Databricks Hands Delta Sharing to the Linux Foundation and Levels It Up - Techstrong.ai

Skip to content

Databricks Hands Delta Sharing to the Linux Foundation and Levels It Up

4.3 min readPublished On: June 11, 2026By Steven Vaughan-Nichols

OpenSharing is the open, vendor‑neutral plumbing for exchanging not just data, but AI models and “agent skills” across platforms and clouds.

People have figured out that one large language model (LLM) is not enough, and forget about relying on a single data lake or one set of agent skills. To really get work done with AI we need to share models, data, and skills and to do that well you need standards. That’s where Databricks and the Linux Foundation come in with OpenSharing.

OpenSharing is an open standard and protocol aimed at unifying how organizations share both data and AI assets, from structured tables to large language model–based “skills,” across different platforms and environments. The project evolves Databricks’ existing Delta Sharing technology into a community‑governed Linux Foundation initiative designed for what both organizations describe as the “agentic era” of AI.

Delta Sharing, for those who don’t know it, defines a standard way for a “data provider” to expose tables or files so that “data recipients” can query that data in real time using their own tools. You don’t have to use Databricks or even run on the same cloud to access data. It’s positioned as the first open, vendor‑neutral protocol for secure data sharing at scale, as opposed to proprietary sharing features locked to a specific data warehouse or cloud.

Databricks bills OpenSharing as the “next chapter” of Delta Sharing. The company claims that even before this standardization move Delta Sharing was already the most widely adopted open data‑sharing protocol, used by enterprises such as Amadeus, Atlassian, LSEG, SAP, Stripe, and The Trade Desk for secure, zero‑copy data collaboration. All these companies have also bought into OpenSharing.

By moving the technology under the Linux Foundation umbrella, the companies are reassuring customers and partners that the protocol will remain vendor‑neutral and that its roadmap will be shaped by a broader open source community rather than a single commercial steward calling the shots. Since that’s essential for a standard to get broad adoption this was a smart move.

In a statement, Linux Foundation CEO Jim Zemlin said, “OpenSharing addresses a critical need for a common, vendor-neutral framework that enables organizations to exchange AI assets securely and interoperably across platforms and ecosystems.”

What’s new, however, is not just governance but scope as well. OpenSharing extends Delta Sharing beyond tables and files to cover AI‑specific artifacts, such as agent skills, AI models, and unstructured data. The idea is that instead of wiring up bespoke integrations or being locked into proprietary marketplaces, enterprises can publish and discover AI assets via standard APIs for discovery, authorization, and access, regardless of which data platform or cloud stack they run.

Databricks is pitching this as the first open protocol that can securely share AI assets across organizations at scale. Before OpenSharing, the company argues, there was “no standard way” to share agent skills or models across corporate boundaries. That forced customers into expensive point‑to‑point integrations or single‑vendor ecosystems. In the OpenSharing model, those assets become addressable resources with consistent semantics and governance hooks, opening the door to richer cross‑company and intra-company collaboration.

Technically, OpenSharing tries to smooth over a fragmented sharing landscape by abstracting away underlying storage and table formats. The protocol can publish data that lives on different platforms and clouds and is explicitly designed to support multiple open table formats, including existing Delta Sharing recipients and Apache Iceberg clients. Databricks highlighted new support for “Iceberg IRC clients.

In this context, by the way, “IRC” refers not to Internet Relay Chat, but to the Iceberg REST Catalog. This is a standardized REST API for managing and querying Iceberg table metadata in a vendor‑agnostic way. The spec allows any compatible catalog (Snowflake, Databricks Unity Catalog, Microsoft Fabric OneLake, AWS Glue, Tabular, etc.) to expose Iceberg tables via a common REST interface. Any compliant client can connect without custom drivers. It’s clearly a very useful extension for OpenSharing.

By building on Delta Sharing’s open connectors and extending to additional engines and formats, OpenSharing aims to enable organizations to share governed, live data and AI assets without replication, even when partners sit on different stacks or in different regions. Storage partners...