India’s Yotta Plans $2 Billion AI Hub with Nvidia Blackwell Architecture
Indian data centre firm Yotta Data Services announced on Wednesday it was developing one of the largest AI computing centres in the Asian continent using the newest Blackwell Ultra chip developed by Nvidia (NVDA.O), opens new tab in a project worth over 2 billion dollars.
It involves a four-year contract worth more than one billion dollars in which Nvidia will set up one of the largest DGX Cloud infrastructures in the region of Asia-Pacific in the facility of Yotta, the firm announced.
The relocation occurs at a time when international cloud providers such as Microsoft (MSFT.O) and Amazon (AMZN.O) are adding AI data centre capacity in India, as the demand for generative AI services grows and there is a trend to localise high-performance computing infrastructure.
The investment is also to be made amid US export controls that have reformed world supply chains of advanced AI chips, compelling companies to extend partnerships in markets of India.
The supercluster, which will become operational by August, will be implemented at the data centre campus of Yotta, which is located near the capital, New Delhi, with more capacity being supplied by its data centre in the financial hub of India, Mumbai.
Yotta, an Indian firm that belongs to the real estate company of Indian billionaire Niranjan Hiranandani, is an Indian partner firm to Nvidia and operates three campuses of data centres in Mumbai, Gujarat and in the vicinity of New Delhi.
The development follows a period of growth of global cloud providers such as Microsoft and Amazon in the capacity of their AI data centres in India, as pressure mounts on generative AI services and moves towards localisation of advanced computing infrastructure.
ET states that it will consume a significant portion of the remaining capacity with the DGX Cloud supercluster, though the national Indian AI Mission will consume some of the GPU power as well, and the Yotta infrastructure will support projects including Bhashini, Sarvam, BharatGen, and Soket, which are constructing Indian-language primaries of AI.
The publication reported that through this deployment, Yotta is going to increase its total GPU footprint by a factor of 3x through about 40,000 GPUs currently to over 75,000 units in the next two years.
The entire investment of the 2-billion-GPU worth has already been secured by Yotta, which is already planning to raise 1 to 1.2 billion via pre-IPO and IPO funding, which ET further reported.
