IndiaAI allocates Rs 111 crore in GPU subsidies; Sarvam bags nearly Rs 99 crore

The IndiaAI Mission was approved by the Union Cabinet in March 2024

The IndiaAI Mission has disbursed over Rs 111.85 crore in subsidies under its GPU allocation scheme, aimed at making high-performance AI computing accessible to startups, researchers, and government entities.

According to data from IndiaAI, a total of 4,423 GPUs (graphics processing units) have been allocated so far under the programme.

The largest single allocation of both GPUs and subsidies was made to Sarvam AI's co-founder Pratyush Kumar, who received 4,096 NVIDIA H100 SXM GPUs through Yotta Data Services.

Sarvam was earlier announced as the first startup to have been selected for building India's foundational AI model under the mission.

The total bill of materials for this allocation stood at Rs 246.71 crore, of which the government subsidised Rs 98.68 crore—by far the largest subsidy under the initiative to date.

The amount of subsidy and allocation of GPUs are set to increase in the coming days, after the Ministry of Electronics and Information Technology (MeitY) selected three more startups — Socket.Ai, Gan.AI, and Gnani —for developing sovereign foundational models.

The scheme

The GPU empanelment initiative aims to provide affordable access to computing power for Indian AI developers and institutions.

While the scheme is open to early-stage startups, MSMEs, academia, and government bodies, the amount of subsidy varies widely, depending on the scale of the project, category of applicant, and type of GPU requested.

One early-stage researcher, Shilpa Singh, was granted an A100 40GB GPU with a total project cost of Rs 284, but received a subsidy of Rs 113.60.

Another startup founder, Manisha Singhal, was allocated an H100 SXM GPU with a total cost of Rs 341.20, but received no subsidy at all.

Government institutions also featured prominently among high-value allocations.

For example, a user identified as Shreevyas H M received a combined subsidy of over Rs 11 crore across multiple allocations of H100 NVL and H200 SXM GPUs through NxtGen Datacenter and Cloud Technologies.

In one instance, the project cost exceeded Rs 1.06 crore for a 12-GPU allocation.

Service providers such as E2E Networks Ltd, NxtGen Datacenter, and Yotta Data Services currently dominate the list of allocations.

According to the data, E2E, in particular, has served a wide range of categories, from early-stage startups and MSMEs to researchers and government entities, often supplying H100, L40S, and A100 GPUs at prices significantly lower than market rates.