Many enterprise Power BI semantic models use Azure Databricks as a data source. When building these models, developers and architects face an early and consequential decision: which storage mode to use. Cost, security, and ease of development and tuning all factor in — but report performance is probably the most important of them, because reports that are slow to load are one of the most common causes of end-user dissatisfaction.

In practice, that decision is often made on intuition rather than evidence. To help change that, we’ve published a new white paper, Modern Power BI Architecture Choices for Reporting on Azure Databricks, benchmarking four ways of serving the same Delta tables to a Power BI report:

• Direct Lake on OneLake — over Delta tables in a Fabric lakehouse or warehouse

• Direct Lake on mirrored Unity Catalog tables — shortcuts, no copy

• DirectQuery — on a Databricks SQL warehouse

• Composite Model on Databricks — DirectQuery combined with Import-mode aggregations

Figure: The four Power BI storage modes benchmarked to evaluate their impact on report performance and scalability.

What the results suggest: there’s no universal winner — but there are clear patterns.

Direct Lake on OneLake performed well across the widest range of situations in this benchmark. It’s highly competitive at smaller and mid-size volumes, and for the typical Power BI workload — where reports are used repeatedly throughout the day — it delivers interactive performance without extra modeling effort.

At the top end of the volume curve, the picture shifts. With billions of rows, a Composite Model with aggregations was the fastest and most consistent pattern, staying sub-100ms on queries the aggregation tables can resolve. The caveat is equally clear: that advantage doesn’t extend to queries that fall through to the underlying DirectQuery source, so the payoff depends on how well your aggregations match real user behavior.

These are just the headline findings — the detailed results vary considerably by data volume, cache state, filter scenario, and query type. We’d encourage you to read the white paper for the full picture before deciding on a pattern.

One thing worth noting up front: this is a point-in-time comparison as of June/July 2026, and both platforms are moving quickly. It also measures the end-user query experience within Power BI, rather than raw database execution speed. Treat it as a guide for running your own testing, on your own data.

Explore the white paper for a full walkthrough of through each pattern in detail — the test setup, what was measured, and how results break down by data volume, cache state, and query type.

Visit the Power BI download center to access the paper and other related resources.

Microsoft has refreshed its Power BI sample reports with modern visual defaults, stronger semantic models, mobile-optimized layouts, and newer authoring features to help creators build clearer, more effective reports.

Upgrading Power BI Dataflows Gen1 is now easier with the Dataflows Upgrade Wizard. Now in preview for eligible workspaces assigned to Fabric capacity, the wizard provides a guided experience to upgrade Power BI Dataflows Gen1 items to Fabric Dataflows Gen2 (CI/CD). The wizard preserves key properties of the existing dataflow and assesses each item before you begin, helping you understand the upgrade scope and expected follow-up actions. 

Modernize at your own pace 

Power BI Dataflows Gen1 remains supported in a legacy state, while new feature investment focuses on Fabric Dataflows Gen2 (CI/CD), as shared in a previous post about the future of Dataflows. The Upgrade Wizard gives dataflow owners a guided self-service path to start that modernization without rebuilding their Power Query logic. 

You don’t need to upgrade your full estate at once. Start with a representative set of dataflows, validate the results, and expand at a pace that works for your organization. For detailed migration planning and inventory guidance, review Migrate from Dataflow Gen1 to Dataflow Gen2. 

Build on the benefits of Dataflows Gen2 

Fabric Dataflows Gen2 (CI/CD) builds on the Power Query authoring experience already used in Power BI Dataflows Gen1, while adding capabilities designed for development, operations, and integration across Fabric. 

Dataflows Gen2 supports a broad set of data destinations: Azure SQL Database, Azure Data Explorer (Kusto), Azure Data Lake Storage Gen2, Fabric Lakehouse tables, Fabric Lakehouse files, Fabric Warehouse, Fabric KQL database, Fabric SQL database, SharePoint Files, Snowflake database, and PostgreSQL database. You can begin with the upgraded dataflow and add destinations when you are ready. 

As a native Fabric item, Dataflows Gen2 (CI/CD) supports Git integration and Fabric deployment workflows. Teams can bring dataflow definitions into source control, collaborate through established development practices, and move changes across development, test, and production environments. Refresh history and Fabric monitoring experiences provide operational details that help teams review runs and investigate failures. 

Capabilities such as Fast Copy, Modern Evaluator, and Partitioned Compute improve performance and scale for different workload patterns. Microsoft’s published Dataflow Gen2 cost and performance benchmarks show refresh times improving by 1.6 to 21 times with Dataflows Gen2. Results vary by data, capacity, connector, gateway usage, and query design, so validate your representative workloads. 

Dataflows Gen2 (CI/CD) also benefits from tiered Standard Compute pricing, with a lower rate after the first 10 minutes of each mashup-engine query. Separate meters apply to high-scale compute and Fast Copy data movement. 

Copilot in Data Factory enhances Dataflow Gen2 authoring with natural language. It can generate transformation steps, create queries, and summarize a query and its applied steps. These capabilities help authors understand and review complex query logic, improving productivity during transformation design. 

The bottom line is simple: upgrade now, then modernize when it adds value. 

Upgrade one or more dataflows from the same workspace 

Ready to begin? Open the menu next to a Power BI Dataflows Gen1 item and select Upgrade to Dataflow Gen2 (Preview). The Dataflows Upgrade Wizard opens with the selected item and lets you add other Dataflows Gen1 items from the same workspace. 

