OpenSharing: An AI Model Sharing Protocol for the Agentic Era
Databricks OpenSharing is an open, cross-cloud AI model sharing protocol that standardizes how organizations exchange data assets, AI models, and agent skills with secure, governed, zero-copy access across platforms and storage environments. Evolving from the Delta Sharing standard launched in 2021, OpenSharing broadens the scope from tabular data to the "full AI stack", including unstructured data and agent skills. Now a Linux Foundation project, it offers common APIs for discovery, authorization, and access so partners can publish once and reach many platforms rather than building custom integrations or relying on single-vendor hubs. It also adds support for Iceberg IRC clients and on-premises storage partners, allowing enterprises to connect local assets directly to cloud platforms without data movement. As Matei Zaharia notes, Delta Sharing proved that customers prefer open standards over lock-in, and OpenSharing extends that preference to AI-era collaboration.
Databricks Marketplace Apps: From Data Catalog to Full AI Workbench
Databricks Marketplace is evolving from a data discovery catalog into a full data ecosystem platform through Marketplace Apps and Genie Agents. Partners can now distribute Databricks Apps that run natively on the platform, covering scenarios such as analytics dashboards, operations portals, AI chat interfaces, and data workflow tools. Databricks reports 5X growth in Apps since launch, with more than 5,000 accounts running Apps in production weekly and the top 100 customers deploying over 250 production Apps on average. Marketplace Commit Drawdown and upcoming transactability aim to let customers tap pre-committed spend for eligible partner solutions, aligning incentives for Databricks sellers and independent software vendors. According to Databricks, over 20,000 customers, including more than 70% of the Fortune 500, now rely on the platform, giving Marketplace participants a large installed base for AI and data products.
AI/BI Dashboards and Themes: Brand-Consistent Analytics at Scale
Alongside OpenSharing and Marketplace Apps, Databricks is expanding its AI/BI tooling so teams can design analytics that feel consistent across business units and products. Workspace themes and shared design patterns give organizations a way to standardize the look and behavior of dashboards, reports, and AI-infused analytics experiences. This matters as agents and AI models consume shared datasets or Marketplace Apps: end users still expect a familiar interface, regardless of which team or partner built the underlying solution. In practice, centralized design systems let data teams define common layouts, color palettes, and interaction patterns once, then apply them to many AI/BI assets. Combined with Databricks-native applications and shared AI models, these capabilities aim to reduce the overhead of maintaining many different front ends, while improving trust and adoption among non-technical business users.
A Unified Data and AI Ecosystem for Cross-Cloud Collaboration
Taken together, OpenSharing, Marketplace Apps, and integrated AI/BI tools position Databricks as a data ecosystem platform that addresses fragmentation in enterprise AI deployments. OpenSharing lowers barriers to cross-cloud collaboration by making AI model sharing and agent skill exchange open and protocol-based, not vendor-specific. Marketplace Apps and Genie Agents turn that shared foundation into complete workflows that can be discovered, installed, and monetized inside the same environment. Designable AI/BI dashboards ensure these capabilities surface in a consistent, brand-aligned way to end users. Instead of stitching together isolated tools for data sharing, model hosting, and analytics, organizations can treat Databricks as a single environment where partners publish, transact, and operate AI solutions. The result is a more coherent ecosystem that supports open standards while encouraging commercial innovation and multi-party collaboration.
