VAST Data, the AI Operating System company, today 

announced a new inference architecture that enables the NVIDIA Inference Context Memory 

Storage Platform – deployments for the era of long-lived, agentic AI. The platform is a new class 

of AI-native storage infrastructure for gigascale inference. Built on NVIDIA BlueField-4 DPUs 

and Spectrum-X Ethernet networking, it accelerates AI-native key-value (KV) cache access, 

enables high-speed inference context sharing across nodes, and delivers a major leap in power 

efficiency. 

As inference evolves from single prompts into persistent, multi-turn reasoning across agents, 

the notion that context stays local breaks down. Performance is increasingly governed by how 

efficiently inference history (KV cache) can be stored, restored, reused, extended, and shared 

under sustained load – not simply by how fast GPUs can compute. 

VAST is rebuilding the inference data path by running VAST AI Operating System (AI OS) 

software natively on NVIDIA BlueField-4 DPUs, embedding critical data services directly into the 

GPU server where inference executes, as well as in a dedicated data node architecture. This 

design removes classic client-server contention and eliminates unnecessary copies and hops 

that inflate time-to-first-token (TTFT) as concurrency rises. Combined with VAST’s parallel 

Disaggregated Shared-Everything (DASE) architecture, each host can access a shared, globally 

coherent context namespace without the coordination tax that causes bottlenecks at scale, 

enabling a streamlined path from GPU memory to persistent NVMe storage over RDMA fabrics. 

“Inference is becoming a memory system, not a compute job. The winners won’t be the clusters 

with the most raw compute – they’ll be the ones that can move, share, and govern context at 

line rate,” said John Mao, Vice President, Global Technology Alliances at VAST Data 

“Continuity is the new performance frontier. If context isn’t available on demand, GPUs idle and 

economics collapse. With the VAST AI Operating System on NVIDIA BlueField-4, we’re turning 

context into shared infrastructure – fast by default, policy-driven when needed, and built to stay 

predictable as agentic AI scales.” 

Beyond raw performance, VAST gives AI-native organizations and enterprises deploying 

NVIDIA AI factories a path to production-grade inference coordination with high levels of 

efficiency and security. As inference moves from experimentation into regulated and revenue-

driving services, teams need the ability to manage context with policy, isolation, auditability, 

lifecycle controls, and optional protection – all while keeping KV cache fast and usable as a 

shared system resource. VAST delivers those AI-native data services as part of the AI OS, 

helping customers avoid rebuild storms, reduce idle-GPU resource waste, and improve 

infrastructure efficiency as context sizes and session concurrency explode. 

“Context is the fuel of thinking. Just like humans that write things down to remember them, AI 

agents need to save their work so they can reuse what they’ve learned,” said Kevin Deierling, 

Senior Vice President of Networking, NVIDIA. “Multi-turn and multi-user inferencing 

fundamentally transforms how context memory is managed at scale. VAST Data AI OS with 

NVIDIA BlueField-4 enables the NVIDIA Inference Context Memory Storage Platform and a 

coherent data plane designed for sustained throughput and predictable performance as agentic 

workloads scale.” 

Experience VAST’s industry-leading approach to AI and data infrastructure at VAST Forward, 

our inaugural user conference, February 24–26, 2026 in Salt Lake City, Utah. Engage with 

VAST leadership, customers, and partners through deep technical sessions, hands-on labs, and 

certification programs. Register here to join.