My Take On The Fabric Conference Updates

I am sure you have seen, there has been a lot of Fabric and Power BI news lately. Not surprisingly, Fabric Conference was also last week! 😀

I won't list all the updates here, you can read Arun's blog, or either of the Fabric or Power BI monthly feature summary blogs to go through the whole list:

• FabCon and SQLCon 2026: Unifying databases and Fabric on a single data platform
• Fabric March 2026 Feature Summary
• Power BI March 2026 Feature Summary

Instead, I will highlight what are the biggest game changers for me, and what you definitely should dive in deeper.

Platform

Database Hub (Early Access)

This is a new control plane for everything database: so Azure SQL, Cosmos DB, PostgreSQL, MySQL and Fabric Database, offering estate-wide observability, delegated governance and Copilot-powered insights. I haven't seen or used it yet, but does this really take away the need to use the Azure Portal and / or SSMS?

Planning in Fabric IQ (New)

A new enterprise planning capability that brings financial and business planning directly into Fabric. So we're talking budgets, forecasts, scenarios and targets. To me it sounds like this is aimed at replacing disconnected planning tools (like Anaplan)? It's a co-operation / acquisition with Lumen, but as far as I got from the sessions and community notes, you don't pay Lumen directly but instead it just costs you CU's from the already present capacity.

OneLake + Databricks Unity Catalog (Preview)

They now added the capability to natively read from OneLake through Azure Databricks Unity Catalog. As I see more and more companies use Databricks combined with Fabric, as either bronze/silver or the whole data platform, I reckon this integration will have advantages for people using both platforms. It still looks like Microsoft is keeping their promise of keep developing an open platform.

Fabric Remote MCP (Preview)

If you haven't looked into MCP servers yet, I highly recommend you do. I also haven't dived deep into it, but I started doing some dabbling around with the Power BI modelling MCP server, for minor tweaks and measure creation, and that was already awesome.

Kurt Buhler and Eugene Meidinger have some good posts about the theory, inner workings and how to get started.

Power BI 

Translytical Taskflows (GA)

Translytical taskflows (TT) a.k.a. writeback for Power BI, although that doesn't paint the whole picture because it can do much more. It uses User Data Functions (Python notebooks) in Fabric to update data in a data store of you choice (e.g. Azure SQL, Fabric SQL Database, Lakehouse) from within certain elements in a Power BI report. Now that TT is GA I reckon much more companies will be able / are allowed to explore this. This will be a game-changer, you don't need seperate licenses for e.g. Power Apps or other 3rd party software that can do writeback.

Be aware though, because there are still some limitations, a.o. PBIR (Power BI enhanced report) and PBIP (Power BI Project) formats cannot be used together with TT.

Direct Lake on OneLake (GA)

Direct Lake now works better with OneLake security and it has more modeling features. The guidance is now clear: use Direct Lake on OneLake by default, unless you have a specific SQL endpiont security dependency. For me, import is still the default, but if you have specific requirements around data freshness or loading very large datasets into memory, you could have benefit by using Direct Lake.

TMDL View in web modeling (Preview)

Finally, we have TMDL view in the web! That makes the web modeling so much more complete, you can copy measures, tables or whole models over. Or adjust sources, parameters, anything! They also implemented a side-by-side diff view before applying changes.

Custom Totals in Tables / Matrices (Preview)

I know a few people that will be very pleased with this... :-) There's been a lot of complaining over the years that "my totals are off!". No, that's just the way DAX works... Well, now you can at least decide for yourself if you want to change it.

Modern Visual Defaults (Preview)

I wanted to call out this one, because I'm not sure if I'm actually a fan of this feature. The new default will be smoothed lines, which can sometimes give false data points in a graph. I'd have to check wether this will turn out good.

Developer

Git integration now has selective branching, where you can branch out for a specific feature and only pull items you need.

Furthermore, two new open-source projects have been added:

• Agent skills for Fabric: these are natural language plugins for GitHub Copilot terminal. This sounds like plugging into the Matrix and learning a new skill :) It also means you can use and share skills from and for the community.

• Fabric Jumpstart: these are reference architectures and single-click sample deployments. These look familiar to the earlier industry solutions, but now completely set up with data inside your own tenant.

Closing

With everything taken into account, Microsoft is betting hard on Fabric as the AI data operating sysem. I'm curious what the coming weeks and months will bring.

As I am typing this blog, I'm now actually in the plane to the MVP Summit in Redmond, Washington at the Microsoft HQ! Hopefully I can dive even deeper into a lot of these topics while I'm there!