With significant subsidies and large-scale GPU installations, India positions itself to develop foundational models and reduce dependence on foreign compute.
Indian government has installed over 17,300 graphics processing units (GPUs) as part of the ₹100 billion IndiaAI Mission, according to government officials and media reports.
This initiative aims to build a public-sector cloud computing platform for researchers and start-ups to train artificial intelligence models, including large language models (LLMs).
To facilitate this, 34,333 GPUs have been approved for procurement across the first two rounds of implementation. A review meeting held on 16 June 2025, chaired by IndiaAI CEO Abhishek Singh, assessed deployment progress, integration with the IndiaAI compute portal, and the allocation of GPU time to end-users.
During the review, cloud service providers (CSPs) such as Yotta, E2E Networks, and NextGen reported significant progress in installation. By contrast, Jio Platforms and CtrlS Datacenter Ltd. had not completed installation as of mid-June.
However, both companies remain within their contractual deadline of 7 August 2025, based on letters of intent issued on 7 February, which allow six months for deployment.
Under this scheme, four start-ups, ‘Sarvam’, ‘Gnani.AI,’, ‘GAN.ai’, and ‘Socket’, have received approval to develop foundational models. Among them, Sarvam, co-founded by Pratyush Kumar, secured the most significant allocation: 4096 NVIDIA H100 SXM GPUs from Yotta Data Services.
This project has been allocated ₹986.8 million in subsidy against a total cost of ₹2.46 billion, and other firms are expected to receive comparable support.
At the same time, the government announced a 100% subsidy on computing infrastructure specifically for companies developing foundational AI models.
Providing further clarity, the report by Moneycontrol states that the scheme provides a 100% subsidy specifically for foundational model development, while other AI workloads, such as inferencing and applications, remain eligible for a 40% subsidy.
Meanwhile, the third round of bidding for GPU capacity has concluded, and the submitted proposals are currently undergoing technical evaluation.
Taken together, these efforts reflect the government’s intention to strengthen domestic AI infrastructure and reduce reliance on foreign compute platforms.
