India’s Centre for Development of Advanced Computing is seeking suppliers to provide AI computing services for government departments, researchers and businesses, through a procurement process that would select qualified providers before placing specific orders.
Its request for suppliers, issued on 7 October, offers a detailed view of how a public buyer proposes to organise those services: identify the capacity, specify how it should be managed and make a provider responsible for delivery. Prices and individual contracts would follow in later tenders.
C-DAC, a scientific society under India’s Ministry of Electronics and Information Technology, places the procurement within the IndiaAI Mission. The document identifies academic institutions and startups among the intended users. The question for them is how the proposed capacity would become a service they can obtain and use.
The plan envisages up to 10,000 AI compute units, with provision to expand according to demand. This is a procurement category covering GPUs and other AI accelerators, rather than a count of identical machines. The figure describes the proposed scale; it does not establish how much capacity has been purchased or delivered. Speaking on 8 October, Electronics and IT Minister Ashwini Vaishnaw described government demand of about 10,000 GPUs and a planned five-year supply arrangement intended to give the industry confidence in demand, Moneycontrol reported.
For a research team preparing to train a model, the specification of the computing environment would matter as much as its headline size. The processors need to work with the memory, storage and network available to the team. Researchers also need software they can use and support when something goes wrong. Those requirements make the surrounding service part of the purchasing decision.
C-DAC’s proposed package covers that broader environment, including hardware, software, networking, storage, security and support. The providers would have to deliver the infrastructure and services needed to run AI workloads, with detailed requirements set during subsequent procurement.
One significant requirement is that capacity committed to C-DAC be identifiable, verifiable and dedicated to its needs. The deployment provisions seek to prevent other customers’ workloads from interfering with those resources. For a public buyer, this creates a way to specify which capacity a supplier is offering and how its availability can be checked.
The underlying infrastructure could be owned by the supplier, leased or placed in another operator’s data centre under a colocation arrangement. The supplier would remain responsible for operating and managing the complete service. The document also prohibits subcontracting the AI compute services to another provider. Access to somebody else’s facilities would therefore leave the supplier accountable to C-DAC.
Data location is another condition. The hosting requirements cover the systems that process data and those used to manage the infrastructure. Data, metadata, logs and backups must remain in India, as must disaster recovery arrangements. This extends the requirement beyond the location where a user’s main dataset is stored.
These provisions concern control over the contracted service: where resources and information reside, how they are separated and who answers for their operation. Their effectiveness would depend on implementation and oversight. They cannot, by themselves, establish independence from foreign hardware or software suppliers.
The document also calls for reports on availability, utilisation, incidents and performance, while leaving detailed service levels to later tenders. These records would give C-DAC a way to compare the service promised with the service delivered. For users, the practical value of that oversight would lie in whether problems are identified and addressed quickly enough to keep their work moving.
The route to a contract has two stages. First, C-DAC would assess suppliers’ eligibility and technical qualifications without requesting prices. Qualified providers would then be eligible for later procurement rounds through the Government e-Marketplace, India’s public purchasing platform. Those rounds would specify the work, technical requirements, service levels, financial bids and contractual terms.
Suppliers would need operational infrastructure meeting the prescribed requirements to participate in that second stage. C-DAC could allocate work to one provider or several, with the allocation method determined later. Admission to the supplier panel carries no guarantee of business, revenue, utilisation or a purchase commitment.
C-DAC has revised the timetable through two corrigenda. Corrigendum 1 (13 October) moved the pre-bid meeting from 14 to 22 October and the start of bid submission from 22 to 24 October; Corrigendum 2 (27 October) moved the submission deadline from 29 October to 13 November 2026 at 15:30 IST and the technical bid opening from 30 October to 14 November at 16:00 IST. Under this revised timetable, submissions for this empanelment remain open until 13 November. The corrigenda do not announce an empanelment outcome or any Stage 2 award. Readers should not assume bids closed on the original 29 October date.
Individual purchases would connect qualified suppliers with the demand C-DAC actually procures; their terms would determine the commercial outcome.
For prospective users, the practical tests are more immediate: when they could obtain a suitable configuration, how reliably it would run their workloads and what terms would govern access. The price paid to a provider and any fee charged to an end user are separate questions. This supplier-selection document does not establish the final user experience or cost.
Later procurement specifications and awards should make the services being bought clearer. Delivery and usage records would then show whether the arrangements meet users’ needs. That is the useful test of this plan: which organisations can obtain computing capacity, on what terms, and whether the service supports the work they need to do.