Indian data center firm Yotta to build $2b Nvidia-powered AI hub

Indian data center company Yotta Data Services announced plans to develop a US$2 billion AI computing hub near New Delhi, using Nvidia’s Blackwell Ultra chips.

The project includes a four-year agreement valued at over US$1 billion, during which Nvidia will set up a large DGX Cloud cluster within Yotta’s infrastructure.

The hub aims to be one of Asia-Pacific’s largest and is expected to be operational by August, with additional capacity from Yotta’s Mumbai facility.

The project aligns with increased investments by global cloud providers like Microsoft and Amazon in India’s AI data center capacity, amid rising demand for generative AI services.

The development also reflects shifts in the global supply chain for AI chips, influenced by US export controls.

Yotta, part of billionaire Niranjan Hiranandani’s real estate group, operates three data center campuses in India and is a partner of Nvidia in the country.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

Government policy and national AI ambitions may help underpin this deal

Indian government policy supports the investment through a 21-year tax holiday meant to draw data center projects 1.
Much of the added computing power is set aside for India’s national AI Mission, a government program to build and deploy domestic AI capabilities. The mission backs foundational language models such as Bhashini and Sarvam 2.
The deal may lift Yotta’s standing in the market. Yotta has about 10,000 advanced Nvidia GPUs, which News18 reported equals roughly 75% of India’s GPU compute capacity 3.
Gupta said the buildout would raise India’s compute capacity by nearly five to six times, easing a shortage that has limited startups and researchers 2.

India’s AI boom creates geopolitical shifts but faces physical limits

The project puts India in a bigger role in a changing global AI supply chain. It also makes the country more attractive to US tech partners as US export controls limit where advanced AI chips can ship 4.
The pace of growth runs into practical limits, including the electricity and water needed to operate large AI data centers.
India’s data center water use is forecast to more than double by 2030. Analysts expect 60-80% of the country’s data centres to face high water stress this decade, which raises long-term operating risk 5.