NEW DELHI: Union Minister for Electronics and Information Technology Ashwini Vaishnaw announced a major expansion in the country’s national AI infrastructure and talent development ecosystem. Speaking at the ‘IndiaAI – Make AI in India, Make AI Work for India’ event in New Delhi, the Minister said India’s national compute capacity has now crossed 34,000 GPUs, bolstered by the addition of 15,916 new units to the existing 18,417.

The announcement was accompanied by the selection of three more startups to develop indigenous foundation models under the IndiaAI Mission, following a rigorous multi-stage expert evaluation process.

Calling the expansion of GPU infrastructure a critical enabler for India’s AI ambitions, Vaishnaw reiterated the government’s commitment to democratizing access to technology. “Technology should not be left in the hands of a few. It is important that a larger section of society can access it, develop new solutions, and get better opportunities. That’s the philosophy behind the IndiaAI Mission,” he said.

The Minister urged the newly selected startups to aim for a top-five global position in their respective domains and highlighted the Mission’s progress across its key pillars: compute infrastructure, foundational models, safety standards, open datasets, and talent development.

As part of this effort, 367 datasets have already been uploaded to AI Kosh, India’s AI-specific open data repository.

The IndiaAI Foundation Model pillar, which focuses on building large-scale AI models trained on India-specific data, received over 500 proposals since the call was launched. On April 26, Sarvam AI was selected to build India’s sovereign LLM ecosystem, including a 120-billion parameter open-source model for public service use cases like Citizen Connect 2047 and AI4Pragati.

Now, three more proposals have been greenlit. Soket AI will build an open-source 120-billion parameter model optimized for India’s linguistic diversity, targeting applications in defense, education, and healthcare. Gnani AI will develop a 14-billion parameter multilingual Voice AI foundation model for real-time speech processing and reasoning. Gan AI will focus on building a 70-billion parameter multilingual model aimed at achieving superhuman text-to-speech (TTS) capabilities.

The compute expansion has been supported by seven industry partners who offered competitive commercials across different GPU categories. These include Cyfuture India Pvt. Ltd., Ishan Infotech Ltd., Locuz Enterprise Solutions Ltd., Netmagic IT Services Pvt. Ltd., Sify Digital Services Ltd., Vensysco Technologies Ltd., and Yotta Data Services Pvt. Ltd. This shared GPU pool will provide a common AI training and inference platform for startups, researchers, and government agencies.

As part of the IndiaAI Applications Development Initiative, the winners of the IndiaAI I4C CyberGuard AI Hackathon were also announced. Jointly organized with the Indian Cyber Crime Coordination Centre (I4C), Ministry of Home Affairs, the hackathon led to the development of AI-based solutions capable of interpreting handwritten FIRs, screenshots, and audio files to better classify cybercrime complaints and detect evolving crime patterns.

IndiaAI, an independent business division under the Ministry of Electronics and Information Technology (MeitY), serves as the implementing agency for the IndiaAI Mission. It aims to democratize the benefits of AI, strengthen India’s position as a global AI leader, and promote ethical and responsible AI use across sectors