Nvidia backs India’s sovereign AI push with gigawatt-scale infrastructure

Nvidia has announced a slew of partnerships with India’s leading infrastructure and technology providers to help the subcontinent build up its sovereign artificial intelligence (AI) capabilities.

During the AI Impact Summit in New Delhi this week, the chip giant unveiled plans to deploy tens of thousands of its latest Blackwell graphics processing units (GPUs) across the country to support the government-backed IndiaAI Mission that fosters AI innovation by enhancing data quality, building AI talent and improving access to AI infrastructure.

For example, Nvidia is working with engineering giant Larsen & Toubro (L&T) to build a gigawatt-scale network of AI datacentres. These facilities aim to keep critical data and model training within India’s borders, a key requirement for government and regulated industries.

“AI is driving the largest infrastructure buildout in human history – everyone will use it, every company will be powered by it and every country will build it,” said Nvidia CEO Jensen Huang. “Together with L&T – an 88-year-old engineering and nation-building leader – we are laying the foundation for world-class AI infrastructure that will power India’s growth and help realise the full vision of India AI. ”

Meanwhile, cloud service provider Yotta has deployed over 20,000 Nvidia Blackwell Ultra GPUs to power its Shakti Cloud, hosted in datacentres in Navi Mumbai and Greater Noida, while E2E Networks will launch a Blackwell cluster on its AI and machine learning platform, hosted at the L&T Vyoma datacentre in Chennai.

The AI infrastructure in these datacentres will cater to the needs of model builders, startups, researchers, and enterprises that are building and deploying AI applications in India, as well as supporting initiatives such as BharatGen, a government-backed project to create a multilingual and multimodal AI ecosystem.

As part of the BharatGen project, a new 17-billion-parameter mixture-of-experts model has been built from scratch using Nvidia’s NeMo framework for pretraining and the NeMo RL library for post-training. The open-source model is designed to power applications across public services, agriculture, security and cultural preservation.

In addition, Nvidia is supporting efforts by India’s $250bn IT services sector to move beyond generative AI chatbots toward agentic AI systems capable of reasoning and executing complex workflows.

Major global systems integrators, including Infosys, TCS, Wipro, and Tech Mahindra, announced they are adopting Nvidia’s AI Enterprise software to build AI agents for global clients.

Wipro showcased a new system developed for a US health insurer that now handles 42% of inbound calls with a sub-200-millisecond latency, while Infosys has integrated Nvidia’s stack to create a 2.5-billion-parameter model for agent development, code generation, refactoring, and software engineering workflows.

“Agentic AI is reshaping India’s tech industry, delivering leaps in services worldwide,” Nvidia said in a blog post, noting that “technology leaders are tapping Nvidia Nemotron open models, data and software to build and deploy agentic and generative AI, accelerating productivity and efficiency.”

The summit also highlighted India’s growing role as a manufacturing hub, with industrial heavyweights Reliance Industries and Tata Motors adopting Nvidia’s Omniverse platform for building physical AI applications.

For example, Reliance is combining Siemens’ digital twin technology with Omniverse libraries for faster, more precise simulation and plant design for its next generation gigafactories, while Tata Motors is using digital twins built on Omniverse libraries along with other AI capabilities to convert camera feeds into intelligent sensors for automated quality checks and real-time safety compliance.

Nvidia said these partnerships aim to “bring physical AI into factories, warehouses and infrastructure”, allowing Indian manufacturers to reduce lead times and improve operational efficiency through high-fidelity simulations.