This announcement arrives in the context of increasing global competition for advanced AI hardware. Notably, the United States has set an annual import limit of 50,000 GPUs for India, highlighting the importance of domestic capacity-building.
Vaishnaw also announced the selection of three new proposals aimed at developing indigenous large language models (LLMs), supplementing the one project that was previously approved. These selections form part of the government’s wider initiative for AI self-sufficiency and innovation in the nation.
On Thursday, Vaishnaw stated, “This is a significant change, and AI is here to stay,” reaffirming the government’s commitment to “democratise” access to artificial intelligence and ready the country for rapid technological advancements.
The cumulative total of over 32,000 GPUs will greatly enhance India’s ability to develop and train AI models, particularly large language models (LLMs) and small language models (SLMs), which are fundamental to generative AI applications.
Open-source models, broader access
Vaishnaw mentioned that the government will back the development of foundational AI models as part of the AI Mission, making them open-source and widely available.
“Our aim is to ensure that AI tools and infrastructure are not monopolized but accessible to all innovators and developers,” he elaborated.
Initially, the government had empanelled companies to provide 18,693 GPUs, of which 14,000 are already in circulation. Jio Platforms and Tata Communications were key suppliers in this initial phase.
These initiatives are a part of a larger effort to limit reliance on foreign AI models and infrastructure, positioning India as a global hub for AI innovation, training, and deployment.