During Build 2025, Microsoft announced important new features for Microsoft Fabric. Prior to the conference, we spoke with Arun Ulag, Corporate Vice President Azure Data at Microsoft, who is responsible for these developments. One thing is clear: Microsoft Fabric is extremely important for the company, but it could also become much more important for Microsoft customers.
Microsoft Fabric is a fabric in the familiar sense of the word, i.e., a conceptual layer that connects separate components. Unlike many other fabrics, however, Microsoft Fabric is also a product, a platform. This means that organizations purchase Fabric capacity on Azure when they want to use it. For many other fabric providers, it is an addition to existing components, which can then be better integrated with each other. With those suppliers, you often cannot purchase it as a line item on your monthly bill. It is simply there, and the components that support it make use of it.
The fact that Microsoft is turning Fabric into a true SaaS service indicates that it sees a lot of potential. From a commercial perspective, of course, because now it can sell something else from Azure. But we also think it shows that Microsoft Fabric is a very important development for the company. This should make Azure even more valuable to customers. That should encourage them to do even more within Microsoft’s cloud environment, which is obviously good for the company in terms of revenue, and so on.
We talk to Ulag about three new additions to Microsoft Fabric: native Cosmos DB integration, a ‘Chat with your data’ feature, and a digital twin builder.
Microsoft Fabric: the first 18 months
Microsoft Fabric has experienced impressive growth in the year and a half since its launch, Ulag says. He quotes Satya Nadella from a recent earnings call, who called it the fastest-growing data and analytics product Microsoft has ever launched. More than 21,000 organizations, including 70% of Fortune 500 companies, now use the platform.
Microsoft Fabric started about 18 months ago as an analytics platform, but has since undergone a significant evolution. “Last November, we took a major step forward by adding real-time intelligence to Fabric,” Ulag said. That immediately made it much more interesting for OT use cases, he said. SQL Server was also added to the platform as a native workload.
Microsoft is now taking another big step by adding Cosmos DB (a NoSQL database) as a native workload to Fabric, alongside the existing native integration with SQL Server. This is the first of three major announcements Microsoft is making at Build 2025 regarding Fabric. “With this addition, Fabric is evolving from an analytics platform to a full data platform,” Ulag said.
Native Cosmos DB integration
Until now, it was already possible to add data from various sources, including Cosmos DB databases, to OneLake, Fabric’s data lake component. However, this was not yet native integration. That is now in public preview.
Why is it important to have native integration of Cosmos DB with Microsoft Fabric? Ulag explains what needs to be done if your databases are not running natively in Fabric: “If you use Cosmos DB or SQL, you still need to set up a separate Cosmos DB instance or SQL instance. You have to configure capacity, manage clusters, configure networks, set up storage, and manage credentials.” All of this places a considerable burden on end users.
With native integration, none of this is necessary anymore, is the point Ulag wants to make. Developers can use Cosmos DB in Fabric without worrying about the underlying infrastructure. This fits seamlessly into Fabric’s Software-as-a-Service approach, where Microsoft handles the complex infrastructure details so developers can focus on creating business value.
The integration also has important benefits for building AI applications. “The distinctions between SQL databases, NoSQL or document databases, in-memory databases, analytical databases, real-time databases, and vector databases are rapidly blurring,” says Ulag. “Developers are mixing and matching these capabilities to build new AI applications and workloads.” The idea here is that this will be much smoother if they use Microsoft Fabric as a foundation.
‘Chat with your data’: Power BI gets an AI boost
The second major announcement has to to do with something Ulag calls ‘Chat with your data’. This new functionality focuses on Power BI, Microsoft’s business intelligence tool, which, according to the company, has more than 400,000 paying organizations and 30 million monthly active business users. It has also been part of Microsoft Fabric from the outset.
The ‘Power BI agent’ enables users to ask questions about company data using natural language. The system uses the entire Power BI database (which is already in OneLake, Fabric’s data lake) and presents answers with Power BI visualizations. Ulag emphasizes that it is important for this agent to provide identical answers to what users would see in Power BI itself. Microsoft has therefore invested a lot of time in this. “This is essential for maintaining confidence in the results,” he rightly notes.
An important component for the quality of this new functionality is the semantic model behind every Power BI report. This model contains business-friendly names, relationships between tables, measures, definitions, calculations, and hierarchies. It is a layer of business logic on top of structured data. This is not new, by the way. Microsoft initially developed the technology voor semantic models for SQL Server about ten years ago. Such a model can be very small, but it can also map an entire enterprise environment.
