Vast Data Unveils Multicloud AI Platform

The platform is built on a "Disaggregated, Shared-Everything" (DASE) architecture, a departure from traditional designs that separates compute from storage. This allows organizations to scale performance and capacity independently, a key requirement for resource-intensive AI workloads. The system state is stored on NVMe SSDs in highly available enclosures, enabling any server to access all data and metadata directly without internode communication. At its core, the VAST Data Platform unifies several components into a single operating system for AI. This includes the DataStore for unstructured data, DataBase which combines features of a database, warehouse, and lake, a global DataSpace for access across environments, and a DataEngine for execution. The goal is to simplify the complex software stack typically required for AI and analytics. The company, founded in 2016 by Renen Hallak, Jeff Denworth, Alon Horev, and Shachar Fienblit, has seen rapid growth, with a recent valuation soaring towards $25-30 billion. It has secured major deals, including a $1.17 billion agreement with AI cloud provider CoreWeave, and counts customers like Elon Musk's xAI and Lambda Labs. This has pushed the company's contracted annual recurring revenue past $500 million. This move positions VAST directly against established data giants like Snowflake and Databricks, as well as cloud providers, by offering a comprehensive, integrated platform rather than siloed services. The deep partnership with Nvidia is a key strategic advantage, with the platform's software running directly on Nvidia-powered servers and integrating a suite of their libraries and APIs. This collaboration aims to streamline AI pipelines by merging storage, database, analytics, and inference into one stack.