India invites bids for 10,000 GPUs under IndiaAI Mission

01.What happened

India's Centre for Development of Advanced Computing (C-DAC), which works under the Ministry of Electronics and Information Technology (MeitY), has issued a Request for Empanelment (RFE) inviting AI service providers to supply up to 10,000 AI Compute Units under the IndiaAI Mission. The tender defines an AI compute unit as a GPU, TPU, accelerator or similar hardware used for AI workloads, so it is not tied to a single chip maker or product.

The RFE was published on the government's eProcure portal on 7 October 2026 under the reference C-DAC/GPU/01/2026, and bids close on 29 October 2026. C-DAC is named in the document as the implementing agency for the mission's AI computing infrastructure. The empanelment approach means the government is building a list of approved providers rather than buying all of the hardware itself, and it can draw on that pool as individual users ask for capacity.

02.Who can use this compute

The capacity is meant for ministries and departments, academia, research institutions, startups, industry and other eligible users, based on the demand that C-DAC receives. The tender lists large-scale AI model training, fine-tuning, inference and high-performance computing among the workloads the platform should support. Officials have said the compute time will be offered to applicants such as startups, researchers and academia as they apply for it.

The RFE allows C-DAC to scale total capacity beyond 10,000 units during the empanelment period if demand requires it. Capacity is to be organised in clusters of 4,000, 1,000 or 512 units. Each cluster must hold at least 512 units at a single data centre in India, and the capacity is to be spread across multiple locations, although the tender does not name the cities.

03.Data must stay in India

The tender sets clear data-sovereignty conditions. All infrastructure, data, metadata, logs, backups and disaster recovery must remain within India, and providers must follow Indian data localisation rules on where data is stored and processed.

C-DAC's data has to be isolated physically and logically from other customers, and providers must prevent any unauthorised access, use, sharing or transfer of its data and other digital assets. In effect, the government wants a shared platform that many users can access at the same time without their work mixing with anyone else's, the ministry said.

04.Why India is buying compute now

The move comes as India works to ease a shortage of AI computing capacity. Under the Rs 10,372 crore IndiaAI Mission, officials say more than 38,000 GPUs have already been onboarded through three bidding rounds, and they project that the new bid will take the country's total past 50,000. Officials have also said the government is building a government-controlled pool of AI compute alongside capacity sourced from private providers.

Speaking on 8 October, Union IT and Electronics Minister Ashwini Vaishnaw said the government's demand estimate is for 10,000 GPUs and that the first tender is for 5,000. He described the plan as a long-term supply arrangement of five years so that the industry can be assured of demand, and said the government is preparing a roadmap for Indian GPUs for inference, with solutions expected in three to four years. Winning bidders will also be allowed to sell GPU capacity globally, a step officials said the industry had requested so that providers stay profitable until demand within India grows.

05.What it means for Indian readers

For students and researchers, the tender is a signal that subsidised, India-hosted compute is being built for public use, which matters because training and fine-tuning AI models is expensive and usually needs foreign cloud credits. For startups and small businesses, the five-year usage guarantee, and the decision to let winning bidders also sell capacity globally, are meant to keep prices workable while demand within India grows. That could lower the cost of experimenting with AI for a small firm that cannot easily afford to import compute from abroad.

The data-localisation rules matter for anyone handling sensitive information. If you run a small firm in health, finance or education, working with an empanelled provider could make it easier to meet requirements that data stay in the country. At the same time, the tender does not name cities, prices or how much will be bought in the first round, so the practical details are still to come. MeitY has separately said it will release a consultation paper on AI regulation within a month, which will shape how these rules develop.