Over 30 companies, including L&T-E2E consortium, Amazon Web Services (AWS), Jio, Google and Hiranandani-backed Yotta, have submitted bids for the Rs 5,000-crore AI compute infrastructure project under the Rs 10,000-crore IndiaAI mission, sources said.
Thursday was the last date for submitting bids. As part of the IndiaAI mission, the government will eventually create a capacity of about 10,000 graphic processing units (GPUs).
Simultaneously, the ministry of electronics and IT (MeitY) has received around 900 applications from startups that are seeking government’s support to utilise the AI compute infra, including GPUs, cloud services, data centre requirements, among others, at subsidised costs, officials said.
The government has finalised a viability gap funding model, under which it will subsidise compute costs by up to 50% for the startups that want to utilise GPUs. The tender is currently limited to creating a minimum capacity of 1,000 such GPUs, and has seen participation from data centre providers, cloud services providers or authorised partners of such solution providers.
“Both global and local companies have shown interest in the IndiaAI mission to create compute infrastructure. We are targeting to empanel the companies by the year end,” said an official. He added that the government is vetting applications of startups and academia to identify serious projects that require compute infrastructure support from the government.
Officials said it is possible to provide the compute support to startups by this year as a few companies such as Yotta and E2E Networks already have Nvidia GPU capacities with them.
Based on the applications received from AI compute infrastructure providers, Nvidia GPUs dominate the potential procurements, even as there are applications with AMD GPUs as well. According to officials, the next leg of procurement will happen based on the demand from startups, academia and researchers.
The AI mission essentially pertains to bringing AI compute capacity under the public-private partnership (PPP) mode through GPU-based servers, allocating early-stage funding to deeptech startups, setting up of innovation centres and developing broader AI sovereign infrastructure in the country.
The compute infrastructure is essential for startups, academia and researchers to test or run their large language models (LLMs) for their AI offerings.
In the absence of government support, procuring GPUs would be a costly affair for the companies as even for running basic tasks at least 100 GPUs are required.
As per the tender document, the selected bidders will have to make available an installed capacity of a minimum of 1,000 AI compute units within six months from the date of signing of the agreement.
Additionally, the empaneled companies are asked to submit purchase orders for the anticipated GPUs within three months of empanelment, or else the government will encash the bank guarantee submitted by companies with the proposal of application.