Figure: Start the Dataflows Upgrade Wizard from a Dataflows Gen1 item.

The upgraded item preserves its ID, schedule, queries, and connections. Keeping the same item ID helps downstream semantic models and reports continue using the upgraded dataflow without source remapping. The dataflow name also stays the same unless the assessment identifies a required naming change.

After the upgrade is completed, the Fabric Dataflows Gen2 (CI/CD) item replaces the original Power BI Dataflows Gen1 item, and the upgrade can’t be reversed. If you need to keep the original Power BI Dataflows Gen1 item, use Save As to create a separate Fabric Dataflows Gen2 (CI/CD) item instead.

Know what needs attention before the upgrade

Before making any changes, the Upgrade Wizard assesses the Power BI Dataflows Gen1 items in your workspace, giving you a clear view of which dataflows are ready and where you may need to act before or after upgrading:

• Ready to migrate: The wizard didn’t identify a manual step needed before or after the upgrade. Review the known limitations because the assessment doesn’t detect every limitation.

• Needs Attention: You can proceed, but review the reasons and complete the recommended actions.

• Upgrade unavailable: You don’t own the dataflow. Ask the owner to upgrade it or use Take over first.

Figure: The assessment shows which dataflows are ready and which need attention.

A Needs Attention status can currently identify incremental refresh, Bring Your Own Lake storage, linked dataflows, possible DirectQuery consumers, or naming changes. Hover over the status to see the reason, then review the Upgrade Wizard documentation for follow-up actions and more information. The scenarios represented by each assessment status may expand as we further enhance the assessment capabilities.

After the upgrade

Once the wizard finishes, complete these follow-up steps to validate your upgraded dataflows and reconnect any dependencies.

• Refresh each upgraded dataflow to load its data.

• Open and save each downstream dataflow to rebind it.

• Complete any remaining recommended Needs Attention actions.

• Update consumers (for example, semantic models or linked dataflows) using the legacy Power BI Dataflows connector to the modern Power Platform Dataflows connector. The assessment doesn’t detect legacy connector use.

Learn more

Ready to explore the Upgrade Wizard? Review the prerequisites, assessment guidance, and complete workflow in Upgrade Dataflow Gen1 to Dataflow Gen2 (CI/CD) using the Upgrade Wizard. We look forward to hearing about your experience during preview.

If your organization uses Power BI, you own something most companies chasing AI are desperately trying to build. You just may not know it by name.

Let me explain.

The invisible thing behind every report

Every Power BI report you have ever opened sits on top of a semantic model. Every single one. No exceptions. The report is the visible part; the semantic model is the machinery underneath that makes it trustworthy.

What does it do? It translates raw data into business meaning. Somewhere in your organization, someone spent weeks deciding what “revenue” actually means. Gross or net? Booked or recognized? Which currency conversion, on which date? Someone fought over what counts as an “active customer” and whether returns subtract from sales this quarter or the quarter of the original purchase.

Those decisions did not stay in meeting notes. They were encoded into the semantic model: the metric definitions, the relationships between customers and orders and products, the hierarchies that let you roll up a region into a country into a continent. That is why two people opening the same report see the same number, and why the CFO trusts the quarterly dashboard enough to present it to the board.

Figure: Every Power BI report sits on a semantic model, whether its users know it or not.

How did you end up owning one? Think about how Power BI works for most people. Someone opens Power BI Desktop and builds a “report.” That one file quietly contains far more than visuals: the instructions for fetching the data (Import, DirectQuery, or the newer DirectLake), the Power Query steps that clean and shape it, the resulting tables, and the definition of every metric. Publish it, and the platform splits it in two. The visuals become the report people open. Everything else becomes the semantic model. Every report author in your company has been building semantic models for years, often without ever using the word.

Most business users have never heard the term ‘semantic model’. That is fine. It is doing its job precisely because you do not have to think about it. But it represents years of accumulated agreement about how your business measures itself. And that turns out to be exactly what AI needs.

Why AI needs your semantic model

Large language models are remarkable at language and unremarkable at knowing your business. Ask a general-purpose AI, “What was our churn last quarter?” and it faces the same ambiguity your analysts fought over years ago. Which definition of churn? Which customers count? Churn measured monthly and annualized, or measured quarterly?

An AI that guesses at these definitions produces answers that are fluent, confident, and wrong. Wrong in the worst way: plausibly wrong, so nobody catches it until the number shows up in a decision.

This is the problem that Fabric IQ addresses. It is the semantic layer of Microsoft Fabric, and its job is to ground AI in your business language. It takes the concepts your organization has already defined and makes them available to AI agents and to Microsoft 365 Copilot, so that when someone asks a question in plain English, the answer is computed from your definitions, not from a statistical guess.

Here is the part that matters for this blog: Fabric IQ does not ask you to start over. It builds on the Power BI semantic models you already have. Your existing models can directly feed Fabric IQ’s business vocabulary, so the concepts you defined once for reporting now serve chat, agents, and automation. Define “customer” once, use it everywhere.

