(Updated) Updates to Consumption Measurement for AI Capabilities in Microsoft Fabric

Updated October 6, 2026: We have updated the timeline. Thank you for your patience. 

[What and why:]

You’re receiving this notification because you're associated with one or more Azure tenants that use AI capabilities in Microsoft Fabric.  

Starting October 1, 2026, Microsoft will update how consumption is measured for select Microsoft Fabric AI experiences. 

This update aligns capacity consumption more closely with the resources required to complete AI-powered workloads and introduces the flexibility to change consumption rates based on model size.  

What is changing? 

Today, most AI capabilities in Fabric are charged based on a fixed consumption model regardless of the models, tools, and processing involved in completing a request. 

Starting on October 1, 2026, select Fabric AI experiences will use a dynamic consumption model that measures the resources required to complete each task. Consumption will continue to be measured and billed through Fabric Capacity Units (CUs). 

The amount of capacity consumed will vary based on the following factors: 

• The AI model used and reasoning effort 
• The complexity of the request 
• The tools and services involved 
• The processing required to complete the task 

What capabilities in Fabric are impacted by this change? 

The updated consumption model will apply to some Microsoft Fabric operations that utilize large language models (LLMs). These include Copilot in Fabric, Fabric data agents, AI Functions, and Fabric IQ ontologies.  

Additional AI capabilities may adopt the updated consumption model over time as new features become available. 

[Rollout schedule:]

Worldwide: Beginning in early October 2026 and expected to complete by late December 2026 

Beginning in October 2026: Ontology AI experiences are the first to adopt the updated model. 

Later this year: Copilot in Fabric, data agents, operations agents, AI Functions, Real-Time Intelligence, and additional AI experiences will move to the updated model. We'll confirm timing in a follow-up notification well ahead of the change.

[Impact on your organization:]

Your existing Fabric workloads and AI operations will continue to operate without interruption. 

After October 1, you may notice changes in Capacity Unit (CU) consumption for AI-powered workloads. Consumption patterns may vary based on the workload being performed and the resources required to complete individual tasks. 

Review the Fabric Capacity Metrics App to understand and monitor the CU consumption changes for these operations.  

[Action required / Recommendations:]

To prepare for this update, we recommend that Fabric capacity administrators: 

1. Review current usage of Fabric AI workloads. 
2. Monitor Capacity Unit consumption after the update takes effect. 
3. Evaluate capacity utilization trends and adjust capacity planning as needed. 
4. Share this information with teams using Fabric AI capabilities.