Chennai: Cloud computing provider Vultr has launched its 34th global cloud data centre region in Chennai, and is building an AI supercluster powered by AMD Instinct MI355X GPUs, starting at 20 megawatts. The company says it is the largest deployment of AMD Instinct GPUs in India to date.
According to the company, each GPU has 288 GB of HBM3E memory and 8 terabytes per second of memory bandwidth. The cluster is being built to handle AI inference and research workloads as well as the training of large language models. Vultr said it plans to expand in 2027 with MI455X GPUs and AMD Helios.
Demand for AI computing capacity is rising fast in India. The investment is seen as significant at a time when startups, research institutions and large companies want data and compute facilities locally.
Experts said a public cloud region in India could help with data sovereignty, lower latency and cost. But they also flagged challenges around power, cooling and water use.
The company said the region will give local AI startups access to global-grade compute within India. Capacity will be scaled up in phases.
The launch fits a wider push to build sovereign AI capacity in India, where governments and firms want compute and data to stay within the country. Several cloud and chip companies have announced India investments this year. Analysts caution that power supply and skilled manpower will decide how quickly such clusters can be scaled, and that pricing and utilisation will determine whether they benefit smaller startups as much as large enterprises.