The context available to AI runs deeper than metric definitions. Semantic models can carry synonyms, so revenue, sales, and turnover all land on the same measure regardless of who is asking. Models can also be deliberately prepared for AI: authors add descriptions, synonyms, and sample questions with approved answers, so Copilot handles the common questions the way your best analyst would. And Fabric IQ can read the reports connected to a model as context too. The names on your report visuals record how your enterprise actually speaks. If a column stored as SalesAmt appears on every dashboard as “Amount,” that tells AI exactly what to call the concept when talking to your users. Years of report building turn out to be years of vocabulary training.

One practical note before anyone schedules a migration project: none of this requires rebuilding your models. Fabric IQ works with the Power BI semantic models you run today, Import and DirectQuery included. The newer Direct Lake models exist in Fabric and are worth a look for other reasons, but nothing about the AI integration forces an upgrade.

Figure: Fabric IQ grounds Copilot and AI agents in your semantic model and learns naming from your published reports.

The same number, everywhere

There is a scenario every data leader dreads. An executive asks Copilot for quarterly revenue and gets one number. The dashboard shows another. Now you have a meeting about which number is right, and trust in both systems drops.

The entire value of wiring AI through your existing semantic models is that this meeting never happens. The report and the AI agent are reading from the same definitions. Same filters, same currency logic, same fiscal calendar. When the numbers match by construction, you do not need to re-audit every AI answer against every dashboard.

Figure: One semantic model means the analyst and the executive get the same answer.

I would go further: an AI rollout that produces different numbers than your reports is worse than no AI rollout. You spent years building trust in your BI numbers. The fastest way to lose it is to introduce a second, slightly different version of the truth and give it a confident voice.

What this means if you want to be a Frontier Firm

Microsoft’s Work Trend Index describes Frontier Firms: organizations that restructure how work gets done around human-agent collaboration. Whatever you think of the label, the direction is clear. AI agents will increasingly answer questions, monitor operations, and prepare decisions that people used to assemble by hand.

Every one of those agents needs to understand your business to be useful. The common assumption is that this requires a massive new data project. For Power BI customers, it mostly does not. The semantic layer is already built. It is sitting behind your reports, refined by years of real usage and real arguments about real definitions.

The practical work is smaller and less glamorous: figure out which of your semantic models are actually trusted, certify and promote them, clean up the ones where two departments define the same measure differently, and promote the good ones into Fabric IQ. That is weeks of curation, not years of construction.

The companies that move fastest on AI will be the ones that recognize which assets they already have, budgets notwithstanding.

You have Power BI. You spent years teaching it what your business means. Time to let your AI learn from the same teacher.

Welcome to the August 2026 Power BI Feature Summary! This month includes updates across reporting, modeling, data connectivity, mobile, embedded analytics, and developer experiences, with a mix of generally available enhancements and new preview capabilities. Let’s take a look at what’s new.

Download Power BI Desktop

Events and Announcements

September 15 | What a Winning Power BI Dataviz Looks Like

Meet the finalists headed to Barcelona to compete for the title of Power BI Dataviz World Champion. Explore the visuals that got them there, hear their tips for creating standout dataviz, and join their virtual cheer squad before they face the ultimate challenge: 20 minutes to create a championship-winning visual live on stage. Reserve your spot.

Community Conference Tickets are Getting Low

The European Microsoft Fabric and SQL Community Conference take place September 28 to October 1 in Barcelona, Spain, bringing together Microsoft experts and community leaders for exclusive content, live demos, and key announcements, along with more than 130 sessions spanning Fabric, Azure AI, Databases, Power BI, and Microsoft Purview. And of course, the live Dataviz World Championships finale!

The most recent event in Atlanta sold out. Don’t miss out and register for the Barcelona event with code FABCMTY200 to save €200.

General

Deprecation of Old File Picker experience in Power BI Desktop

In April we deprecated the old file picker experience in Power BI Desktop. Starting in October, users still utilize versions of Power BI desktop from March 2026 or earlier will no

longer be able to save and share Power BI files to One Drive and SharePoint. Please ensure you update Power BI desktop to keep using this functionality.

Note: No action is required from users as part of this deprecation; this is simply an informational announcement.

Copilot and AI

Updates to required semantic model permissions for Fabric Apps

Currently, users need Build permission on the underlying semantic model to view and interact with a Fabric App created with the data app template. In early September, consumers will need only Read permission on the semantic model, along with access to the Fabric app. This change supports least-privilege of access by no longer requiring the broader Build permission for app consumption. App authors will still require Build permissions on the semantic model.

Learn more about creating a Fabric app connected to a semantic model.

Copilot Summary and Copilot Narrative can now read visuals hidden behind bookmarks

Authors commonly use report bookmarks as tabs to organize content by showing and hiding visuals on a page. In these designs, certain visuals are hidden by default and revealed only when a display-only report bookmark is triggered. Copilot previously skipped these visuals entirely, so its Summary responses reflected only part of the report.

Copilot Summary can now read and consider these bookmark-revealable visuals in its responses. This covers only visuals that are hidden by default and made visible by a display-only report bookmark triggered through a bookmark button or bookmark navigator. Authors can also select these hidden visuals when configuring the Copilot Narrative visual. Visuals are read in place with no changes to the bookmark state, and RLS and OLS permissions stay fully enforced.

To learn more, refer to Summarize a report with Copilot and Create a narrative visual with Copilot for Power BI.

