NVIDIA is deepening its push into a crucial growth market by entering a slate of partnerships with Indian companies to deploy its AI chips and technology.
Speaking to reporters, CEO of US-based chip designer Nvidia Jensen Huang said that his company has teamed up with India's most valuable firm, Reliance, to build infrastructure for AI applications in India. He also announced that Tech Mahindra would develop Indus 2.0, an AI model in Hindi, using Nvidia's chips and software. Infrastructure providers Tata Communications and Yotta Data Services also said they had plans to buy tens of thousands of Nvidia H100 chips by year-end.
Huang was speaking at the company's continuing AI Summit event in Mumbai, in a season when Indian technology service providers are in a rush to build up their AI capabilities. Infosys, Wipro and other IT companies have already started using Nvidia's software for developing custom AI applications for corporate clients.
"In the future, India will be a country that exports AI, not software," said Huang.
Wipro said that it has trained 225,000 employees on AI platforms by Nvidia, while Tata Consultancy Services said that it has trained 50,000 staff as AI associates. More than 500,000 developers in India are now joining Nvidia's developer program, the company said.
Flipkart and Zoho, software providers, will also use Nvidia's technology to build large language models in Hindi.
More partnerships will push those existing ones with Reliance Industries to build large language models for Indian languages. That deal included plans for AI cloud infrastructure and training for employees, too.
Few Indian startups are also using Nvidia's technology: CoRover.ai said its chatbot for Indian Railways handles 150,000 queries in multiple languages every day, and claims it has facilitated more than 10 billion customer interactions since launch.
To compete with its main rivals, AMD and Intel, over India's burgeoning AI chip market, the American graphics chip giant Nvidia has trained more than 100,000 Indian AI developers.