Logo - Keyrus
  • Playbook
  • Services
    Data advisory & consulting
    Data & analytics solutions
    Artificial Intelligence (AI)
    Enterprise Performance Management (EPM)
    Digital & multi-experience
  • Insights
  • Partners
  • Careers
  • About us
    What sets us apart
    Company purpose
    Innovation & Technologies
    Committed Keyrus
    Regulatory compliance
    Investors
    Management team
    Brands
    Locations
  • Contact UsJoin us
Expert opinion

13

Microsoft Fabric Explained: Capabilities, Pricing, and Real-World Use Cases

Ryan Koch, Managing Consultant, Keyrus NORAM

Microsoft products are synonymous with business tech. According to a report by Land Base, 3.9 million organizations are using Microsoft products, with the majority based in the U.S. So, when Microsoft Fabric was released in 2023, there was, of course, excitement, but also some questions. Mainly, what is Microsoft Fabric, and where does this fit into the current Microsoft ecosystem?

In this article, we’ll break down what Microsoft Fabric is, its core capabilities, the strengths and limitations of the platform, and we’ll cover a real-world use case so you can see how your organization might be able to use it.

What is Microsoft Fabric?

Microsoft Fabric is a SaaS data and analytics product that combines numerous Microsoft products into one platform. Microsoft has products such as Azure Data Factory, Azure Synapse, and storage accounts. With Fabric, these platforms have been combined. Microsoft Fabric, or “Fabric”, consolidates data integration, engineering, warehousing, data science, real-time analytics, databases, and Power BI on one shared data lake, OneLake, under a single SKU. The goal of this was to make it easier for organizations that were managing numerous Microsoft products to have it billed under one line item.

  • TL;DR: Microsoft Fabric is like Microsoft Office, but for data. The way that PowerPoint, Word, and Excel all fall under Microsoft Office is how multiple data platforms fall under Microsoft Fabric.

Fabric is meant for a variety of users and business functions. It’s not only meant for data professionals, AI professionals, and analysts, but also IT teams, finance departments, and business users. The goal of having multiple products under one umbrella is to make it easier for multiple teams to collaborate.

These products are still offered on Azure, so using the phrasing “in the past” isn’t right here. Better to just say that Azure offers these services and, to simplify things for customers, Microsoft consolidated the services under a single SKU and shared data lake

Core Capabilities

As we mentioned, Fabric combines multiple platforms under one umbrella that span data engineering, data warehousing, visualization/analytics, and AI. But within Fabric, they are referred to as “workloads”. These are all built on and reference the same OneLake data lake, using various data ingestion options. The purpose is that all the different workloads in Fabric reference the same underlying storage accounts or data and to help reduce the risk of discrepancies or inaccuracies.

  • As with all data platforms and solutions, the output is only as good as the input. If your data is incomplete, disjointed, and ungoverned, the work done with that data will not be reliable.

  • If you’re handling sensitive data in your OneLake, it’s crucial to protect Personally Identifiable Information (PII) such as names, emails, and healthcare identifiers with dynamic data obfuscation.

The main workloads within Microsoft Fabric are Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence, Databases, Power BI, and Fabric IQ.

Workload

What it does

Typical users

Data Factory

170+ connectors, pipelines, Dataflows Gen2 (Power Query), copy jobs, mirroring

Data engineers, analysts

Data Engineering

Lakehouses, Spark notebooks and jobs (Python, Scala, SQL, R)

Data engineers

Data Warehouse

Serverless T-SQL warehouse over Delta tables, cross-database queries

SQL developers, BI teams

Data Science

ML experiments, models, MLflow, integration with Azure AI Foundry

Data scientists

Real-Time Intelligence

Eventstreams, Eventhouse (KQL), real-time dashboards, Activator alerts, Maps

Ops, IoT, fraud teams

Databases

SQL database in Fabric, Cosmos DB in Fabric for operational apps

App developers

Power BI

Semantic models, reports, Direct Lake mode, apps

Analysts, business users

Fabric IQ

Ontology (business entities and relationships), data agents, planning

Business and AI teams

As with all modern data platforms, there’s also heavy integration with AI in Microsoft Fabric. Often, we see Fabric Data Agents used to have AI review an existing data lakehouse. They can be configured to query your data and integrated into Copilot Studio and AI Foundry agents.