Reporting

Modern visual defaults and customize themes formatting panes (Generally Available)

New reports created in Power BI Desktop and the Power BI service start with the Fluent 2 base theme, giving you a polished, consistent look without formatting each visual. Use the Theme pane to set the base theme, adjust colors to your brand or pick from one of the available color palettes, change text styles, set specific visual properties, page settings, and the filter pane look and feel across your entire report, and import or export custom theme, in both Desktop or web.

Figure: Customizing a report’s base theme and color palette in the Power BI Theme pane.

In this release, we also removed all font overrides from the base theme, so the Text section in the Theme pane now works as intended. Change a font in one place to update titles, callouts, labels, and general text across every visual in your report, including new visuals added later.

For a deeper look at what’s included, see the Deep dive into Modern Visual Defaults and Customizing Theme Improvements and Deep dive into modern visual defaults and formatting your entire report blog posts. Learn more about the base themes and the Theme pane in the Visual defaults in Power BI reports documentation.

Date picker for Slicer visual (Generally Available)

Date pickers combine relative selections that roll forward as time passes with a calendar and slider for manual date range or single date selection. You can anchor relative options to today, the first date, or the last date in your column.

Figure: Enabling single-date selection for the Date picker slicer in the Selection controls settings.

New in this release, the Single date setting under Visual > Slicer settings > Selection controls restrict the slicer to one date at a time—the calendar and slider allow only a single date, and the relative options are limited to selections that return a single date. And you can now clear selections from a header icon in addition to slicer header.

For more information, see the Slicer visual in Power BI documentation.

Center value for donut chart (Generally Available)

Donut charts can now show value in the center, no overlaid card required. Enable the center value and the donut displays it right in the middle.

You control the details in the formatting pane: value format, display units, and font. Add an optional label above or below the value, place an image above or below it, or change the background filling or image of the center. Need something other than the total? Add a custom measure to the center value in the formatting pane.

Figure: A donut center can show a plain total or a custom value, label, image, or background to match your report.

The center value stays in sync with your data: filtering, cross-highlighting, drill-down, and selecting a slice all update it automatically. And because it lives inside the visual, it stays aligned across resizing, mobile, and dashboards, and works with screen readers.

To add a center value, select your donut chart, open the formatting pane, and turn on Center value.

To learn more, refer to Add a center value to a donut chart.

Comments support for reports in org apps

Users can now collaborate on reports in org apps using comments. Comment on report pages and visuals, start discussions with colleagues, and use ‘@mentions’ to notify others and bring them into the conversation. Comments also capture the current report context, helping others understand exactly what you’re seeing when providing feedback or discussing insights.

Figure: Comments pane open on a report in an org app comment box highlighted.

To learn more about, see comments in Power BI reports.

Matrix: Expand and collapse for column headers (Generally Available)

You can now expand and collapse column headers in the matrix visual, just like you already can for row headers. When your matrix has more than one field in the Columns field well, +/- icons appear on the column headers so you can combine groups or reveal more detail directly in the visual. You can also customize the color and size of the +/- icons under Column headers > +/- icons in the format pane.

Figure: Customizing the color and size of expand and collapse icons for matrix column and row headers.

To learn more, see Expand and collapse row and column headers.

Matrix: Set the default freeze state for row headers in the format pane

Report authors can now set the default freeze state for matrix row headers directly in the format pane. Turn the Row headers toggle on or off under Layout > Freeze to control whether row headers stay visible or scroll away when consumers scroll horizontally. This default is saved with the report and applies to everyone who views it.

Figure: Setting the default freeze state for matrix row headers in the Visual formatting pane.

This new setting is especially useful for mobile layouts. On phones, portrait orientation limits horizontal space, so you might want to unfreeze the row headers on your mobile layout to give data values more room, while keeping them frozen on the desktop layout, which typically uses landscape orientation.

Figure: Setting the default freeze state for matrix row headers in a mobile layout.

The right-click Freeze row headers and Unfreeze row headers options are still available and, as before, remain transient — they apply only to the current viewing session, so consumers can flip the state temporarily without changing the report author’s default. To learn more, see Freeze row headers.

OneLake file URLS for report visuals and maps (Generally Available)

You can now use links to image files stored in OneLake as an image source anywhere where Power BI accepts an image URL. This works in the image visual, table and matrix cells, the card visual, button and list slicers, as well as for custom icons in the Icons cell element when you use the Field value format style. OneLake file links work even though they require authentication—Power BI signs in on your behalf when the report loads, so you don’t have to make images anonymously accessible to use them in your report.

Copy the file URL from OneLake and paste it into any image URL field or into a column or measure set as Image URL.

 Example: https://onelake.dfs.fabric.microsoft.com/{workspace-id}/{item-id}/Files/{path-to-image

Figure: Using a OneLake file URL as the image source for an image visual in Power BI.

Reference layer maps and shape map files can also be linked from OneLake file URLs. Keep your files in OneLake and use them in your Power BI reports today!

For more information, see the Display images in a table, matrix, or slicer and Image visual in Power BI documentation.

Outer padding for bar, column, line, ribbon, and waterfall charts in formatting pane (Generally Available)

Column, bar, clustered, line, ribbon, and waterfall charts now include an Outer padding setting under Layout, giving you control over the space between the edges of the plot area and the first and last categories. This is helpful when you want your chart to fill the visual—especially in scenarios where you’ve turned off the category axis and are relying on a legend, like a clustered column chart used as a comparison view. Then, you can simply resize the visual to get the look you want on your report page. Reducing outer padding on line charts with categories also lets you show more data before a scrollbar appears.

