Databricks Opens Up On-Prem Storage

Databricks is breaking down the barriers that have kept vast amounts of enterprise data locked away on-premises. The company today announced the Databricks storage ecosystem, a new approach designed to bring its Data Intelligence Platform directly to data residing in private clouds, on-premises data centers, and edge environments.

The core of this new offering is the OpenSharing protocol, an open-source standard that enables Databricks to connect securely to these disparate data sources without requiring any data movement. This addresses a critical pain point for organizations constrained by data sovereignty regulations, massive data gravity, or the latency demands of edge computing.

The 'Data That Can't Move' Problem

For years, the prevailing strategy was to migrate all data to the cloud. However, this approach is increasingly unfeasible for many enterprises.

Strict data sovereignty and regulatory mandates, like GDPR and HIPAA, often prohibit moving sensitive datasets. The sheer economics of storing and transferring petabytes or exabytes of data also make cloud migration a non-starter for some, leading companies to even repatriate workloads. Furthermore, edge applications in retail, manufacturing, and telecommunications demand low-latency access to local data that cloud round-trips can’t provide.

These aren't niche concerns; they represent a fundamental shift from "migrate everything" to "govern everything."

Enter the Databricks Storage Ecosystem

The new Databricks storage ecosystem is built to serve this evolving landscape. By leveraging the OpenSharing protocol, storage partners can expose their data estates directly to Databricks' serverless compute, generative AI capabilities, and Large Language Models (LLMs).

This means organizations can now run Databricks workloads, including training AI models on classified data or analyzing network telemetry, directly on their existing on-premises infrastructure. The benefit is immediate: isolated data becomes active, AI-ready assets without the cost, complexity, and compliance risks associated with data migration or duplication.

This integration provides a unified catalog across hybrid environments, allowing customers to use Databricks Serverless Compute, Genie, and AgentBricks to query and reason over data that never leaves its original location. This is not a future vision; these integrations are available today.

Key Launch Partners

Databricks has launched with support from several major storage providers:

• MinIO: Offering General Availability, MinIO AIStor natively implements OpenSharing, connecting Databricks to on-premises Apache Iceberg™️ and Delta tables under Unity Catalog governance.
• Everpure (formerly Pure Storage): In Private Preview, Everpure provides an OpenSharing connector for secure, gated access to object storage data within Databricks workspaces without replication.
• Qumulo: Slated for Private Preview in July 2026, Qumulo's NeuralSearch integrates with OpenSharing to allow secure, non-replicating sharing of data across core, cloud, and edge environments.
• VAST Data: Expected in Private Preview in August 2026, VAST Data's AI Operating System will support OpenSharing, bridging Databricks workflows with hybrid infrastructure data without massive movement.

This initiative marks a significant step in Databricks' strategy to unify data governance across the entire enterprise data estate, regardless of where the data physically resides.