How Everything Works Together

As we mentioned, everything with Microsoft Fabric lives and runs on OneLake. From a customer perspective, as you set up a Fabric capacity in your Azure environment, OneLake will get set up at the same time. As you start creating workloads, all the data is stored in one central location. As your team and workloads grow, everyone is referencing the same data infrastructure. For the compute, everything draws on capacity units that are all tied to the Fabric SKU.

If you think of the structure from a layer perspective, it goes, bottom to top: OneLake (storage) → shared CU capacity (compute) → workloads (Factory, Engineering, Warehouse, Science, Real-Time, Databases, Power BI) → Fabric IQ semantic layer → Copilot, agents and Microsoft 365 apps, with Purview and OneLake security spanning all of them.

  1. OneLake (aka the OneDrive for data): Every Fabric tenant gets a single logical data lake built on Azure Data Lake Storage Gen2. Each workspace and item stores its data there automatically, leaving nothing to provision.

  2. Open formats: Tables are stored as Delta Lake on Parquet. Spark, T-SQL, KQL, and Power BI (via Direct Lake) all read the same (no copy per engine). OneLake also exposes ADLS-compatible endpoints and an Iceberg REST Catalog API, so tools such as Databricks and Snowflake can read the data too.

  3. Zero-ETL access to data elsewhere: Shortcuts are pointers to data in ADLS, Amazon S3, Google Cloud Storage, Dataverse or on-premises sources, with no data moved and no storage charged. Mirroring keeps a near-real-time, read-only replica of operational databases (Azure SQL, Snowflake, Oracle, SAP Datasphere, Google BigQuery, Cosmos DB and more) in OneLake using the source's change data capture.

  4. Capacity-based compute: All workloads draw on a shared pool of Capacity Units (CUs) bought as an F SKU. Fabric uses bursting (borrow more compute for short jobs) and smoothing (spread the cost over time) so bursts don't need a bigger SKU. Surge protection, overage billing, and the Capacity Metrics app help keep usage under control.

  5. SaaS on Azure: Microsoft runs the infrastructure, upgrades and scaling. Identity is Microsoft Entra ID; network controls include Private Link, workspace IP firewall rules, outbound access protection and customer-managed keys (all GA). Governance plugs into Microsoft Purview.

Pricing Overview

Fabric is priced as compute capacity plus storage: you buy a capacity (F SKU) measured in Capacity Units, every workload shares it, and OneLake storage is billed separately per GB. It’s important to remember that Fabric’s pricing is not consumption-based.

Compute: Sizes run from F2 (2 CUs) to F8192, each double the last. You can read more detailed and current information on the Microsoft pricing page.

For storage, OneLake hot storage runs about $0.023 per GB per month, with cool and cold tiers and a BCDR option available; shortcuts consume no storage, and mirroring includes free replica storage of 1 TB per CU (so an F64 gets 64 TB). On licensing, anyone publishing or viewing Power BI content on a capacity below F64 needs a Power BI Pro (or Premium Per User) license, while at F64 and above viewers can use a free license, and non-Power BI work such as pipelines, notebooks, and warehouses requires no Pro license at all. Other cost levers include Autoscale Billing for Spark (paying only for active Spark compute at PAYG rates), overage billing instead of throttling, surge protection, a dedicated Copilot capacity, and a 60-day free trial capacity.

Strengths and Advantages

Unified Billing Structure: The biggest advantage of Microsoft Fabric is the unified billing structure. Everything is billed under a single SKU, simplifying finance team efforts.

Always Connected and Integrated: Because everything is built on a single OneLake, everything within Microsoft Fabric is connected and integrated with each other from the beginning. There’s no need to configure one tool with another.

Scalable Foundation from the Start: For organizations experiencing growth and adding multiple tools to their tech stack, there usually isn’t oversight around how all these tools grow and scale. Fabric has everything set up to be scalable from the start. As you grow, so does your data foundation. The deep integration with other Azure services makes Fabric a no-brainer for existing Microsoft customers looking to scale analytics.

AI-Ready: Copilot, data agents, and the Fabric IQ semantic layer ground AI on governed business data.

Considerations and Limitations

Newness: Even though Fabric was released in 2023, it’s still relatively new to the market. This has led to some functionality growing pains, where all the bugs haven’t been fully flushed out.

Functionality Gaps: Despite combining multiple tools on one platform, there are still gaps in functionality for each tool. Features that are available on a standalone Azure product may not be included in the Fabric version.

Indirect Cost Visibility: While billed under a single SKU, CU consumption per workload is less intuitive than per-service Azure billing.

