Yotta spends $2 billion to deploy top Nvidia chips

Yotta Data Services is investing $2 billion (Rs 16,600 crore) to deploy Nvidia’s latest AI chips in its Noida hyperscale data center campus, expected to go live by August this year.

This marks the first-ever deployment of Nvidia’s Blackwell B300 graphics processing units (GPUs) in India to create one of Asia’s largest AI superclusters, Sunil Gupta, chief executive of Yotta, told ET.

Alongside this, Nvidia will anchor Asia’s first DGX Cloud supercluster within Yotta’s infrastructure, consuming nearly half of the new GPU capacity under a four-year contract valued at about $1 billion, Gupta said.


“Nvidia is creating one of Asia’s largest DGX Cloud clusters on our supercluster. They will deploy about 10,300 GPUs to serve their global APAC customers and run their own models and services,” he said.

A substantial portion of the remaining capacity will be dedicated to India’s national AI Mission. “There are more than 500 applications from startups to access affordable compute. Many have not received GPUs yet. There is huge pressure on capacity. This expansion will increase India’s compute capacity almost five to six times,” he said.

The infrastructure will support initiatives such as Bhashini, Sarvam, BharatGen and Soket, which are building foundational Indian-language AI models.

With this deployment, Yotta’s total GPU footprint will rise sharply. The company plans to scale from about 40,000 GPUs today to more than 75,000 GPUs over the next two years, supported by large-scale data centre expansion in Mumbai and Delhi.

Over time, Yotta aims to build up to 2 gigawatts of data centre capacity at its Mumbai campus, enough to host 25 large data centre buildings.

Yotta, which is preparing for an initial public offering, has already secured funding for the entire $2-billion GPU investment and is now planning to raise $1-$1.2 billion through pre-IPO and IPO funding.

“Our target is to raise $1-$1.2 billion. About $400 million will come from HNIs (high net worth individuals) and family offices, and the rest from global private equity funds,” Gupta said.

He added that large global investors, including marquee private equity firms, are in advanced talks, though names could not be disclosed at this stage.

Beyond infrastructure, Yotta is also building a token-based AI marketplace platform, allowing enterprises and developers to consume AI models via APIs instead of leasing raw GPU capacity.

“We are building a platform-as-a-service where enterprises can just consume models and pay by tokens. That will be the real enterprise offtake of AI infrastructure,” Gupta said.