The reason having a semantic model is important with the new functionality of Power BI is that “LLMs struggle with structured data without a semantic model,” explains Ulag. With 10 million semantic models already running in the Power BI service, he believes customers have a huge head start.
‘Chat with your data’ will not only be available in Power BI itself, but will also appear soon in Microsoft Teams and later in Microsoft 365 Copilot.
Digital Twin Builder for modeling physical and logical
The third announcement is a Digital Twin Builder. This addition enables users to model physical or logical entities in Fabric and connect them to various data sources that Fabric can already ingest data from. This provides a fully digital representation of a machine, an entire factory, or a business process.
Note, however, that this is an ‘image’ consisting solely of data, not a visual representation of an aircraft engine, for example. But if you put all the data in the right relationship to each other within a digital twin in Microsoft Fabric, you have all the data you need to gain a good insight and overview. The idea is that, thanks to the relatively recent addition of real-time intelligence, you also retain that insight and overview. For example, telemetry from a machine can be linked directly to the digital twin model, providing real-time insight into the performance of that machine or of an entire factory with machines.
What about data sovereignty?
An important question these days, especially outside the US, is what happens to the sovereignty of organizations’ data in these kinds of environments. Since Microsoft Fabric seems to be tightening its ties with the data even further, we can imagine that customers will have questions about this. Is an organization’s data safe in this new SaaS offering from Microsoft?
When we ask Ulag this question, he emphasizes the openness of the environment. Microsoft Fabric always returns data in an open format, such as Apache Parquet, Delta Lake, and Iceberg. In addition, Microsoft uses the same REST API for Fabric that it has been using for years (Azure Data Lake Gen2), which customers are already familiar with. The point he wants to make is that customers will always be able to access all their data in Microsoft Fabric. “We are the custodians of their data and want to make sure that it is completely open,” he summarizes. It is also worth noting that Microsoft Fabric runs in organizations’ own tenants, not somewhere else.
The above openness is, of course, a good thing. However, it does not fully answer our question. Organizations will currently want to know whether they can actually always access their data, especially given recent developments in this area. Data may be stored in open formats and be fully accessible via existing APIs, but if Microsoft decides (or is forced to decide) to block access for any reason, an organization will still not be able to access it. This is not unique to Microsoft Fabric, but applies to all services that run outside an organization.
Will Microsoft Fabric become the Office for data?
Microsoft Fabric is part of Azure, into which Microsoft has invested a great deal of time, money, and effort. In 18 months, a lot of progress has been made, that much is clear. We estimate that it is still far from complete. Microsoft will want to quickly expand Fabric to include even more functionality. It will only become more important and better. There is undoubtedly enormous potential in Microsoft Fabric.
The goal of Microsoft with Fabric is best visualized by something Ulag said during our conversation. “Thirty years ago, we had Word, which had to compete with Word Perfect, and Excel, which competed with Lotus. Then we realized that the real opportunities didn’t lie in those individual products, but in helping customers with their productivity. That’s why we created Office. It’s not just a bundle of different components, but products that work together and offer a single, uniform experience. We’re now doing exactly the same thing with Fabric.”
So Microsoft is following a familiar path with Fabric. More integration to make the platform as a whole better, more important, and therefore more ‘sticky’. Customers will eventually use it more and more, and Microsoft will thrive. There’s nothing wrong with that, in our opinion, but it’s something organizations need to be well aware of before they get started. Because once you’re in, experience shows that it’s difficult to get out again.
Availability
The native integration of Cosmos DB into Microsoft Fabric is now available in public preview. The agent for Power BI is available now, while the one for Teams will arrive within a week and the agent in Microsoft 365 Copilot in a few months. We are not entirely sure when the Digital Twin functionality will be available, but we suspect it will be immediately after the announcement. If that is not the case, we will update this information later.
Finally, it’s worth pointing out the fairly generous free trial version of Microsoft Fabric. Developers can sign up for this without needing a credit card or Azure subscription, according to Ulag. They will then receive $17,000 worth of Azure capacity to spend over two months. According to him, Microsoft is doing this so that developers can actually build something that showcases the power of Fabric.
Also read: Microsoft Fabric intends to be a data platform for the AI era