US cloud provider Vultr has launched a 20 MW AI supercluster in Chennai, its largest deployment in India and one of its two largest AMD AI infrastructure deployments anywhere.
The numbers
- 20 MW of capacity in the Chennai facility.
- Around 10,000 GPUs in the initial configuration, AMD Instinct MI355X.
- March 2027 is when Vultr plans to have the cluster operational, according to the Times of India.
- 600 MW of AI capacity in India targeted by 2028, and 1 GW by 2029, though the company has not committed to firm timelines.
- Up to 5,000 people could eventually work at its Chennai operation as the facility comes online and services expand.
Vultr hasn't disclosed the investment for Chennai, or for its existing data centre operations in Delhi-NCR, Mumbai and Bengaluru. In 2027 it plans to add AMD's MI455X GPUs and the Helios rack-scale architecture.
Who it is for
The cluster is built for large language model and foundation model training, aimed at AI labs, startups, enterprises, research institutions and universities. Vultr CMO Kevin Cochrane said India is the company's largest market outside the US.
Vultr's pitch is price. Piyush Gupta, its vice president for India, APAC and the Middle East, told the Times of India the company prices cloud infrastructure 50% to 90% below traditional hyperscalers.
The StartupTN piece
Vultr has also signed an MoU with StartupTN to give eligible Tamil Nadu startups up to $45,000 in cloud computing credits plus preferential pricing.
My read
Chennai keeps collecting AI compute; L&T's Vyoma and CapitaLand are already building here. What sets this one apart is that it's an AMD deployment in a market where most of the conversation has been about Nvidia. For Tamil Nadu startups, the StartupTN credits are the benefit you can use today. The 600 MW and 1 GW targets have no firm dates, so I'd read them as ambitions until sites are named.
Sources: Vultr launches 20-MW AI supercluster in Chennai, The New Indian Express; US cloud provider Vultr eyes 1 GW AI capacity in India, Times of India
