Dan Williams
I just came back from the Databricks Data & AI Summit and already started playing around with some of the new features they announced.
Databricks packed the 2025 Data & AI Summit with news that reshapes how engineers, data leaders and business users will build on the Lakehouse. And I have to say I’m more than pleasantly surprised with the innovation that they continue to push.
Here are the 6 releases that stood out to me. I’ll talk about what they are, why they matter, and what the early-day limitations are.
1
What is it? A no-cost, full-fidelity Databricks workspace that replaces the old feature-limited Community Edition. Learners get a small cluster, Unity Catalog, Genie, sample notebooks and more, and they're able to use a personal email to sign up.
Why it matters? It removes the cost barrier for students, hobbyists and new hires. Also demos and PoCs no longer accrue compute charges, accelerating adoption and advocacy. And third, it familiarises newcomers with the real Databricks UI, not just a sandbox with limited features.
What are the current limits? We get tiny clusters with restricted DBUs, meaning no heavy ML training or heavy computational workloads. Also production workloads still require a paid tier.
2
What is it? A managed, “base-tier” Postgres service (built on Neon) that keeps hot rows in memory while cold data lives inexpensively in object storage (S3 or ADLS). Delta and Iceberg files remain fully open allowing you to exit gracefully if needed.
Why it matters? It cuts OLTP costs while retaining familiar Postgres semantics. Another great feature is that you can easily branch off a new clone for testing or for AI agent workloads. And third, it offers tight Lakehouse integration: Unity Catalog governance, Delta Sharing and direct Databricks SQL queries.
What are the current limits? It's a brand new architecture so we might see maturity gaps. Databricks believes that Lakebase is the future of all databases so I'm really bullish on this announcement.
3
What is it? In a nutshell Databricks Apps are managed containers (Streamlit, FastAPI, React etc.) that inherit Unity Catalog permissions. They help teams to ship secure data-centric applications without DevOps complexities.
Why it matters? One platform hosts everything: data, backend and UI. This way you get simpler pipelines and faster insight delivery.
Databricks Apps are also secure by default, with governance inherited from the Lakehouse.
What are the current limits? Out-of-the-box styling is minimal so teams still need front-end polish for a production-grade look and feel.
4
What is it? A framework for production-grade GenAI agents that auto-generate evaluation tests, search prompt/vector/fine-tune combinations and plot cost-vs-quality trade-offs.
Why it matters? Objective numeric scoring replaces “gut-feel” evaluation. Also agent Bricks helps teams tune for the cheapest model that still clears quality targets. Third, it offers rapid feedback loops, allowing business users to refine behaviour without deep ML skills.
What are the current limits? Unfortunately it's still in beta. Also synthetic tests can miss edge cases. But most importantly, sensitive data must be handled with care.
5
What is it? This is by far the best announcement due to it's potential. It's a free, AI-powered migration tool that parses, converts and validates Teradata, Oracle, SQL Server and other legacy code into Databricks SQL.
Why it matters? This is easy. It can cut months of manual rewrite work. Also there's no licence fee: you only pay for Databricks compute during validation.
What are the current limits? Like with any LLM-based converter, logic can be mistranslated. This happens more often than not but the tool has incredible potential.
6
What is it? A single portal for business users that unifies search, chat, dashboards, alerts and catalog browsing.
Why it matters? It simplifies data discovery with one intuitive entry point. It also promotes self-service analytics and cross-team collaboration.
What are the current limits? Adoption relies on change-management, especially in large enterprises with "legacy" BI portals.
This year’s Summit signals a clear strategy: lower the barrier to entry, converge the stack, and push GenAI from experimentation to governed production. While each release comes with early-stage caveats, the direction is clear in my opinion.
But what announcements are in your top 5? What new feature are you most interested in?
If you enjoyed this article, connect with me on Linkedin as I’d love to know how you’re taking advantage of Databricks in your organisation.
Share this:
