30, June 2026
by: Grace McCarthy
This is the first of three posts on what was announced at the Data + AI Summit 2026 and what it actually means, written by Ramen Bhar, Principal Data Engineer at Arreoblue.
Every June, Databricks holds a summit, and every June, it announces a very large number of new products. This year, there were twenty-three, by my count.
This is the reference edition: what got announced, what stage it’s actually at, and what it’s for. Prefer the highlights? Part 2 has the five that matter most. Want the strategic read? That’s Part 3.
One flag before we start: “private preview” and “generally available” get shown on very similar slides, but mean very different things. I’ve noted the status of each item so you know which is which.
Genie One (released) is the headline product. It’s stopped being something you ask questions and become something that does the work: drafting documents, building reports, and taking actions through connected tools on web, mobile, Slack, and Teams, on a schedule, with alerts. You pay for what you use, with a small free monthly allowance.
Genie Ontology (launched at the summit) is the quieter but arguably more important half of that story: a constantly updated map of what your business data means, built by reading tables, documents, tickets, and fifty-plus connected apps. Genie is only ever as good as this layer underneath it. If your context also lives in Snowflake or Salesforce, you’ll likely still need something alongside it.
Genie Agents (released) turns a good one-off Genie conversation into a reusable, governed agent the whole team can run again.
Genie Code (released, expanded) is a workspace for data and ML engineers, now able to import Tableau and Power BI reports directly into Databricks dashboards.
Genie ZeroOps (private preview) watches your live pipelines and models, and when something breaks, traces the cause, drafts a fix, tests it on a copy, and waits for a human to approve it.
Genie App Builder (private preview) turns a plain-English description into a working internal app connected to your governed data. It sits alongside Serverless Micro Apps, which switch off when nobody’s using them, so an idle app costs nothing.
Unity AI Gateway (announced, rolling out) is the control centre for AI in your business: models, agents, tools, wherever they run. It tracks and caps spend, routes work to cheaper models where sensible, and can require human approval before an agent does anything sensitive.
Unity Catalog Metrics (available) lets you define a number, revenue, churn, whatever it may be, once, and use that same definition everywhere.
Business Glossary (preview soon) and Domains (public preview) give you a shared vocabulary and a way of organising data by business area.
Catalog Federation (announced, demoed live with Mastercard) lets one set of governance rules cover your data across different clouds and regions, without copying it.
Automatic Identity Management (Entra ID released on AWS and GCP; Okta in public preview) sets up users, groups, and permissions automatically from your existing identity system.
Lakebase (released, expanded) is a ready-to-use, Postgres-compatible database that sits alongside your analytics data, with one-click failover, instant test copies, automatic query tuning, and built-in search.
LTAP (coming soon) is the more ambitious idea underneath Lakebase: one shared copy of data for both apps and analytics, instead of two copies kept in sync by a pipeline that can, and eventually will, break. It’s the earliest-stage item on this list,, so I won’t plan around it yet.
Lakehouse//RT, powered by a new engine called Reyden (beta), runs real-time queries in under 100 milliseconds directly on governed data. Databricks claims up to 16x faster than the dedicated real-time databases people usually run alongside a lakehouse; test that figure on your own workloads before taking it as given.
Lakeflow updates: Designer, a drag-and-drop and plain-English pipeline builder, is released, and produces real, editable code underneath. Connect now has over 100 ready-made connectors, and Jobs can trigger when the data is actually ready, not just at a fixed time.
AI Runtime & real-time ML (available, expanded) gives you GPUs with no setup for training and fine-tuning, plus low-latency serving for predictions like fraud detection and personalisation.
CustomerLake (private preview) is a customer data platform built inside the lakehouse instead of bolted on as a separate marketing tool. Early customers include HP, Circle K, AB InBev, and Getnet by Santander. It’s the least mature product in this section, entering a market with some well-established players.
Lakewatch (announced) is a security monitoring system built into the lakehouse, backed by Databricks’ agreed purchase of security company Panther, its third security acquisition. It also monitors what AI agents themselves are doing, flagging things like personal data leaks.
Databricks Apps (available) bundles App Spaces, Genie App Builder, and Serverless Micro Apps, plus an expanded Marketplace where twenty partners launched finished apps at the summit.
OpenSharing (launched as a Linux Foundation standard) is an open way to share data and AI assets across platforms without copying anything, with on-premises vendors including MinIO, NetApp, and HPE already on board.
Security & compliance updates cover automatic user management, more flexible access rules, simpler private connections for serverless workloads, and broader certification coverage, including HITRUST.
That’s the full list. Part 2 narrows this down to the five announcements I’d actually pay attention to. Part 3 sets out where all of this is taking Databricks.
If you are interested in finding out more, please contact us.
Have a question or want to make an enquiry?
Contact us
