IndiaAI Compute: Model Builders Got Rs 1,058 Cr, Researchers Rs 10 Cr

01.What the IndiaAI compute numbers show

The IndiaAI Mission was approved in March 2024 with an outlay of Rs 10,372 crore to build the country's AI ecosystem, and its largest component is subsidised access to GPUs. A new analysis by ThePrint of all 304 approvals listed on the IndiaAI Compute Portal, as updated on 8 October, shows who that subsidy is reaching and who it is not. Of the Rs 1,275 crore the Centre has committed so far, nine organisations building AI foundation models have received Rs 1,058 crore. Researchers, students and fellows, who were also meant to benefit, have received Rs 10.14 crore between them, or 0.8 percent.

The skew is sharpest at the top. Most approved users get 40 percent of their bill paid, but 30 approvals were paid in full, and they account for Rs 1,089.56 crore, or 85 percent of all subsidy. Sarvam AI's entry, which showed a subsidy of Rs 98.68 crore in 2025, now equals its full bill of Rs 246.72 crore.

02.Where the money actually went

ThePrint grouped the 304 approvals by type of recipient. Sixteen approvals to foundation-model builders make up 83 percent of the subsidy, and the government has paid 98.9 percent of their bills. One hundred and twenty-three approvals to government departments and public sector units account for 12.5 percent, with 46.8 percent of their bills paid. The 55 approvals to researchers, students and fellows account for 0.8 percent, with 39.7 percent of their bills paid.

Some individual entries stand out. A Bengaluru startup founded by an IISc professor, filed under the category Researchers or Academia, was approved for 2,080 Google TPUs worth Rs 162.86 crore, five weeks after the government had put the TPU capacity of the entire pool at 1,050. The next two largest grants, of Rs 177.27 crore and Rs 163.04 crore, are in the names of Gnani.ai co-founder Ananth Nagaraj and Soket AI founder Abhishek Upperwal, both picked in May 2025 to build foundation models. Government bodies including the CBI, the Rajya Sabha, the National Informatics Centre and Prasar Bharati received Rs 38.97 crore of free compute, nearly four times what all individual researchers together received.

03.Why researchers say this matters

Researchers ThePrint spoke to said the concentration leaves out university labs, which train the engineers these companies hire and work on AI outside language, in areas such as climate, health and materials. We need to get rid of language fixation, an IIT assistant professor said, speaking anonymously. Language is not the only frontier. Prathosh A.P. of IISc said that because of these infrastructure issues, good engineers tend to leave the country.

The wider context is stark. India had 50,460 AI authors and inventors in 2025, the second-highest number after the United States, according to Stanford University's AI Index 2026. Yet it also recorded the largest net outflow of AI talent of any country the report tracked. Globally, the cost of compute has already pushed model-building out of universities: of the notable AI models released in 2025, 93 came from industry and just two from academia.

04.How other countries handle academic compute

Other governments pay for researchers' compute more directly. In September, the US National Science Foundation set up an operations centre to make its National AI Research Resource permanent; its two-year pilot had given American researchers and educators compute at no charge. Britain's AI Research Resource awards projects up to 10,000 GPU hours on public supercomputers such as Isambard-AI in Bristol.

That contrast is the point of the comparison. Where the US and UK route free or heavily discounted compute to individual academics, India's portal so far has concentrated its largest and fullest subsidies on a small group of well-funded model builders.

05.What it means for India

The timing matters. The government is weighing letting GPU suppliers sell capacity abroad because, as an official told Business Standard, there is not enough uptake of GPUs in India right now. That is a sign the shared pool is not being used as widely as intended, and the data suggests one reason: the subsidy is flowing to a handful of model builders rather than to the wide base of students, researchers and small teams the scheme was sold as democratising.

For Indian students and early-career researchers, the practical takeaway is that cheap compute exists but is not automatic: you apply through the IndiaAI Compute Portal, you must fit a named category, and you still pay a share of the bill. For startups and MSMEs, the same portal offers H100, H200, A100, L40S and TPU capacity at up to 40 percent off. The larger question the researchers raise is whether a scheme meant to widen India's AI base can do that when 0.8 percent of its subsidy reaches the people it was meant to widen it to.