Figure: Adjusting outer padding and spacing between categories to control how charts fill the plot area.

To use it, select the visual and, in the Format pane, expand Columns, Bars, or Lines > Layout, then set Outer padding to 0%. To fully fill the plot area, also set Space between categories to 0%.

For more information, see the Customize x-axis and y-axis properties documentation.

Azure map reference layer shape-matching improvements (Generally Available)

Shape matching in Azure Maps reference layers now delivers significantly better performance. When you match your Power BI data to shapes in a reference layer—for example, joining sales data to a set of custom territories or regions in a GeoJSON file—reports now render faster, and interactions like filtering, cross-highlighting, and zoom respond more smoothly. No configuration change is required; existing reports automatically benefit from the improvements.

Figure: Improved Azure Maps reference layer performance, with faster rendering shown in the Performance analyzer.

For more information, see the Add a reference layer to the Azure Maps Power BI visual documentation.

Azure map loads filtered selection and autozoom (Generally Available)

The Azure Map visual has a 30,000 data point rendering limit, and previously, points beyond that limit didn’t display—even when you filtered the report to focus on a smaller area whose data hadn’t originally loaded. Now, when you filter to a smaller selection, the visual reloads to include the newly relevant data points that weren’t part of the original 30,000. This means you can start with a broad view of your data and drill into any subset—by region, category, time period, or through cross-filtering—and be confident the map shows the complete picture for what you’ve filtered to, not just what fit in the initial load.

Figure: Filtering the report to Washington reloads the relevant ZIP code shapes and automatically zooms the Azure Maps visual to the selected area.

The Azure Map visual also now automatically zooms to fit the currently filtered area. When you or a report viewer applies filters, use slicers, or cross-filter from another visual, the map reframes to the extent of the visible data instead of staying at the original viewport. This keeps the map focused on the data being analyzed and removes the need to manually pan and zoom after every filter change.

For more information on Azure Map, see the Azure Maps visual for Power BI documentation.

Slicer visual dropdown border color, open icon, and hierarchy expand and collapse icon colors in formatting pane (Generally Available)

Based on your feedback, we’ve added more slicer formatting options beyond the selection icon, giving you greater control over the slicer’s appearance. When you add more than one field to the slicer to have a hierarchy in vertical list or dropdown style, make sure the icon color matches your report. And dropdown style, offering a compact slicer to your report has more formatting options.

Figure: Customizing the slicer dropdown border, open icon, and accent bar in the Visual formatting pane.

The new top-level Dropdown section in the Visual formatting pane lets you customize the dropdown box in both its open and closed states, including border color (with conditional formatting), rounded corners, and the open icon color (with conditional formatting) and transparency. A new Accent bar option adds a color bar to the dropdown box with settings for color (with conditional formatting), position, transparency, and width.

For hierarchy slicers, the top-level Hierarchy section lets you color the expand and collapse icons (with conditional formatting) and choose which icon to use.

For more information, see the Slicer visual in Power BI documentation.

Modeling

More control over Semantic Model refresh in Power BI Service

We’ve added new refresh options in Power BI Service to give you greater control over how semantic models are refreshed.

Previously, selecting Refresh always performed a schema sync followed by a data refresh. While this works well in most scenarios, there are times when you may want to refresh data without updating the model schema or vice versa. For example, in Direct Lake semantic models, a Lakehouse table might have changed by adding new columns, but you may want to load the latest data without bringing those schema changes into the model.

Figure: New model refresh options in Power BI Service Model view.

The Refresh button now includes three options:

• Refresh schema and data: Updates the schema and then refreshes data.
• Sync schema only: Applies schema changes from the data source, such as new columns or data type changes.
• Refresh data only: Loads the latest data while preserving the current semantic model schema.

In addition, refresh operations can now be performed at the table level. This allows you to refresh a specific table instead of the entire semantic model and choose whether to refresh its schema, data, or both.

Figure: New table refresh options in Power BI Service Model explorer.

These enhancements provide more flexibility and control over semantic model maintenance, helping you refresh exactly what you need while minimizing unnecessary changes and processing.

Refer to the Data refresh in Power BI documentation for more information.

Convert Direct Take Tables to Import Storage Mode in Web Modeling (Preview)

Direct Lake on OneLake supports composite models by allowing for mixing Direct Lake tables with Import and DirectQuery tables. You can now convert individual Direct Lake tables to Import storage mode directly in Power BI web modeling, giving you a seamless way to build a Direct Lake composite (mixed mode) semantic model.

Open a Direct Lake on OneLake semantic model in web modeling, select one or multiple tables, and change the Storage mode from Direct Lake to Import in the Properties pane. The modeling flow guides you through configuring a connection and refreshing the table into Import mode. Because Direct Lake and Import tables support regular relationships, performance remains unchanged.

Figure: Change the Storage mode property from Direct Lake to Import to convert a single table. A warning dialog will be surfaced before continuing.

This feature will be rolling out over the next few weeks. Refer to the Direct Lake in web modeling documentation for more information.

Direct Lake Calculated Columns (Preview)

You can define calculated columns directly in your semantic model using DAX in web modeling and Power BI Desktop, without changing the table storage mode or modifying data upstream.