Migration effort: Moving legacy warehouses and pipelines still needs assessment, code conversion and testing.

Capacity sizing takes skill: Shared CUs mean one heavy Spark job can throttle Power BI reports if capacities aren't planned and monitored.

Real-World Use Case

In fleet and transportation, questions about diagnostics, parts, and warranties come up every day. The answers are often spread across policies, warranty terms, and maintenance records in different systems. Fabric brings that data together in OneLake, while Fabric Data Agents let technicians ask questions from Teams or a Copilot Studio agent, and get answers grounded in their organization’s own data.

An example to bring that to life is fleet maintenance. A technician looking at a failed component can quickly confirm whether a warranty applies, capturing savings that might otherwise go unrealized. Maintenance teams can draw on past work order history across asset groups, learning from previous failures and corrective events to improve preventative maintenance routines, and asset uptime. This is all accomplished by the tech asking a question within Teams in Fabric, which utilizes Fabric Data Agents.

The data foundations are in place, the agent does the searching, and the technician stays in command of the decision. That is architected intelligence in practice.

Keyrus & Microsoft

Standing up a Fabric capacity is the easy part. Making OneLake a foundation the whole business can trust, where every pipeline, report, and agent draws on the same governed data, is the actual work. It is what turns a Fabric pilot into a platform that compounds value over time.

At Keyrus, we call that discipline being an Architect of Intelligence: designing the Operating System of the intelligent organization, where data and AI are not bolted onto a workflow but embedded into how the business already runs. We operationalize intelligence.

Our proprietary Human Orchestrated Model™ (HOM™) frames how we approach a Fabric build like the one in this article, across three layers:

  1. Intelligence Foundations. The OneLake architecture, governed lakehouses and warehouses, and semantic models that every workload and agent inherits. With dbt jobs now generally available in Fabric, our teams bring the same modeling practices we use across platforms directly into OneLake. A Fabric Data Agent is only as trustworthy as the data beneath it.

  2. Human in Command. Workspace security, Purview governance, PII protection, and review steps that keep people setting intent and approving outcomes while pipelines and agents execute.

  3. Performance Steering. Measuring whether Fabric shortened the path from question to answer, not just whether it went live, and right-sizing capacity so compute matches how the business actually uses it.

Technology amplifies. Culture and governance determine whether that amplification compounds into something durable or evaporates the first time a number doesn’t match.

As a Microsoft partner, Keyrus helps organizations move from a Fabric pilot to a governed, production-ready platform that teams across the business can trust, from the first report to the first agent. Learn more about our Microsoft partnership or contact us to start your Microsoft Fabric rollout.

Read More: Escaping the Data Pipeline Trap in Microsoft Fabric

Microsoft Fabric is an all-in-one, AI-powered cloud platform that unifies data engineering, warehousing, data science, real-time analytics, and business intelligence (Power BI) into a single SaaS solution. It streamlines data management by utilizing OneLake, a centralized data lake, to eliminate data silos.

Microsoft Fabric is built for anyone who works with data across an organization. That includes data engineers, data scientists, analysts, and AI professionals, as well as IT teams, finance departments, and business users. Because every team works from the same platform and the same data, Fabric makes cross-team collaboration simpler.

Azure Data Factory and Azure Synapse are standalone Azure services that need to be connected to other tools and billed separately. Microsoft Fabric consolidates those capabilities into one integrated platform on OneLake, so services are connected from the start and costs fall under a single capacity rather than multiple consumption-based bills.

Keyrus is proud to be a Microsoft funding, reselling, and delivery partner and to have worked on numerous Microsoft Fabric projects. We know that data is unquestionably a key to success for businesses. When used intelligently, it opens unique opportunities for facing present and future challenges. At Keyrus, we enable organizations to deploy the capabilities to make data matter: by leveraging data and AI to start making smarter, more impactful decisions.

Microsoft Fabric includes eight main workloads: Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence, Databases, Power BI, and Fabric IQ. Together they cover data ingestion, Spark-based engineering, T-SQL warehousing, machine learning, streaming analytics, operational databases, reporting, and a semantic layer that grounds AI agents in business data.

OneLake is the single, organization-wide data lake that every Microsoft Fabric workload uses, often described as the OneDrive for data. It stores tables in open Delta Lake and Parquet formats, so Spark, T-SQL, KQL, and Power BI read the same data without copies. Shortcuts and mirroring bring in external data with zero ETL.

