India is buying a lot more chips. The Ministry of Electronics and Information Technology has floated a fresh tender for 10,000 additional AI compute units — GPUs, TPUs and similar accelerators — to grow the country's national pool of AI computing power that startups, researchers and academics can rent at subsidised rates. It is the latest concrete step in India's push for sovereign AI capacity.
The tender has been issued through the Centre for Development of Advanced Computing (C-DAC), the implementing agency for the IndiaAI Mission. A Request for Empanelment invites AI service providers to supply the compute units, with the paperwork allowing capacity to scale beyond 10,000 if demand from ministries, academia and industry warrants it.
A public stockpile of compute
The new capacity sits on top of more than 38,000 AI chips already onboarded under the IndiaAI Mission, a programme funded at over Rs 10,000 crore across five years. The Request for Empanelment is not a one-off purchase order: empanelled providers stay on the roster, and C-DAC can draw down more capacity during the empanelment's validity as demand arrives from eligible users.
The technical conditions are telling. Each cluster must offer at least 512 AI compute units at a single data centre inside India, with capacity spread across multiple locations. The stated purpose is dedicated compute for ministries, departments, academia, research institutions, startups and industry — a shared national resource rather than capacity reserved for a few large players.
Why subsidised compute is the strategy
The scarcest resource in modern AI is not talent or ideas but compute — the specialised chips needed to train and run models. The use cases listed in the tender spell out the ambition: large-scale model training, fine-tuning, inference and high-performance computing, the full lifecycle of building and running AI systems.
A national pool lets Indian startups and researchers rent serious computing power cheaply, instead of depending entirely on foreign cloud providers. That is the whole bet: turn computing from a rich-company luxury into public infrastructure.
The delivery problem it's trying to dodge
The tender also reads as insurance. The mission currently has access to around 30,000 GPUs against commitments of 45,000, with some private providers unable to deliver their pledged capacity amid a sharp rise in equipment costs, according to official sources cited in the Indian press.
Alongside this bid, the government is working on a 3,000-GPU install with C-DAC and an 8-exaflop supercomputer project developed with Abu Dhabi's G42 and Cerebras in partnership with C-DAC and the Mohamed bin Zayed University of AI, which would be owned by the Indian government as a gift from the UAE government. A bigger public pool means India's AI ambitions lean less on private pledges that may not arrive.
For Indian startups, the message is straightforward: the country's public compute bet keeps getting bigger, and the queue for subsidised GPUs is about to get shorter.
By Hamza Faisal
