Microsoft Fabric Updates Blog

Understanding Fabric Eventstream Pricing

In this blog post, we’ll walk through Eventstream’s pricing model to give you a clear understanding of how it works and help you navigate it with confidence.

By the end of this post, you will be able to:

  • Comprehend how Eventstream pricing is structured across its components.
  • Understand the relationship between Eventstream components and billing meters.
  • Review detailed pricing examples to support precise and confident cost estimation.

Eventstream Components & Cost Drivers

First, let’s summarize the components of a Fabric Eventstream:

  1. Input sources
  2. Streams (default and derived)
  3. Operators (using No-Code or SQL)
  4. Destinations

Each component maps to one or more of the four Eventstream operation types (billing meters). These operation types define how Eventstream activity is represented in the Fabric Capacity Metrics app and how they contribute to Fabric consumption.

The following reference table outlines how Eventstream components correspond to each operation type. Note that certain components may map to one or more operation types:

Eventstream ComponentOperation TypeDescription of Operation TypeOperation unit of measureFabric consumption rate
Eventstream Per HourFlat chargePer hour0.222 capacity unit (CU) hours
Streams (default and derived)Eventstream Data Traffic Per GBData ingress and egress volume in default and derived streams
(includes 24-hour retention)
Per gigabyte0.342 CU hours
Operators and some input sourcesEventstream Processor Per HourComputing resources that the processor consumesPer hourStarts at 0.778 CU hours and autoscales 2 per throughput
Input sources leveraging connectorEventstream Connectors Per vCore HourComputing resources that the connectors consumePer hour0.611 CU hours per vCore

Table listing Eventstream pricing components with their operation types, descriptions, units, and consumption rates, including hourly charges, data traffic per GB, processor usage, and connector vCore costs.

Breakdown of each component with associated operation type:

Flow diagram: Input Source → Default Stream → Operator → Derived Stream → Destination.

Flow diagram: Input Source → Default Stream → Operator → Derived Stream → Destination.

Input Source

Input source(s) represents where your data comes from in an Eventstream. You can connect to different types of sources, and pricing is based on how they operate.

Most Eventstream sources fall under ‘connectors’, meaning they incur the cost of Connector vCore per hour meter (e.g., Azure SQL DB, Cosmos DB, Service Bus, etc.).

Other sources do not incur the connector cost and instead incur the Eventstream Processor Per Hour meter charge (e.g. Azure Event Hubs, Azure IoT Hub).

Billing is determined by how each source brings data into Fabric.


Default and Derived Streams

There are two types of streams in an Eventstream topology. Default Stream is present in all Eventstreams, and additionally, you can create a Derived Stream

A derived stream refers to a logical stream of data which is created by applying transformations to the default stream after adding operators such as filtering and aggregations. Derived streams can be accessed independently from default streams for additional consumption or analysis through the Real-Time Hub.

For billing purposes, the Eventstream Data Traffic per GB meter is calculated based on the volume of data entering (ingress) and exiting (egress) each stream within your Eventstream topology, as illustrated below:

Flow diagram with two branches: Input Source → Default Stream → Transform Operator → Destination; and Input Source → Default Stream → Filter Operator → Derived Stream → Destination. Shows ingress/egress labels and data sizes (1 GB, 1 MB).
Flow diagram with two branches: Input Source → Default Stream → Transform Operator → Destination; and Input Source → Default Stream → Filter Operator → Derived Stream → Destination. Shows ingress/egress labels and data sizes (1 GB, 1 MB).

Operators

Operators (configured through no-code options or SQL) enable the application of processing logic within your Eventstream, including filtering, enrichment, and routing. Billing is determined by the number of processing routes defined in the Eventstream, representing the paths through which data flows from sources, through processing steps, to their respective outputs:

Flow diagram with three processing routes from Default Stream: Processing route 1: Filter → Manage Fields → Destination 1 (push), Processing route 2: Filter → Derived Stream, Processing route 3: Aggregate → Destination 3 (push).
Flow diagram with three processing routes from Default Stream: Processing route 1: Filter → Manage Fields → Destination 1 (push), Processing route 2: Filter → Derived Stream, Processing route 3: Aggregate → Destination 3 (push).

The Eventstream Processor Per Hour meter is determined by the number of processing routes defined within your Eventstream.

Separately, the throughput setting serves as a lever to manage how the system scales. Eventstream automatically adjusts compute resources to maintain low latency, and billing reflects the compute that is actually consumed. By configuring throughput, you establish an upper limit on system scaling, which can also support cost management.


Destinations

In Fabric Eventstreams, destinations fall into two categories:

  • Push destinations perform processing and send data, they are charged per the Eventstream Processor Per Hour meter, these include:
    • Lakehouse
    • Eventhouse (with transformation applied)
    • Activator
  • Pull destinations fetch data from your Eventstream (rather than Eventstream pushing the data), these destinations do not incur a processing cost on the Eventstream side, these include:
    • Eventhouse (Direct ingestion)
    • Custom endpoint

Pricing Walkthrough: Example Scenarios

We will now examine several examples to calculate CU consumption based on the Eventstream components involved.