Microsoft Fabric is priced as compute capacity plus storage. Organizations buy an F SKU, ranging from F2 to F8192, which all workloads share, and pay for OneLake storage separately per GB.Current rates are on Microsoft’s pricing page.

Microsoft Fabric supports AI through Copilot, Fabric Data Agents, and the Fabric IQ semantic layer. Fabric Data Agents can be set up in a few steps to query data in OneLake and integrate with Copilot Studio and Azure AI Foundry agents, letting users ask questions in plain language from tools like Microsoft Teams.

It depends on capacity size. On capacities below F64, anyone publishing or viewing Power BI content needs a Power BI Pro or Premium Per User license. At F64 and above, report viewers can use a free license. Non-Power BI work in Fabric, such as pipelines, notebooks, and warehouses, does not require a Pro license.

Key strengths of Microsoft Fabric include unified billing, built-in integration across workloads, a scalable foundation, and AI readiness. Considerations include its relative newness, some feature gaps compared with standalone Azure services, less intuitive cost visibility per workload, migration effort for legacy systems, and the need to plan and monitor shared capacity carefully.

Start small with a pilot capacity, typically F2 to F16, and use the Capacity Metrics app to measure real usage before reserving a steady baseline. A free 60-day trial capacity is also available. Partners like Keyrus help organizations move from a Fabric pilot to a governed, production-ready platform.

Yes, for organizations already in the Microsoft ecosystem. Fabric supports single-tenant environments with isolated site-specific workspaces, letting regional governance standards sit centrally while individual sites manage their own data and reporting.

1. Exploding Costs: Inconsistencies in pipelines lead to financial issues. Inefficient code burns through Fabric Capacity Units (CUs) faster than necessary. This is the biggest challenge and “cost” of falling into the Trap. 2. Team Bloat: You find yourself needing a larger team just to handle the bugs, maintenance, and clean up technical debt, rather than delivering new insights. 3. Significant Manual Maintenance: If a new auditing standard or error logging requirement is introduced, you must manually update dozens, or even hundreds, of separate pipelines. 4. Longer Training for New Hires: New hires struggle to onboard because every project follows a different structure, requiring weeks of training just to understand the local "flavor" of engineering.

Eventstreams in Microsoft Fabric provide a real-time, event-driven data ingestion experience. It is Ideal for streaming scenarios such as IoT telemetry, clickstream analytics, and log ingestion.

Notebooks in Microsoft Fabric offer a flexible code-first approach to data ingestion and transformation. It is Ideal for data engineers and advanced analysts who prefer Python, Spark, or SQL for full control. It supports both batch and interactive processing within the Fabric ecosystem & it can be fully integrated with Lakehouse, Warehouse, and OneLake for seamless data operations.

Dataflow Gen2 is a revamped version of Power BI Dataflows (Power Query in the cloud), now part of the Data Factory Experience in Microsoft Fabric which is designed for analysts to ingest, clean, and load data from cloud-based sources with no-code or low-code transformations and supports data output destinations, offering more flexibility than traditional Dataflows.

Data Pipeline is a tool within the Data Factory Experience in Microsoft Fabric which is designed for analysts and engineers to build and orchestrate ETL and data ingestion workflows. It is similar to Azure Data Factory / Synapse Pipelines, though not full feature parity yet.

There are 4 core data ingestion options in Microsoft Fabric: Data Pipeline, Dataflow Gen2, Notebook, and Eventstream. Data pipeline orchestrates ETL/ELT workflows. Dataflow Gen 2 is a low-code/no-code ETL in Power Query. Notebook is a cook-first ingestion using Python, Spark, or SQL. Eventstream is real-time streaming ingestion. There are other options like database mirroring which we’ve excluded.

Continue reading
  • Blog post

    Keyrus introduces a new AI-enabled delivery platform for business transformation

  • Expert opinion

    How to build data foundations for agentic AI in healthcare

  • Webinar

    How Holland Bloorview, Snowflake and Keyrus are putting AI-assisted medical coding into practice

  • Success story

    Enterprise AI Governance That Speeds Up Value: How Keyrus and Unframe Help Teams Stay in Control as They Scale

  • Expert opinion

    What Is a Utility Common Information Model (CIM)? Understanding Interoperability, Governed Data, and AI

Logo - Keyrus
New York City

252 West 37th st., Suite 1400 New York, NY 10018

Phone:+1 646 664 4872