Neysa 在孟買部署 8,192-GPU AMD Instinct MI350X 集羣
Neysa Deploys AMD GPU Cluster in Mumbai for AI Training and Inference
Neysa 透過 Tytan 裸機 AI 基礎設施平台,在孟買部署並啟用由 8,192-GPU 組成的 AMD Instinct MI350X 集羣,支援大型語言模型訓練、微調、生成式 AI 和高吞吐量推理。集羣結合 AMD EPYC 處理器及 AMD ROCm 軟件生態,客戶可選擇裸機、裸機搭配 Slurm 或裸機搭配 Kubernetes。
Neysa has deployed an AMD Instinct MI350X GPU cluster in Mumbai through its Tytan bare-metal AI infrastructure platform.
The 8,192-GPU cluster is now operational and will support large language model training, fine-tuning, generative AI and high-throughput inference workloads.
The deployment brings together AMD Instinct MI350X GPUs, AMD EPYC processors and the AMD ROCm software ecosystem.
Neysa said the cluster was brought online in a single phase, with global customers already having contracted capacity for AI deployments in India.
The infrastructure is designed to give customers dedicated access to accelerators, along with control over network architecture, storage, deployment models, and service-level requirements.
Customers can choose between bare-metal deployments, bare metal with Slurm, or bare metal with Kubernetes.
The deployment also expands Neysa's silicon-agnostic infrastructure strategy.
Through Tytan, the company allows customers to select accelerator architectures and operating configurations based on their workloads, while Neysa manages cluster deployment and operations.
Neysa said the MI350X cluster is intended to serve both Indian enterprises and global AI companies establishing compute capacity in the country.
The company also operates Velocis, a managed AI cloud platform that combines GPU compute, orchestration, inference, observability and cost management.
"The question for large AI deployments is no longer simply how many GPUs are available. It is whether the infrastructure can be deployed at scale, operated with certainty and configured for the workload," said Sharad Sanghi, Co-Founder and CEO of Neysa. “Neysa Tytan gives AI offtakers control: dedicated bare-metal capacity and the freedom to choose the silicon, architecture and operating model that fit their requirements. With AMD Instinct MI350X on Tytan, we are expanding that choice while delivering the scale global customers need to deploy AI infrastructure in India.”
In October, Neysa partnered with ST Telemedia Global Data Centres (STT GDC) to deploy GPU infrastructure supporting up to 9,000 GPUs at the STT Noida 2 facility. The first phase is expected to go live in Q1 2027 and will support AI training and inference through Neysa's Velocis platform.
The multi-year arrangement is designed to serve Indian enterprises and global compute customers, with data residency and predictable latency among its stated objectives.
In February, Blackstone and co-investors entered into definitive agreements to provide up to $600 million in equity financing to Neysa, with the company planning to raise another $600 million in debt. The funding was intended to support its expansion plans, including deploying more than 20,000 GPUs across India.
Neysa's announcement also marks another major step for AMD in scaling its GPU infrastructure in India.
In October, Vultr announced plans to deploy 10,000 AMD Instinct MI355X GPUs in a 20 MW AI supercluster in Chennai. The facility is expected to become India's largest AMD Instinct supercluster and the second-largest globally. Vultr has also outlined plans to expand the infrastructure in 2027 with AMD's next-generation MI455X GPUs and Helios rack-scale architecture.
來源:Analytics India Magazine · analyticsindiamag.com