India, Nvidia in talks on sovereign GPU development

India is in talks with Nvidia to explore the development of a sovereign graphics processing unit (GPU) and local manufacturing of advanced data processing devices.

Ashwini Vaishnaw, India’s minister for electronics and information technology, said the discussions also included edge computing systems such as Nvidia’s DGX Spark, which is designed for AI applications and can operate without internet connectivity.

The talks focused on supporting secure AI model inferencing for use in sectors like railways, healthcare, and education.

Nvidia currently holds over 80% of the global GPU market and has seen high demand as companies and governments invest in AI infrastructure.

Vaishnaw shared the meeting details on social media, noting India’s interest in building domestic GPU capabilities and manufacturing advanced AI hardware locally.

🔗 Source: YourStory

🧠 Food for thought

Implications, context, and why it matters.

India’s GPU ambitions face a steep timeline and capability gap

Four new semiconductor projects approved in August 2025 total INR 46 billion and target compound semiconductors (materials such as Silicon Carbide used for power and radio chips) plus packaging facilities (final assembly and interconnect of chips), not advanced logic fabs (leading-edge processor manufacturing plants) used for high-end GPUs 1.
Ten sanctioned semiconductor projects total INR 1.60 trillion 1. Work centers on assembly, testing, and packaging (the post-fabrication stages known as backend) plus compound semiconductors, not advanced logic fabrication. New units now ship commercial chips 2, yet design and manufacturing depth for sovereign GPUs (domestically designed and manufactured processors under national control) still trails Nvidia.
Talks remain exploratory. GPU design and fabrication need advanced packaging first, which stacks or tightly connects multiple chips to act as one system. The 3D Glass Solutions site in Bhubaneswar is a materials and packaging firm facility for such interconnects 1. Full GPU sovereignty sits on a longer horizon.

Government AI compute expansion creates immediate opportunities for infrastructure providers

India has deployed 38,000 GPUs at about ₹65 per hour 3. This push creates room for system integrators (firms that assemble hardware and software into turnkey systems) plus data center operators (companies that run large-scale compute facilities) to partner as the IndiaAI Mission rolls out over its five-year, ₹10,371.92 crore budget 3.
Secure, offline AI for railways, healthcare, and education could trigger tenders for edge computing deployments (processing data on or near the device where it is generated) 3. Hardware OEMs (original equipment manufacturers) and AI software vendors can target air-gapped setups (physically isolated from external networks) in public sector work.
AIKosh, the government’s national data repository, hosts 5,500+ datasets and supports 11,000+ registered users 3. BharatGen AI, an Indian language foundation model, covers 22 Indian languages 3. Developers can build industry-focused tools for government and enterprise customers using these sovereign resources, which reduces reliance on foreign AI platforms.