VAST Data and AMD have expanded their collaboration to help AI cloud providers and enterprises build high-performance AI infrastructure designed to support training, inference and agentic AI workloads at scale.
The collaboration combines the VAST AI Operating System with 6th Gen AMD EPYC processors and AMD Instinct GPUs, bringing together compute, data management and software to address the growing demands of production AI environments. The companies said the joint platform is intended to improve infrastructure efficiency as organisations deploy increasingly complex inference, retrieval-augmented generation (RAG) and agentic AI applications.
The expanded collaboration includes new AI infrastructure reference architectures developed alongside DriveNets, support for AMD’s latest EPYC processors in VAST’s next-generation platforms, and software integrations with TensorMesh and EmbeddedLLM. It also introduces KV cache optimisation capabilities, automated cache lifecycle management for enterprise compliance, and networking integration using the AMD Pensando Pollara 400 AI NIC.
According to VAST Data, early testing using an AMD Instinct MI355X GPU demonstrated a 9X improvement in time-to-first-token and 9.7X higher token throughput through KV cache offloading for high-concurrency agentic AI workloads.
“AI is entering an operational phase where infrastructure efficiency matters as much as model performance,” said John Mao, Vice President, Global Technology Alliances at VAST Data. “The industry is discovering that inference is fundamentally a data problem. Success depends on how effectively organizations can bring data, compute, memory and intelligence together as a single system.”
Derek Dicker, Corporate Vice President, Enterprise Business Group at AMD, added: “Our expanded collaboration with VAST combines AMD EPYC CPUs and Instinct GPUs with the software foundation customers need to accelerate inference, improve infrastructure efficiency and deploy AI at scale.”
According to VAST Data, the collaboration reflects growing demand for AI infrastructure that supports large-scale inference and agentic AI alongside traditional model training.
Discussion about this post