Nvidia widens India bet with sovereign AI, Blackwell clusters and giga factories

AT a clip when bharat is positioning itself as an rising planetary unreal intelligence (AI) data centre hub with significant investments on the agenda, tech giant Nvidia has announced a number of partnerships that span what the company calls a five-layer AI cake. At the India AI Impact Summit 2026, Vishal Dhupar, Managing Director, South Asia at Nvidia, calls AI essential infrastructure “just as electricity or the internet was in previous generations”.Nvidia’s three-pronged focus on India is defined by investments for the country’s AI mission, which include infrastructure expansion as well as frontier model development. Secondly, Nvidia is making a case for its Nemotron foundational models, as well as AI enterprise software, as the foundation for agentic and generative AI in enterprises across call centres, healthcare, software development and telecommunications. Finally, Nvidia is partnering with India’s biggest manufacturers to build AI factories using Nvidia AI hardware.Nvidia’s investments in India are set to be defined by a number of partnerships that span the cake, as Dhupar calls it, covering compute infrastructure, open models, enterprise AI agents and manufacturing. This comes as Ashwini Vaishnaw, Union Minister for Electronics and IT, again emphasised at the summit that, as part of the government’s AI Mission 2.0, there will be a need to ready large infrastructure for sovereign models.Nvidia has partnerships in place with Yotta, Larsen & Toubro, and E2E Networks for AI infrastructure projects, or large AI factories that drive innovation from India. “Yotta is augmenting its Shakti cloud powered by more than 20,000 Nvidia Blackwell Ultra GPUs. This provides large-scale, sovereign AI infrastructure to the nation,” Dhupar says. He also details L&T’s plans for new sovereign gigawatt AI factory infrastructure, with the roadmap including expansion in Chennai and a new facility in Mumbai. E2E Networks is building an Nvidia Blackwell cluster.“These developments will help our researchers, startups, and enterprises build AI models for India, and for the world,” Dhupar says, noting that these models provide a springboard for sovereign AI development within the country.Critical to Nvidia’s infrastructure push in India will be the company’s Nemotron open models, which are already being used by Indian AI developers including Sarvam, BharatGen and Chariot. Nvidia insists Nemotron isn’t just a model or a family of models (these have fared well in benchmarks for biomedical AI, climate, agentic AI and autonomous systems), but an entire toolkit with datasets, libraries and everything needed to build agentic AI.“Nemotron covers a wide range of capabilities, from vision and speech to document understanding and safety. We recently released the Nvidia Nemotron 3 Nano, a highly efficient language model. Model builders can use these open tools to create their very own sovereign AI models,” he says, confirming that following the success of Nemotron 3 Nano, the larger versions called Super and Ultra are coming soon.Estimates suggest India’s tech industry will generate $350 billion in revenue by 2030, with as much as $134 billion worth of investments lined up to extend manufacturing capacity across the automotive, renewable energy, construction and robotics industries. Nvidia says it is partnering with brands across these industries, including Tata Group and Tata Motors, Torrent Power, Havells and Power Grid Corporation of India, to deploy its CUDA-X libraries for high-performance computing, Nvidia AI for enterprises, and the Omniverse open platform for complex workflows.Nvidia confirms that as many as 800,000 developers in India are part of its developer programme, while more than 4,000 startups are listed as part of the Inception initiative in the country.