Key behaviors

• Only the User Context expression context is supported.
• User-context-aware DAX functions such as USERCULTURE() are supported.
• Columns do not materialize and are evaluated at query time.
• Security context is respected (RLS and OLS).
• Columns cannot be used in relationships because they do not materialize.
• Expressions that do not reference user-aware functions or secured columns behave like any other DAX calculated column.

Figure: Creating a concatenated location field.

Figure: Month names displayed using USERCULTURE().

“Customer Location” returns the same value for every user unless the security context differs. “Month Name” adapts to the language of the person viewing the report.

Figure: Month names are localized based on the user’s culture using USERCULTURE().

No preview switch is required – open a Direct Lake on OneLake semantic model in the web or edit in Desktop, add a calculated column, and start authoring in DAX. To learn more, see the Create calculated columns in Power BI Desktop documentation.

Data connectivity

Accessibility, Dark Mode, and Usability Upgrades

Theming improvements for better accessibility

In this release, we’ve invested in a range of theming improvements that address color contrast and readability across both the light and dark themes. Users who rely on higher-contrast visuals will notice cleaner, more legible interface elements throughout the Power Query UI.

Highlights:

• Improved color contrast across interface elements in both light and dark themes.
• More consistent theming that keeps controls, text, and icons clearly distinguishable.
• A stronger foundation for accessibility that benefits all users, regardless of theme preference.

Figure: Better contrast for connectors and white background for icons.

Better contrast for connector icons in Dark Mode

Connector icons are how you quickly recognize your data sources, but in Dark Mode, icons with transparent backgrounds could blend into the surrounding UI and become hard to identify. We’ve fixed that. Now, when a connector icon has a transparent background, it automatically receives a white background in Dark Mode.

The result: connector icons stand out clearly and stay easy to identify across the new Power Query UI, making it faster to find and select the right data source briefly.

To learn more, refer to the Get data in Power BI Desktop documentation.

Mobile

Rotate view in the mobile report footer

The Rotate view button is now available in the Power BI Mobile report footer, letting you switch layouts with one tap for larger visuals and a wider view—even when your phone is locked in portrait orientation.

This is especially helpful for reports that don’t have a mobile-optimized layout, which open in portrait by default. Just tap Rotate view to get a wider, more readable view, then tap again to switch back.

This feature is rolling out soon to Power BI Mobile for iOS and Android – stay tuned!

Figure: Tapping Rotate view in the report footer to get a larger view of the report.

To learn more about the actions in the report footer, refer to the Explore reports in the Power BI mobile apps documentation. To learn how to build a mobile-optimized layout for your reports, see the mobile layout documentation.

Export data to Excel in the Power BI mobile app

Export data from Power BI report visuals directly from your phone or tablet. With Export data now available in the Power BI mobile app for iOS and Android, you can quickly take the data behind a visual into Excel for further analysis or sharing, without switching to the Power BI service on another device.

To export, open a report visual in focus mode or open the visual’s More options (…) menu, and select Export data. The export starts automatically using the supported Excel format for that visual. The exported data reflects your current filters, slicers, drill state, permissions, and row-level security. When the file is ready, use your device’s sharing options to open, save, or share it.

Figure: Exporting data from a report visual to Excel in the Power BI mobile app, via focus mode or open the visual’s More options (…) menu.

To learn more, see Export data from a Power BI visualization.

Embedded Analytics

Enhancements to Power BI Embedding in SharePoint Online

Embedding Power BI reports and report pages in SharePoint Online is now easier. With our new UI, you now have the option to directly select the workspace you want to embed in SharePoint, rather than copying and pasting the full URL. Simply navigate to SharePoint online, add Power BI content, and the new menu will allow you to browse workspaces and reports you have access to.

Alongside embedding reports and report pages, you can now also embed a visual. Simply select the “Embed a single visual” toggle to select the specific visual you want to embed.

Figure: Embedding a Power BI visual with the new SharePoint Online Experience.

More details about these enhancements can be found in the Power BI and SharePoint Documentation.

Developers + API’s

Faster PBIP development with instant reloads and VS Code integration

A faster PBIP editing experience is here. If you’re working with Power BI Projects (PBIP), you can now edit project files outside Power BI Desktop and reload those changes instantly, without restarting the application. Power BI Desktop automatically detects changes made to your files and prompts you to apply them with a single click.

Figure: Apply external changes prompt in Power BI Desktop.

To make editing even easier, we’ve also added a built-in entry point that opens your project directly in Visual Studio Code.

Figure: Open Visual Studio Code entry point in Power BI Desktop.

Whether you’re making edits in VS Code or using external tools to update your project files, you can now move seamlessly between your tools and Power BI Desktop without breaking your flow. Learn more

Third party custom visuals 

Need more visual options for your report? Power BI reports support custom visuals built to fit your exact needs. Browse hundreds of options in AppSource, also available directly from Power BI Desktop, or develop your own and add it to a single report, share it across your organization, or publish it to AppSource for the community. You can even develop a custom visual tailored to your specific requirements with AI coding tools like GitHub Copilot Free. 

The following are some updates from the community-based custom visuals you can use in your reports today. 

PowerGantt Chart by Nova Silva

The PowerGantt Chart helps teams plan, track, and communicate project timelines directly in Power BI. This month’s update is shaped by your feedback, and we’re grateful to the community of users whose suggestions keep guiding how the visual evolves. We’ve added three enhancements that give you finer control over how schedules and baselines are displayed.

What’s New?

