New Delhi: Union IT Minister Ashwini Vaishnaw announced that under the IndiaAI Mission, 18,693 graphics processing units (GPUs) have been empanelled to strengthen AI research and development.
Additionally, he highlighted that with this achievement, the country is embarking on the development of its own large language model (LLM) under the ambitious IndiaAI Mission.
This decision comes shortly after a Chinese AI lab introduced the low-cost foundational model, DeepSeek, signalling the need for India to develop its own homegrown solutions to address its unique linguistic and cultural context.
As part of the initiative, the government has selected 10 companies to supply a total of 18,693 high-end GPUs, essential for training machine learning models. These GPUs will be used to develop foundational models that power AI applications.
The minister noted that approved by the Prime Minister in March 2024, the IndiaAI mission is already making notable progress. The first major pillar of this mission, the ‘common compute facility,’ has been established as of August 2024. It aims to empower startups, IT companies, and research institutions in India to advance AI capabilities, with a particular focus on creating foundational models and LLMs tailored to India’s needs.
A cornerstone of this initiative is the empanelment of high-end GPUs, a vital resource for AI research. Vaishnaw revealed that the IndiaAI mission set a target to empanel 10,000 GPUs, but far exceeded expectations.
In total, 18,693 GPUs have been empanelled, including 12,896 H100 GPUs, 1,480 H200 GPUs, and models like MI 325 and MI 300x. These GPUs are among the most powerful in the world, marking a significant achievement for India’s AI infrastructure.
Vaishnaw also pointed to the DeepSeek project, which has already been trained on 2,000 GPUs. He emphasized that the development of India’s own foundational models and sector-specific applications is now possible, thanks to the enhanced computational power provided by this initiative.
The IndiaAI mission is not just about technological infrastructure, but also about democratizing access to AI tools and fostering inclusive growth. “This is a reflection of our Prime Minister’s commitment to democratizing technology for inclusive growth,” said Vaishnaw, emphasizing that the initiative will allow Indian startups and academic institutions to develop AI models tailored to solve pressing national issues.
With the goal of fostering innovation, the IndiaAI mission has officially launched a call for proposals from startups, IT services companies, and academia. The aim is to generate foundational models, sector-specific solutions, and AI-driven applications. Proposals will be evaluated by a technical body based on rigorous criteria, similar to the evaluation structure used for India’s semiconductor mission.
“Every proposal that we receive, will be reviewed by our technical committee. This will ensure that only the most promising, technology-driven projects are funded,” Vaishnaw added.
The mission is structured around four key pillars, with applications, safety, and research playing crucial roles in its growth. The second pillar, focused on applications, has already seen the approval of 18 AI applications across agriculture, learning disabilities, and climate change. These applications are expected to address significant challenges faced by the population at scale, with many more projects in development.
Furthermore, the IndiaAI mission is taking a proactive approach toward AI safety. The establishment of an AI Safety Institute, a virtual hub-and-spoke initiative, is aimed at addressing concerns related to AI privacy, algorithmic integrity, and broader ethical issues. Eight projects have already been approved, with five more open for proposals focused on privacy and algorithmic purity.
With these advancements, India is positioning itself as a global leader in AI, with a clear focus on inclusive, impactful growth powered by cutting-edge technology and innovation.
Also Read –
Budget FY25-26: What India’s Telecom, IT and Tech Industry Want from Nirmala Sitharaman
Discussion about this post