India may need up to 2 lakh GPUs as AI use scales to crores, says MeitY’s Abhishek Singh

Abhishek Singh, Additional secretary, MeitY, in conversation with Dr Nalin Mehta, Managing Editor, Moneycontrol.

• India needs 100,000 to 200,000 GPUs for AI expansion
• Government to incentivise private sector GPU investments
• Public funding to support AI in sectors like health and education

Did our AI summary help?

India will need to sharply scale up its graphics processing unit, or GPU, capacity as artificial intelligence adoption expands to millions of users, said Abhishek Singh, Additional Secretary, MeitY, outlining plans to boost compute infrastructure with private sector participation and public funding support.

In conversation with Dr Nalin Mehta, Managing Editor, Moneycontrol, at the Rising Bharat Summit, Singh said the country currently has 62,000 GPUs, which were sufficient for training the first set of AI models. “We have 62,000 GPUs. That was sufficient for training the first set of models,” Singh said.

However, he cautioned that the requirement would multiply as AI services reach a wider base of users.

“Given the size of our country, if 10-20 crore people start using these services for various purposes, we will need to upend the backend infrastructure. 100,000 to 200,000 GPUs will be the minimum we will need to meet the demand of India,” Singh said.

Singh said the government is preparing a policy aimed at incentivising private players to invest in computing infrastructure to support the broader AI ecosystem.

“We are coming with a policy which incentivises the private sector to invest in the computing infrastructure. Part of GPUs will be offered by the private sector, incentivised by the government for the larger AI ecosystem,” he said.

The move signals a hybrid model in which government-backed incentives would catalyse private capital to build out the hardware backbone required for AI deployment at scale.

Singh added that the government would also directly fund AI inferencing across key public-facing sectors.

“Inferencing in applications across key public sectors like agriculture, education, healthcare will be funded by govt in many ways,” he said.

For sensitive and strategic areas, the approach would differ. “In strategic sectors like defence, we will set our own GPU infrastructure within the govt datacentres,” Singh said.