• Baseline customization – Set the baseline bar size relative to the actual task bar and move milestone icons onto the baseline so planned versus actual delivery is easy to compare.
• Squared bar edges – Choose squared edges on item bars for a cleaner, more technical look.
• Vertical grouping – Group tasks vertically to organize large project structures by phase, team, or workstream.

Whether you’re managing product launches, construction schedules, or multi-team programs, these refinements make complex timelines easier to read briefly. Project managers, PMOs, and analysts gain sharper baseline comparisons, cleaner layouts, and better navigation across busy projects. Try the PowerGantt Chart for FREE now on your own project data by downloading it from the AppSource.

Questions or remarks? Visit us at: NovaSilva.

Slicer Panel by Powerviz

Powerviz Slicer Panel is an advanced, all-in-one slicer that combines lists, dropdowns, sliders, buttons, date pickers, switches, and text slicers in one panel. With built-in templates, full customization, and import/export, it makes filtering smarter, simpler, and more interactive.

Key Features

• Multiple Slicers in One Build: Bring together a wide variety of slicer types, including dropdown, expanded list, date, button, text, slider, and switch slicers, all within a single, unified panel, to filter your entire report from one place.
• Display Mode: Switch between Canvas, Pop-up, and Pop-up by Slicer.
• Orientation & Position: Arrange slicers and dock the panel top, bottom, left, or right.
• Panel Header: Show slicer icon, text, applied count, search, clear, and saved slicers.
• Slicer Action Icons: Edit, expand/collapse, sort, clear, rank, duplicate, or delete each slicer.
• Add Tabs: Group slicers into tabs for organized filtering.
• Saved Slicer: Store filter selections as reusable states.
• Conditional Formatting: Style slicer values with rule-based formatting.
• Advanced Search: Search across slicers and fields instantly.

Other features included Templates, Import/ Export Themes, Search within Slicer, Ranking, Sorting, and more.

Try Slicer Panel visual for FREE from App Source

• Check out all features of the visual
• Step-by-step instructions
• YouTube Video
• Learn more about visuals
• Follow Powerviz on LinkedIn

Figure: New slicers. Smarter filtering. One unified panel.Take your Power BI reports to the next level with the advanced Slicer Panel by Powerviz.

Figure: From multiple slicer types to display modes, orientation, and saved slicers, every feature is designed to make your filtering more flexible, interactive, and effortless.

Drill Down Map PRO by ZoomCharts

When visualizing data with geographic coordinates, what better way to do it than literally placing it on a map? That’s why map charts are a growingly popular visualization type in Power BI reports, and the Drill Down Map PRO custom visual by ZoomCharts expands on the capabilities of map charts.

• Node Clustering: Multiple nearby nodes can create clusters and even display the values as pie charts. Simply zoom in to drill down.
• Base Layer Customization: Choose between AzureMaps or any custom tileserver, use your own images as the base layer, or disable it entirely.
• Custom Shape Layers: Enable up to 10 individually customizable shape layers. Use preset shapes or import your own KML/GeoJSON files.
• Conditional Formatting: Automatically apply color fill to each area by comparing their values against other shapes or by using each shape’s own reference value.
• And More: Paginated tooltips, custom tooltip fields, auras, node images, lasso tool.

Figure: Reveal deeper insights with clustering, pie charts, custom shape layers, and flexible base map options tailored to your reporting needs.

Figure: Customize every aspect of your map with conditional formatting, node images, auras, advanced tooltips, and intuitive selection tools.

Drill Down Map PRO works incredibly well with other visuals by dynamically cross-filtering data, enabling you to build even more insightful and user-friendly Power BI reports.

Get Drill Down Map PRO on AppSource

Closing

That’s a wrap for the August 2026 Power BI Feature Summary. This month’s release includes improvements across reporting, modeling, data connectivity, mobile experiences, embedded analytics, and developer workflows, along with continued investments in Copilot, accessibility, and Direct Lake capabilities. We hope these updates help you build, manage, and share data experiences more effectively. As always, we appreciate your feedback and look forward to hearing how you’re using these features in your organization.

Power BI Projects (PBIP) make reports and semantic models file-based, enabling source control, automation, and other developer workflows. You can edit those files directly in a code editor or through external tools. Until now, changes made outside of Power BI Desktop wouldn’t appear until you restarted the application. With this update, Power BI Desktop detects those changes automatically, so there’s no restart required.

Edit anywhere and see changes immediately

Power BI Desktop now detects changes made directly to PBIP files, regardless of whether those changes come from an editor, external tool, or automated process.

When a PBIP file changes on disk, Desktop detects the update and prompts you to reload the latest version of your project.

Apply external changes prompt in Power BI Desktop

Selecting Apply external changes reloads the updated files into Power BI Desktop without restarting the application.

This creates a much smoother experience when moving between Power BI Desktop and the tools you already use.

Open your project directly in Visual Studio Code

Power BI Desktop now includes an Open in VS Code option for PBIP projects.

Selecting it launches Visual Studio Code with the project folder already loaded as the active workspace, giving you immediate access to report and semantic model files.

Open Visual Studio Code entry point in Power BI Desktop

Benefits

These updates make it easier to use PBIP files with external tools and automated processes while continuing to work in Power BI Desktop.

Typical scenarios include:

• Editing project files in Visual Studio Code.
• Updating files through scripts or automation.
• Working with AI-assisted development tools that modify PBIP artifacts.

