Vultr plans 20MW AMD Instinct AI supercluster in Chennai
The MI355X-powered deployment will target AI labs, startups, enterprises and research institutions seeking in-country compute for model training, fine-tuning and inference.
Vultr plans to deploy AMD Instinct MI355X GPUs in a 20MW AI supercluster in Chennai, expanding access to large-scale AI computing infrastructure in India as demand grows for locally hosted workloads that meet data residency, sovereignty and latency requirements.
The deployment will be India’s largest AMD Instinct deployment, according to Vultr.
It will also give the privately held cloud infrastructure provider two of the world’s largest AMD AI infrastructure deployments globally, alongside its flagship supercluster in Springfield, Ohio.
The Chennai facility is designed to serve AI labs and foundation model developers, AI-native startups, enterprises, and research and academic institutions. It will support large language model and foundation model training, fine-tuning and high-throughput inference.
The deployment creates a potential infrastructure base for enterprise AI projects, including workload deployment, model optimisation, cloud migration and managed services for Indian channel partners.
However, Vultr has not announced any Indian channel partners, partner programme or commercial arrangements for taking the new capacity to market.
MI355X GPUs to power AI workloads
The Chennai supercluster will use AMD Instinct MI355X GPUs, each equipped with 288GB of HBM3E memory and 8TB/s of memory bandwidth.
The specifications are intended to support the demanding compute and memory requirements of increasingly complex AI models.
J.J. Kardwell, CEO of Vultr, says organisations deploying AI at scale need local infrastructure to address data sovereignty, data residency and latency requirements.
The Chennai deployment will allow organisations in India to train and fine-tune models and run inference on infrastructure located within the country, Kardwell says.
The announcement does not disclose the number of GPUs planned for Chennai, the investment involved, the expected operational date, or whether the stated 20 MW refers to IT load or total facility power.
The announcement also does not identify customers that have committed to using the facility.
These details will help determine the scale of capacity available to customers and how quickly enterprises and service providers can build workloads on the infrastructure.
Vultr extends AMD collaboration
The Chennai deployment expands Vultr’s collaboration with AMD across its AI infrastructure and cloud computing portfolio.
The companies’ relationship spans AMD Instinct MI300X, MI325X and MI355X GPUs, as well as planned adoption of MI455X GPUs and AMD Helios rack-scale AI architecture.
Vultr says it plans additional capacity expansion in 2027, including MI455X GPUs and the Helios architecture. The announcement does not specify where that capacity will be deployed or confirm that these technologies will be installed in Chennai.
The new supercluster will also use AMD ROCm, the company’s open software ecosystem for GPU computing. ROCm provides access to AI frameworks, libraries and models, giving developers greater flexibility when building and deploying AI workloads.
Jaya Jagadish, country head and SVP Engineering at AMD India, says access to compute at this scale can help Indian researchers, startups and enterprises turn AI ideas into applications. She also highlights ROCm’s role in offering developers more choice in how they develop and deploy AI.
The combination of GPU infrastructure and an open software ecosystem could be relevant to AI implementation and workload optimisation projects for systems integrators and managed service providers. The companies have not, however, announced specific software integrations, implementation partners or a channel enablement initiative linked to the Chennai facility.
Vultr says its global footprint comprises 34 cloud data centre regions serving enterprises, AI innovators and government agencies in 17 countries across six continents.
The company describes its infrastructure as secure, compliant and available at predictable cost.
The Chennai supercluster adds to the shift towards geographically distributed AI infrastructure, bringing compute closer to where data is generated and applications are deployed.
The extent of the opportunity for Indian channel partners will depend on the capacity available, customer demand and the routes Vultr establishes for partners to deliver services around the infrastructure.