India’s AI Ambitions: 34,000 GPUs and Three New LLM Startups Propel Self-Reliance

Published: 03 Jun 2025

Author: Precedence Research

Share :

June 1, 2025

To achieve self-reliance and participate in the AI supremacy race, the national Artificial Intelligence computing capacity of India has exceeded 34,000 GPUs entirely, according to a statement released by the Ministry of Electronics and Technology (MeiTY).  

While talking at an event held in New Delhi on 30th May, India's union electronics and IT minister mentioned its intent behind this initiative. Technology is a double-edged sword and thus can be used by anyone, irrespective of an individual’s agenda. Therefore, it has become essential to democratize technology and make it available to everyone with safety standards and transparency instead of giving access to only a few people with wealthy backgrounds.  

When the AI mission in India started, it was expected to achieve the target of 10,000 GPUs. However, owing to the dedication and enthusiasm of engineers and assistance, nearly 18,000 GPUs have already been deployed and are being used. Further, to support industry partners, an additional 15,916 GPUs have been added to the previous 18,417 empaneled GPUs. This initiative will offer a common computational AI platform for developing indigenous models and AI algorithms to train and infer them by considering the Indian context. 

Further, he added that 367 datasheets have already been shared and uploaded to AI Kosh. Such steps will help to reverse brain drain and the development of a comprehensive ecosystem comprising foundational models, high computing capacity, safety standards, and others. All these efforts are aimed at making a self-reliant and inclusive AI ecosystem within India. 

New Indian AI startups have been selected to build foundational models in India. SARVAM AI will build India's sovereign LLM ecosystem and develop an open-source 120 billion-parameter AI model to support governance and public service access by using use cases such as 'AI4pragati’ and others. Again, including Sarvam, three more startups have been selected with rigorous multi-stage evaluations done by experts. Soket AI, an open-source model, will optimize the nation’s linguistic diversity in sectors like defense, education, and healthcare. On the other hand, Gnani AI will build a one billion-parameter voice AI foundation model to provide real-time speech processing in different languages along with highly developed reasoning potential.