Databricks’ Ecosystem Overhaul: From Platform to Operating Fabric
Databricks’ latest announcements describe a unified data platform ecosystem that combines real-time analytics, security tooling, and open AI collaboration protocols to function as an end-to-end data and AI operating fabric for enterprises. Instead of stitching together point solutions for streaming, governance, security, and model distribution, Databricks wants data teams to work from a single lakehouse architecture that can serve humans, applications, and agents at the same time. At its Data + AI Summit, the company introduced Lakehouse//RT for Databricks real-time analytics, agreed to acquire AI SOC provider Panther to reinforce enterprise data security, and elevated Delta Sharing into OpenSharing, a wider AI model sharing protocol that also covers agent skills and unstructured data. New Marketplace and Apps extensions round out the vision by drawing independent software vendors and data providers further into the lakehouse, reducing integration friction across the stack.
Lakehouse//RT: Real-Time Analytics Without a Separate Serving Layer
Lakehouse//RT puts Databricks real-time analytics directly on governed Delta Lake and Apache Iceberg tables, instead of pushing teams toward a duplicate serving database. Powered by the new Reyden compute engine, it is designed to deliver millisecond query latency for tens of thousands of concurrent users and AI agents querying the same lakehouse architecture. Early customers have seen up to 16x better performance than existing real-time serving stacks, with response times as low as 10ms on smaller datasets and sub-100ms performance on larger ones. Every Lakehouse//RT query runs inside Unity Catalog’s governance model, so there is no extra permissions tier, no separate change data capture pipelines, and no proprietary formats sitting off to the side. For organizations building agentic applications that need fresh operational context, the message is clear: the lake itself can now serve as the primary real-time analytics engine.
Panther and the Security Lakehouse: AI-Native Threat Detection
Databricks’ agreement to acquire Panther brings AI-native security operations into the same lakehouse architecture that powers analytics and AI workloads. Panther is an AI SOC platform built for the emerging security lakehouse model, intended to replace legacy SIEM tools that struggle with high costs, limited data access, and manual workflows. The product comes with more than 100 out-of-the-box data integrations, detection-as-code capabilities, and agentic SOC workflows that can automate large parts of threat investigation. According to Databricks, the goal is to help organizations “detect more threats, investigate every alert, and fight AI-driven attacks with AI” by unifying security data on a common platform. In practice, Panther’s approach lets enterprises centralize logs, telemetry, and AI system events, then apply agents and analytics at scale, reinforcing enterprise data security without standing up a separate, siloed security stack.
OpenSharing: An AI Model and Skills Sharing Protocol for the Agentic Era
OpenSharing extends the open source Delta Sharing protocol into a broader AI model sharing protocol that also covers agent skills and unstructured data. Now governed as a Linux Foundation project, it is described as the first open protocol that can publish and consume AI models, agent skills, and data assets through common discovery, authorization, and access APIs, regardless of platform. This moves Databricks from data-only exchange to a richer AI asset network that spans models and skills as well. OpenSharing widens the cross-platform collaboration field by adding support for Iceberg IRC clients, so asset providers can reach new recipients without custom integration work. With new on-premises storage partners, customers can connect on-premises data directly to cloud platforms with no data movement. The project is available on GitHub, signaling Databricks’ intent to keep this layer open rather than tying AI collaboration to a single vendor marketplace.
Marketplace, Apps, and the Push Beyond Point Solutions
Alongside the platform and protocol work, Databricks is expanding its Marketplace and Apps model to make the lakehouse architecture a commercial hub for data, AI, and partner solutions. The company now supports transactability on Databricks Marketplace, letting customers draw down pre-committed spend on eligible partner offerings that run on or share data into Databricks. Partners can also list Databricks Apps and Genie Agents, opening the door to new usage-based models such as “pay-per-question” for agent-powered experiences. Databricks reports more than 20,000 customers and an annual run rate growing more than 65% year over year past $5B, giving partners a large installed base to sell into. For buyers, the promise is fewer point tools: real-time analytics, enterprise data security, AI asset exchange, and a growing partner network come together into one data platform ecosystem that can replace fragmented stacks.
