VAST Data has expanded its collaboration with AMD to help cloud providers and enterprises build infrastructure for AI training, inference and agentic AI applications.
The partnership combines the VAST AI Operating System with sixth-generation AMD EPYC processors, AMD Instinct graphics processors and AMD Pensando networking technology.
The companies are addressing the growing infrastructure requirements created by AI agents, reasoning models and large-scale inference services.
AI inference occurs when a trained model processes new information and generates a response or prediction. As organizations deploy agents with longer conversations and more persistent context, performance increasingly depends on managing data, memory and processor resources efficiently.
VAST’s Disaggregated Shared Everything architecture provides unified access to storage, databases, streaming data and AI services across distributed environments.
The expanded collaboration includes the use of sixth-generation AMD EPYC processors in VAST’s next-generation CBox and EBox platforms.
VAST, AMD and DriveNets have also developed an AI infrastructure reference architecture combining AMD Helios rack-scale systems, the VAST AI Operating System and DriveNets networking.
The architecture supports training, inference, reinforcement learning and key-value cache workloads.
A key-value cache stores information previously processed by an AI model, allowing the system to reuse context rather than recomputing it for every request.
VAST said early testing with an AMD Instinct MI355X processor produced a ninefold improvement in time to first token and 9.7 times higher token throughput when offloading the cache to VAST infrastructure during high-concurrency agentic workloads.
The results were supplied by VAST and were not independently verified by AMD. Actual performance may vary depending on hardware and workload configuration.
The collaboration also provides automated expiration and deletion of cached information that may contain personal or sensitive data.
KEY QUOTES:
“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.”
John Mao, Vice President of Global Technology Alliances at VAST Data
“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.”
Derek Dicker, Corporate Vice President of the Enterprise Business Group at AMD
