India has under 2% of the world’s $1 trillion worth of compute infrastructure, which is several times less than countries like the US and China which together have nearly 60%, a top Nvidia executive said on Monday. At the same time, there is an opportunity for India to become the ‘AI (artificial intelligence) factory of the world’ as more and more infrastructure becomes available in the country.
“If we can basically build the infrastructure in accelerated computing quickly and faster, research will happen, innovation will happen, and more importantly, you will add $1 trillion to the economy,” said Vishal Dhupar, managing director for South Asia at global processing unit (GPU) making giant Nvidia.
Speaking at the Startup Mahakumbh in the national capital, Dhupar highlighted that India contributes only about 2% of the world’s AI research, and that the paucity of compute infrastructure is the key determinant.
“There’s a direct correlation to that,” Dhupar said. “Indians who contribute to the research but are not based in India are contributing 12% of AI research, because there’s infrastructure available there. I think that itself signals why infrastructure is really, really important.”
While the US and China each earmark about 4% of their GDP towards research through infrastructure, India spends only a percent, he noted.
The government’s recent announcement regarding the India AI Mission, where it will invest in 10,000 GPUs to make compute more accessible, is a ‘great start’, Dhupar said.
“But more importantly than that, the Indian business houses have figured out that now, computing is not about data centres, but converting data centres into compute units that will produce intelligence. In other words, it’s becoming factories.”
This gives India the opportunity to go from being the ‘back office of the world’ to ‘the office of the world’.
Codifying Indic languages and culture in large language models (LLM) is also a ‘mammoth opportunity’ for India, Dhupar said.
Vision and good research are important for India to fulfil its AI ambitions, and for the latter, adequate infrastructure availability is ‘in progress’, Dhupar said, pointing to Nvidia’s recent partnership with data centre company Yotta Data Services to bring 16,000 GPUs to India.
“We’re going to also work with more business houses to bring it (compute infrastructure) here and cater for the sensitivities of this country, helping you to innovate and build here so that you can maximise the disruption that is needed in our country and hopefully add the trillion dollars,” he added.
Dhupar said that the Nvidia Inception programme for startups has over 1,600 Indian startups in its fold, of which over 400 are AI startups and 60 are generative AI startups.
‘Hanooman’, a series of open-source Indic language AI models named after Hindu deity Hanuman, is the result of one such project by Indian Institute of Technology-Bombay. Developed in partnership with Seetha Mahalakshmi Healthcare (SML), Hanooman can generate both text and speech in several Indian languages. The first version of the model is scheduled to be launched in late March.
