May 11, 2026

Power BI Embedded with Direct Lake — What You Need to Know

ByPascal Kiefer

Direct Lake is one of the most significant storage mode innovations in Microsoft Fabric — and it now works with Power BI Embedded. This means you can deliver real-time analytics from OneLake directly to your embedded portal, without any data imports or scheduled refreshes. Here is what it means in practice, what you need to know technically, and how the Embedsy Portal handles it for you.

What is Direct Lake?

Direct Lake is a storage mode for tables in a Power BI semantic model, available in Microsoft Fabric. Unlike Import mode — which copies data into the semantic model on a schedule — Direct Lake reads data directly from Delta tables stored in OneLake, the unified storage layer in Microsoft Fabric.

This means your reports always reflect the latest data, without waiting for a refresh to complete. And because Direct Lake uses the same VertiPaq engine as Import mode, query performance is significantly faster than DirectQuery, which has to round-trip to the source database for every query.

For a full technical overview, see the Microsoft Direct Lake documentation.

Why Direct Lake matters for embedded analytics

For organizations delivering analytics through an embedded portal, Direct Lake changes what is possible.

Always-current data

With Import mode, your reports are only as fresh as your last scheduled refresh. With Direct Lake, data updates in the lakehouse are immediately reflected in reports — no refresh lag, no stale dashboards. For use cases like operational monitoring, financial reporting, or customer-facing portals where data freshness matters, this is a meaningful improvement.

High performance at scale

Direct Lake loads parquet-formatted files directly into memory using the VertiPaq engine. This delivers query performance that matches or exceeds Import mode, even on large datasets — without the overhead of importing and storing a copy of the data.

No data duplication

Import mode stores a copy of your data inside the semantic model. Direct Lake eliminates that duplication — data lives once in OneLake and is accessed directly. This reduces storage costs, simplifies data governance, and ensures a single source of truth across your Fabric environment.

DirectQuery fallback

Depending on the Direct Lake variant used, the model may fall back to DirectQuery mode in certain scenarios — for example when using Direct Lake on SQL endpoints and the data cannot be loaded directly. This fallback is handled transparently, though it may result in slower performance for those queries. For most production workloads on an appropriately sized Fabric capacity, fallback should be rare.

Two technical requirements to be aware of

Power BI Embedded with Direct Lake mode has two specific technical requirements that differ from standard embedding. Both are handled automatically by the Embedsy Portal — but if you are building your own solution, you need to implement these explicitly.

1. V2 embed token required

Direct Lake mode requires the newer V2 Generate Token API — the older V1 APIs are not supported. The V2 API supports generating tokens for multiple reports and datasets in a single call, which is both more efficient and the only option compatible with Direct Lake.

See the Microsoft documentation on generating embed tokens for full details on the V2 API.

2. Fixed Identity cloud connection required

When using a service principal to embed Direct Lake reports — which is the standard approach for embedding for your customers — you must configure a Fixed Identity cloud connection on the Direct Lake semantic model. This is a dedicated connection that uses the service principal as the authentication method, rather than SSO.

This is a one-time configuration step done in the Fabric workspace, but it is a prerequisite that is easy to overlook. Without it, embedding will fail for Direct Lake semantic models.

Microsoft provides a step-by-step guide: Specify a fixed identity for a Direct Lake semantic model.

Fabric capacity is required

Direct Lake semantic models require a Microsoft Fabric capacity — they do not work on Power BI Embedded A SKUs alone. If you are currently using an A SKU for embedding, you would need to move to an F SKU (Microsoft Fabric capacity) to use Direct Lake. F SKUs support both Power BI Embedded and the full Fabric platform, making them the recommended choice for most organizations already working within the Fabric ecosystem.

For a detailed breakdown of capacity options and costs, see our post on Pricing of Microsoft Fabric and Power BI Embedded.

Using Direct Lake with the Embedsy Portal

The Embedsy Portal already supports Power BI Embedded with Direct Lake mode — no additional configuration or custom development required on your side. Once the Fixed Identity connection is configured in your Fabric workspace, add the report as you would any other. The portal handles V2 token generation automatically, and your users see real-time data with no refresh delays.

Steps to get started

Make sure you are on a Microsoft Fabric capacity (F SKU).
Create or migrate your semantic model to use Direct Lake storage mode. See the Direct Lake overview on Microsoft Learn.
Configure a Fixed Identity cloud connection on the semantic model in your Fabric workspace. See Microsoft's guide on specifying a fixed identity.
Add the report to the Embedsy Portal as you would any other report — no additional steps required.

Have questions about any of these steps? We are happy to walk you through it.