Vultr and AMD plan 20 MW AI supercluster in Chennai with MI355X chips

Vultr, a US-based cloud computing company, announced on 22 October 2026 that it will build a 20-megawatt artificial intelligence (AI) supercluster in Chennai using AMD's Instinct MI355X graphics processing units (GPUs). The company said the facility will be India's largest deployment of AMD Instinct chips and, together with its site in Springfield, Ohio, one of its two largest AMD AI installations. It is aimed at AI labs, companies building foundation models, AI start-ups, enterprises and research institutions in India that need computing power to train and run AI models. Vultr has not disclosed the number of GPUs, the investment or when the cluster will go live.

Computing power is the bottleneck for AI. Training a large language model or running AI services for millions of users needs thousands of specialised chips, and most of that capacity is concentrated in the United States. Indian AI companies have often had to rent computing abroad, which raises costs and creates concerns about where data is stored. The government's IndiaAI Mission has been subsidising access to GPUs to close this gap, and a large commercial cluster in Chennai adds to the capacity available inside the country.

The choice of AMD is also significant. Nvidia dominates the market for AI chips, and its GPUs are often in short supply. AMD's Instinct line is the main alternative, and large deployments like this one help build a second source of computing for developers. The trade-off is software: many AI tools are written for Nvidia's CUDA platform, and moving them to AMD's ROCm can take extra work, although support has been improving.

Chennai has become one of India's main data-centre hubs because of its undersea cable landing stations, which connect India to the rest of the world, and its power supply. A 20 MW AI site is large by Indian standards for GPU computing, though small compared with the largest AI campuses being built in the United States.

The announcement comes amid a rush of AI infrastructure investment in India, from global cloud companies and Indian conglomerates alike. For Indian start-ups and researchers, more local capacity could mean lower prices and shorter waits for GPUs, provided supply keeps pace with demand.

Short-term: Indian AI developers gain another option for renting high-end GPUs within the country once the site is running. No date has been given for that.

Long-term: If built and expanded as planned, the cluster would strengthen Chennai's position as an AI computing hub and give Indian companies a larger alternative to Nvidia-based capacity, which could help lower costs.

Who is affected: AI start-ups and research labs in India, enterprises adopting AI, data-centre and power suppliers in Tamil Nadu, and competing cloud providers.

Questions and Answers