Vultr 将在金奈部署 20MW AMD MI355X GPU 集群
Chennai to Host One of the World’s Largest AMD GPU Clusters
云基础设施公司 Vultr 宣布将在印度金奈部署 20MW 的 AMD MI355X GPU 集群，这将是印度最大的 AMD Instinct GPU 部署，也是全球最大规模之一。MI355X 于去年 6 月发布，配备 288 GB HBM3E 内存，高于 Nvidia B200 的 180 GB。Vultr 在全球 17 个国家运营 34 个云数据中心区域，并计划明年进一步扩容，引入下一代 MI455X GPU 和 Helios 机架级 AI 架构。
正文
当前显示原文，中文全文译文尚未就绪。
Vultr, the cloud infrastructure company, has announced it will deploy a 20MW cluster of AMD MI355X GPUs in Chennai.
This will be India’s largest AMD Instinct GPU deployment, and one of the world’s largest, alongside Vultr’s supercluster in Springfield, Ohio, in the US.
“Enterprises deploying AI at scale need infrastructure they can access locally to meet data sovereignty, data residency, and latency requirements,” said J.J. Kardwell, CEO of Vultr.
“Vultr’s Chennai supercluster, powered by AMD MI355X GPUs, gives organisations in India the capacity to train, fine-tune, and run inference on in-country infrastructure.”
Jaya Jagadish, country head & SVP of engineering at AMD India, said, “India has the talent and ideas to build the next generation of AI. Access to compute at this scale will help turn more of those ideas into reality.”
“Through our collaboration with Vultr, we’re expanding access to AMD Instinct compute in India, backed by the AMD ROCm open software ecosystem, giving researchers, startups and enterprises greater choice and flexibility in how they develop and deploy AI,” she added.
AMD launched its Instinct MI355X GPUs last June.
The accelerator offers 288 GB of HBM3E memory, compared with 180 GB on Nvidia’s B200, alongside competitive peak compute performance.
However, hardware specifications alone do not determine real-world performance. NVIDIA retains an advantage through its mature CUDA software ecosystem, extensive developer adoption and integrated AI infrastructure stack.
It also continues to dominate the AI accelerator market, although its share varies by market definition and measurement period.
Vultr operates 34 cloud data centre regions across 17 countries worldwide. The company said that it plans significant additional capacity expansion with AMD next year, including the next-generation MI455X GPUs and the Helios rack-scale AI architecture.
AMD is also pursuing larger deployments through cloud providers and AI labs. Alongside Vultr, TensorWave operates an 8,192 AMD GPU cluster, one of North America’s largest deployments. The company has since expanded its cloud offerings to include the newer MI355X GPUs and announced plans to adopt AMD's next-generation GPUs.
Oracle announced plans to begin deploying 50,000 MI450-series GPUs in its cloud infrastructure from the third quarter of 2026, with further expansion planned for 2027 and beyond.
OpenAI expects to bring AMD’s Helios rack-scale systems online from the fourth quarter of 2026, with deployments accelerating through 2027. Anthropic, meanwhile, has announced a multiyear partnership to deploy up to 2 GW of MI455X GPUs.
Meta has also committed to a multi-gigawatt AMD GPU deployment, working to scale capacity and validate AMD’s next-generation infrastructure.
Chennai is one of the country’s major data centre hubs. Vultr's announcement follows CapitaLand’s investment.
The Singapore-based firm announced a ₹1,940 crore investment in its Chennai data centre, with 34 MW of IT capacity and 53 MW of gross power capacity. Other major players expanding in the city include STT GDC India, CtrlS Datacenters and Sify Technologies, which have announced or developed facilities to meet growing demand for cloud computing and AI infrastructure.