VAST Data

VAST AI Operating System

Fundamental

Data, not the GPU, increasingly sets the throughput of an AI factory - VAST is the platform built for that.

The data problem in the AI era

The most expensive hardware in a data center now spends much of its life waiting. GPUs sit idle while data crawls to them from slow, tiered storage - industry GPU utilization has been cited as low as ~5% (VentureBeat). VAST's framing is blunt: “this isn't a GPU problem, it's a data problem - your GPUs are starving.”

Why traditional storage fails here

Tiered storage and generic cloud object stores were built for cost-per-terabyte, not for feeding GB200 / GB300 NVL72 racks. Warm AI data lands on slow tiers, and the GPUs stall waiting for it to be staged.

The bottleneck moved

As clusters scale into thousands of accelerators, the limiting factor stops being FLOPs and becomes how fast the data layer can keep every GPU fed. Throughput is now a storage problem.

~5% GPU utilization figure cited by VentureBeat; actual utilization varies widely by workload and deployment.

AI factory & token factory

NVIDIA frames the modern AI data center as a factory: it takes data as raw material and turns it into intelligence, measured as token throughput. The output of that factory is set by the whole line - and the data layer, not just the GPU, governs how many tokens come off the end.

Raw data → data layer → GPUs → tokens. Starve any stage and the whole line slows. The same logic that makes NVIDIA infrastructure and inference serving throughput-bound applies upstream: the factory only runs as fast as its slowest input.

See the site's NVIDIA Infrastructure and Inference & Serving pages for how the GPU and serving stages turn fed data into tokens.

One platform, not a stack of point solutions

A typical AI data estate is a stack of separate products - a data lake, a warehouse, a vector DB, a streaming bus, and a compute cluster - each with its own copy of the data and its own ETL between them. VAST collapses all of that into one platform on a single namespace spanning files, objects, tables, and vectors, under one permission and security model.

Data lake + warehouse

Files, objects, and tables share one namespace - no copy between the lake and the warehouse, no nightly ETL.

VectorStore

Embeddings live next to the source records they describe, so RAG retrieval never queries a stale, separate index.

Streaming + compute

A Kafka-compatible broker and serverless functions run inside the platform, triggered by data events.

One security model

Files, objects, tables, and vectors are governed under a single permission and audit model - not four.

The payoff is eliminating the copies and ETL that move data between systems - fewer pipelines to break, one source of truth, and no consistency lag between what the warehouse, the vector index, and the agents see.

The VAST AI Operating System

Unveiled in May 2025, the AI OS is a layer cake: Storage (DASE) at the base, the Database / VectorStore on top, the DataEngine adding serverless compute and an Event Broker, and the AgentEngine running agents and MCP at the top - with InsightEngine, the NVIDIA real-time RAG stack, spanning the layers. Click a layer to see what it does and why it matters for AI.

InsightEngine · NVIDIA

Real-time RAG stack spanning every layer - NVIDIA accelerated retrieval over the full platform.

AgentEngine

Agents + MCP

What it is: A runtime for AI agents and tools that live next to the data, exposed through the Model Context Protocol (MCP) under one permission model.
Problem it solves: Agents normally run far from the data they reason over, copying context across systems and losing governance at every hop.
Why it matters for AI: Running agents in the data plane means tool calls hit fresh, governed data with no extract step - context stays current and access stays controlled.

What is shipping today

The engines below are at different stages. As you read the deep dives, keep maturity in mind - most of the platform is generally available, the agent layer is newer, and parts of the broader AI OS are still on the roadmap.

Generally available

Storage (DASE), DataBase, VectorStore, DataSpace, DataEngine & Event Broker, and InsightEngine are shipping today.

In preview

AgentEngine & MCP is the newest layer - announced and demoed, rolling out, not yet fully GA.

On the roadmap

The wider AI OS vision - see the Roadmap page for what is announced versus already shipping.

Maturity reflects VAST's public announcements and may change; confirm current availability with VAST before positioning.

Removing the storage-vs-GPU tradeoff

The old choice was fast storage you couldn't afford at scale or cheap storage that starved the GPUs. VAST's architecture aims to remove that tradeoff: feed GPUs directly while keeping all warm AI data on flash.

GPUDirect Storage

Data moves straight from storage into GPU memory, bypassing the host CPU and bounce buffers on the read path.

NVMe-oF fabric

Every compute node addresses the full all-flash pool over the network, so there is no local-vs-remote data tier.

All-flash at HDD-like cost

VAST-reported economics put all-flash capacity near hard-disk price points, so warm AI data stays on fast media.

All-flash-at-HDD-like-cost economics are VAST-reported; realized cost depends on data reduction rates and configuration.

Where to go deeper

Each pillar of the platform has its own deep dive - the architecture underneath, the data and vector layers, the compute and agent engines, the NVIDIA stack, and the roadmap.