India's government has selected 20 indigenous foundation-model proposals for support and sanctioned more than 93 lakh hours of subsidised computing capacity, part of a push to cut dependence on foreign AI infrastructure. The disclosure came in a written Lok Sabha reply from Jitin Prasada, minister of state for electronics and information technology.

The numbers

The 20 proposals were picked from 506 applications under the IndiaAI Mission — 12 large multimodal models and 8 small language models. Intellectual property stays with the developers, not the government.

Some of these are already live: Sarvam AI's 30-billion and 105-billion parameter models and Indus, Gnani.AI's speech-to-speech model, BharatGen's multilingual models, and Avataar AI's video-generation model.

The infrastructure behind it

IndiaAI Mission was approved in March 2024 with a five-year outlay of Rs 10,371.92 crore, aimed at building domestic capability across computing, foundation models, applications, research, and responsible AI.

The government has empanelled 15 computing service providers and approved subsidised computing support for 237 projects, totalling 93.18 lakh GPU hours. It's also placed a purchase order for a roughly 1.1-exaflop AI computing system at the National Informatics Centre's data centre in Shastri Park, Delhi.

Eleven national-level hackathons have produced 62 AI prototypes so far, 20 of which are deployed in public-sector institutions. Fifty-eight AI centres of excellence have been approved across states and union territories — work has started on 22 of them, in 13 states and UTs. Thirteen more projects are working specifically on AI safety: bias, privacy, algorithm audits, explainability, machine unlearning.

The wider push

Alongside IndiaAI, the government is building related infrastructure. The National Mission on Interdisciplinary Cyber-Physical Systems, outlay Rs 3,660 crore, has set up 25 technology innovation hubs at academic institutions covering AI, robotics, cybersecurity, quantum technology, and fintech.

On hardware, 12 semiconductor manufacturing projects with committed investments of Rs 1.64 lakh crore have been approved under Semicon India, three of which have started commercial production. The Cabinet approved a follow-on Semicon 2.0 on July 15, outlay Rs 1.27 lakh crore, aimed at Indian-designed chips, fabrication and packaging units, and the materials and equipment supply chain around them.

What the government is admitting

The reply states plainly that India's AI computing ecosystem still depends on imported GPUs bought through approved service providers. IndiaAI and the semiconductor missions are framed as the mechanism for closing that gap over time — this is an acknowledgment that the dependency exists now, not a claim that it's solved.

The reply also cites Stanford's Global AI Vibrancy Report from November 2025, which ranked India third globally in AI competitiveness and second-largest contributor to AI projects on GitHub.

Source: Times of India — Govt backs 20 indigenous AI models, clears 58 centres of excellence