Nvidia deepens investment in India’s AI ecosystem

Nvidia is doubling down on efforts to expand its presence in India’s burgeoning artificial intelligence (AI) market by driving adoption of its compute platform and fostering collaborations with key players in the local technology ecosystem.

At the Nvidia AI Summit India this week, the company unveiled a series of partnerships and initiatives, highlighting how its technology is driving breakthroughs in various sectors and positioning India as a global AI leader.

Since establishing operations in India in 2004, Nvidia has set up design centres in multiple cities and upskilled more than 200,000 developers through its Deep Learning Institute, recognising India’s potential as a key player in the global AI arena.

“As we stand at the renaissance of the AI revolution, my country is excited to have an opportunity to embrace AI,” said Vishal Dhupar, managing director for South Asia at Nvidia, adding that India’s tapestry of culture and sensibilities can be encoded into tokens and help position the country as the “intelligence capital of the world”.

A central theme of Nvidia’s focus globally, as well as in India, is the development of so-called “AI factories” that leverage advanced chips, software libraries and networking capabilities to tackle complex challenges. At the summit, Nvidia announced key partnerships with Indian tech companies including Yotta, E2E Networks and Netweb to build AI factories for the Indian market.

For example, Yotta, whose Shakti Cloud is the first Indian cloud platform to offer Nvidia inference microservices (NIM), has launched an accelerator programme to support startups. On the hardware front, E2E is enabling access to Nvidia’s H200 graphics processing units (GPUs) and InfiniBand networking, while Netweb is producing Tyrone servers based on Nvidia’s MGX reference design.

Nvidia is also supporting Indian companies in building large language models (LLMs) tailored for India’s regional and linguistic requirements. For instance, it is working with Flipkart, Krutrim, Gnani AI, Sarvam AI and Zoho to build and manage the lifecycle of local LLMs using the Nemo framework and optimise model performance with NIM. These partnerships address the need for safer conversational AI, multilingual speech-to-speech capabilities, and AI-powered customer service.

In addition, Nvidia has developed the Nemotron-4-Hindi-4b model, a small language model tailored for Hindi language tasks. The model was built by pruning a larger Nemotron model while preserving accuracy and incorporating Hindi training data. Nvidia also used Nemotron to generate synthetic Hindi training data, addressing the challenge of limited data availability in sovereign AI development.

Anne Hecht, senior director of enterprise AI at Nvidia, noted that the Nemotron-4-Hindi-4b model is already being deployed by the Centre for Computational Brain Research at the Indian Institute of Technology (IIT) Madras to answer neuroscience-related queries, as well as IIT Delhi to uncover antibiotic prescription patterns in the fight against antimicrobial resistance.

With agentic AI architectures expected to be more widely adopted, Nvidia has introduced an NIM agent blueprint for customer service, enabling enterprises to create AI agents that can draw on short- and long-term memory, allowing for more natural interactions and multi-turn conversations with customers.

Hecht said Indian technology consulting giants, such as Infosys, Tata Consultancy Services, Tech Mahindra, and Wipro, are training nearly 500,000 professionals to build industry-specific agentic AI applications using NIM agent blueprints.

Nvidia is also championing the development of robotic systems in India. The company highlighted several Indian robotics startups in its Inception programme, such as Addverb, Ottonomy, and Ati Motors, that are using Nvidia AI, Omniverse and Jetson to develop logistics robots, last-mile delivery robots, and autonomous vehicles for factories and warehouses.