Get started

Trying it out is simple:

1. Open a PBIP project in Power BI Desktop.
2. Select Open in VS Code.
3. Make a change to your project files and save it.
4. Return to Power BI Desktop and select Apply external changes when prompted.

You can now move seamlessly between Power BI Desktop and your editing tools, making it easier than ever to work with PBIP files wherever your workflow takes you.

Explore the documentation to learn more about Power BI Projects (PBIP).

One of the most common questions I hear from customers as they move to an allocation-based billing model is:

“What levers do I actually have if I want to control how many resources my workloads consume?”

Many customers are looking for a way to prevent specific workloads from scaling aggressively and consuming more resources than they’re comfortable with. That’s where Custom SQL Pools can help significantly.

Custom SQL Pools, currently in preview and soon to be generally available, were designed to give customers more control over how resources are allocated across workloads. Their primary purpose is workload isolation and governance, but with allocation-based billing, governing resource allocation can have a direct impact on overall consumption.

Figure 1: Custom SQL Pools – application based classification

Understanding the tradeoff

Every workload wants resources.

Your Power BI dashboards want low latency. Your ETL workloads want throughput. Your ad hoc analysts want concurrency.

Without workload governance, multiple workloads can compete for the same resources and can scale aggressively when they become active. While that’s desirable for some latency-sensitive workloads, customers often want different workloads isolated from one another. A business-critical Power BI dashboard may require priority access to resources, while an ETL process may be allowed to run longer in exchange for consuming fewer resources. Custom SQL Pools provide those isolation boundaries and governance controls, allowing customers to make intentional performance versus consumption tradeoffs.

A simple example

Imagine a bursty reporting workload. Without a Custom SQL Pool, a query might:

• Scale to 10 vNodes.
• Run for 10 seconds.
• Be billed based on those 10 allocated vNodes during the allocation interval.

Now imagine that same workload is assigned to a Custom SQL Pool limited to 50% of the warehouse resources.

The query might:

• Scale to only 5 vNodes.
• Run for 20 seconds instead of 10.
• Be billed against a smaller allocated resource footprint.

The query took longer to complete, but it consumed fewer allocated resources.

In other words:

You’re trading performance for lower resource allocation. For many workloads, that tradeoff may not make sense. For some workloads, it absolutely does.

The real-world scenario

Numerous customers use Custom SQL Pools effectively to control how resources are allocated to their workloads.  

One customer had a long-running ETL process that wasn’t latency sensitive. For this customer, they didn’t care whether the load finished in 20 minutes or 30 minutes.

What they did care about was controlling how aggressively the workload consumed resources. They used SQL Pools to restrict the throughput available to that ETL process, effectively smoothing resource usage over a longer period of time. The workload ran longer, but consumed resources at a more controlled rate to further utilize the 24-hour smoothing behavior of background operations.

That’s a great example of where workload governance and consumption governance become closely related.

Good Candidates for SQL Pools

Custom SQL Pools are particularly useful when you have workloads that:

• Are not latency sensitive.
• Can tolerate longer runtimes.
• Generate large bursts of activity.
• Compete with other workloads for resources.
• Need more predictable resource consumption.

Examples include:

ETL and Data Loading

Nightly data loads often care more about successful completion than finishing a few minutes faster.

Restricting resource allocation may be a perfectly reasonable tradeoff.

Background Processing

Data preparation, maintenance operations, and other non-interactive workloads can often run with fewer resources.

Power BI Reporting

Many organizations want to ensure reporting workloads cannot consume all available warehouse resources during business hours.

Using Custom SQL Pools to allocate a fixed percentage of resources to reporting workloads provides predictable behavior while preventing contention with other workloads.

What Custom SQL Pools are not

It’s equally important to be clear about what SQL Pools do not do.

They are not:

• A spending limit.
• A monthly budget cap.
• A mechanism to stop charges after a threshold.
• A “$100/month maximum” control.

Instead, they serve as guardrails that define the maximum share of warehouse resources each workload can access. While they are not a cost optimization feature in themselves, the governance they provide can indirectly affect overall consumption.

Looking ahead

We’re continuing to invest in workload management capabilities.

A common ask we hear is for resource governance to follow the identity rather than the application. For example, a customer may want executive reporting users, production service principals, or critical business processes to receive different resource allocations than ad hoc analyst workloads.

Identity-based classifiers are one example of where we’re headed, but the broader goal is to give customers more flexibility in how workloads are governed. We’re also working toward making the built-in workload boundaries configurable, allowing customers to adjust resource allocation between the default SELECT and NONSELECT workload groups without needing to define custom classifiers. For customers that simply want different resource allocations for reporting and ingestion workloads, this provides a simpler path to workload governance.

Ultimately, the goal is to give customers more explicit control over how resources are allocated and governed, making it easier to align resource allocation with business priorities while managing performance and consumption in a predictable way.

Takeaway

Custom SQL Pools is primarily a workload management feature, not a cost-control feature. However, in an allocation-based billing model, governing resource allocation and governing consumption become closely related.

If you’re willing to let a workload run a little longer, Custom SQL Pools can provide a practical mechanism for reducing the resources that workload is allowed to consume.

For many customers, that’s exactly the lever they’re looking for. Get started with Custom SQL Pools today! Learn more in the documentation: Configure Custom SQL Pools in the Fabric Portal.