Platforms Abstracting Storage Vendors
VAST Data
The real threat is not that Databricks or Snowflake build better storage hardware, it is that they make storage choice invisible inside a larger data workflow. Once a team can land files, govern access, run SQL, attach AI processing, and share results inside one control plane, the storage layer starts to look like a commodity input. That is why VAST is pushing beyond arrays into catalog, database, and compute, especially for hybrid and on premises AI deployments.
-
Databricks is already moving in this direction. Unity Catalog now governs managed volumes for unstructured data, lets teams store files and apply access controls inside Databricks, and exposes those assets across engines. In practice, that means image, audio, video, and PDF workflows can stay inside the Databricks environment instead of forcing a separate storage layer to own metadata and access.
-
Snowflake has made a similar move from the warehouse side. Its unstructured data features let customers store, secure, and process files such as recordings, PDFs, and medical images inside Snowflake with Snowpark, which pushes the platform closer to being the place where both data and compute live together. That makes standalone storage vendors easier to route around for cloud first customers.
-
VAST still has a concrete opening where cloud platforms are weakest. Large enterprises often run mixed environments, with sensitive data on premises, model training near GPU clusters, and strict latency or sovereignty constraints. VAST is building a unified system that serves files, objects, SQL queries, and preprocessing on the same infrastructure, so the customer does not have to split those workloads across a cloud warehouse and a separate storage estate.
The market is heading toward fewer standalone infrastructure layers and more bundled data operating systems. Databricks and Snowflake will keep pulling storage functions upward into analytics and AI, while VAST will keep pushing downward from storage into database and compute. The winning wedge for VAST is becoming the default data plane for hybrid AI, where the cloud platforms still cannot fully absorb the workload.