AMD GPU, they will be the lowest bidders for their respective model categories.
The selection will prioritise the most competitive commercial bids, identifying the lowest bidders (L1, L2, L3, etc). The empaneled companies will need to offer their compute services at the price set by the lowest bidder.
Preference for compute infrastructure utilisation will be given to the lowest bidder, followed by the next lowest, ensuring fair and cost-effective access to resources. The empaneled companies will provide GPUs and related infrastructure, which will be available on an hourly basis to beneficiaries like startups, researchers, and academicians.
A major highlight of the IndiaAI mission is the creation of a robust AI compute infrastructure comprising 10,000 GPUs. This facility is crucial for startups and researchers to test and deploy large language models (LLMs) and other AI solutions. The government has earmarked Rs 5,000 crore from the total mission budget for this purpose.
As part of the empanelment requirements, companies must install a minimum of 1,000 AI compute units within six months of signing the agreement.
The government has earmarked `5,000 crore out of Rs 10,000 crore for AI compute infrastructure. It will implement a viability gap funding model, subsidising compute costs by up to 50% for startups, researchers, and academic institutions.
