VAST's race to own AI data

This marks a land grab for the AI operating layer, not just another storage upgrade. VAST is arguing that AI has hardware and an early killer app, but still lacks the system software that lets companies store raw data, index it, query it, and run processing jobs in one place. That is why it is moving from selling flash arrays into selling the control plane for AI workloads, with database, catalog, and compute services bundled on top of storage.

• The product shift is concrete. A team can keep training files, vector data, metadata, SQL queries, and Spark or Python jobs on the same system instead of moving data between a storage array, warehouse, ETL tool, and compute cluster. That lets VAST sell a much bigger bundle per account than a traditional storage vendor.
  2 sacra
• The comparison set changes when VAST goes up stack. It still competes with WEKA, Pure, Dell, and DDN on AI storage, but once it adds catalog, database, and execution layers it starts competing with Databricks and Snowflake for control of how AI data is organized and used, especially in on premises and hybrid environments.
  2 sacra 3 sacra 4 sacra
• The economics already reflect that broader ambition. VAST reached about $200 million ARR by January 2025, sells deals that often exceed $1 million annually, and describes customer expansion coming from more data, more workloads, new accounts, and added infrastructure services in parallel. That is the pattern of a platform vendor trying to become core infrastructure.
  1 sacra 2 sacra

The next phase is a race to become the default data substrate under AI factories and enterprise agent systems. If VAST keeps turning storage deployments into broader software adoption, it can move from a niche infrastructure supplier into one of the few companies that own where AI data lives, how it is queried, and which workloads run closest to it.