TAIPEI (Taiwan News) —  US chip giant Nvidia is expanding ties with India and has released a Hindi-language AI model to tap into the local market.

While India has not been a traditional leader in the semiconductor industry like Taiwan and Silicon Valley, the country is building up its infrastructure to compete with major hubs, Reuters reported. Nvidia and other chip companies are focusing on India, seeking to capitalize on the country’s abundant talent pool of skilled engineers and vast market potential. 

Nvidia CEO Jensen Huang (黃仁勳) is in India this week for the company’s AI summit in Mumbai. "Nvidia is AI in India,” Huang said, referring to its computing infrastructure. “By the end of this year, we will have nearly 20 times more computing here in India than just a little over a year ago.”

Businesses in India are working to develop AI models incorporating the country’s diverse languages, and for use in tools such as AI customer service assistants and translation. Nvidia is partnering with India’s biggest technology giants, including Tech Mahindra, Infosys, TCS, Wipro, Reliance Industries, and Ola Electric. 

Nvidia will train developers to design AI agents using its software. Furthermore, Reliance Industries and Ola Electric will use the company’s "Omniverse" simulation technology to test factory plans in a virtual world.

In addition, Nvidia announced its small language model, “Nemotron-4-Mini-Hindi-4B,” with 4 billion parameters, that Indian businesses and startups can adopt for their AI models. Tech Mahindra was the first to use the new model to develop its custom AI model “Indus 2.0” with the Hindi language and its dialects. 

Unlike large language models, such as the one used for ChatGPT, small language models are more attractive for companies with smaller resources.

In September 2023, Nvidia announced a collaboration with Reliance Industries to develop AI supercomputers and large language models for India’s local languages. Nvidia and Tata Group also revealed similar joint AI infrastructure plans.