Example 1: Custom endpoint to Eventhouse (direct ingestion, no processing)

Flow diagram: Source (Custom endpoint) → Default Stream → Destination (Eventhouse Direct Ingestion), with ingress and egress labeled as 1 GB per day.
Flow diagram: Source (Custom endpoint) → Default Stream → Destination (Eventhouse Direct Ingestion), with ingress and egress labeled as 1 GB per day.

 The set-up:       

  • Source: Custom endpoint
  • Destination: Eventhouse (pull: direct ingestion)
  • Data volume: 1 GB per day (or ≈ 0.04 GB per hour)
  • Processing: Passthrough
  • Eventstream uptime: Continuous, 24 hours/day

For this set-up, the meter calculation would be as follows:

Operation in Capacity Metrics AppDescriptionOperation unit of measureFabric consumption rateCalculation
Eventstream Per HourFlat chargePer hour0.222 CU hour0.22 per hour
Eventstream Data Traffic per GBData ingress & egress volume in default and derived streams
(Includes 24-hour retention)
Per GB0.342 CU hour0.08 GB (.04 GB ingress+.04 GB egress) of data X .342 = 0.03 per hour  
Eventstream Processor Per HourComputing resources consumed by the processorPer hourStarts at 0.778 CU hour and autoscale per throughput0 X .778 = 0 per hour  
Eventstream Connectors Per vCore HourComputing resources consumed by the connectorsPer hour0.611 CU hour per vCore0 X .611 = 0 per hour  
    SUM = 0.25 CU per hour  
Table showing Eventstream capacity metrics with operation types, descriptions, units, consumption rates, and example calculations for hourly charges, data traffic per GB, processor usage, and connector vCore costs, ending with a total of 0.25 CU per hour.

This Eventstream topology consumes 0.25 CU per hour.

Recommended SKU for this Eventstream alone: F2 (87% unused)


Example 2: Content based routing with 1 connector source, processing logic, and 2 destinations.

Flow diagram: Source (Connector) → Default Stream → two branches: Operator 1 (Filter) → Destination 1 (push), Operator 2 (Manage Fields) → Destination 2 (push). Includes ingress 1 GB/day and egress 1 GB per branch.
Flow diagram: Source (Connector) → Default Stream → two branches: Operator 1 (Filter) → Destination 1 (push), Operator 2 (Manage Fields) → Destination 2 (push). Includes ingress 1 GB/day and egress 1 GB per branch.

The set-up:        

  • Source: 1 connector source (for example Azure SQL CDC)
  • Destination: 2 destinations (Eventhouse tables with pre-processing)
  • Data volume: 1 GB per day (or ≈ 0.04 GB per hour)
  • Processing: Filtering and routing data
  • Eventstream uptime: Continuous, 24 hours/day

For this set-up, the meter calculation would be as follows:

Operation in Capacity Metrics AppDescriptionOperation unit of measureFabric consumption rateCalculation
Eventstream Per HourFlat chargePer hour0.222 CU hour0.22 per hour
Eventstream Data Traffic per GBData ingress & egress volume in default and derived streams
(Includes 24-hour retention)
Per GB0.342 CU hour0.12 GB (.04 GB ingress + (2 x .04 GB egress)) x .342 = 0.04 per hour
Eventstream Processor Per HourComputing resources consumed by the processorPer hourStarts at 0.778 CU hour and autoscale per throughput2 X .77 = 1.55 per hour  
Eventstream Connectors Per vCore HourComputing resources consumed by the connectorsPer hour0.611 CU hour per vCore1 X .611 = 0.61 per hour  
    SUM = 2.42 CU per hour  

Table showing Eventstream capacity metrics with operation types, descriptions, units, consumption rates, and example calculations. Includes hourly charges, data traffic per GB, processor costs, connector vCore costs, and a total of 2.42 CU per hour.

This Eventstream topology consumes 2.42 CU per hour.

Recommended SKU for this Eventstream alone: F4


Example 3: Aggregation processing with scaling traffic (comparing 1 GB/hour vs. 10 GB/hour).

This example shows how the same processing logic results in different processing charges based on the amount of data. As traffic increases, Eventstream scales compute, and the Eventstream Processor Per Hour cost grows accordingly.

The set-up (common across except for traffic levels):

  • Source: 1 connector source
  • Destination: Data Activator
  • Data volume: varies by scenario
    • Example 3A: 1 GB/hour
    • Example 3B: 10 GB/hour
  • Processing: 1 aggregation operator
  • Eventstream uptime: Continuous, 24 hours/day

Example 3A: 1 GB per hour

Operation in Capacity Metrics AppDescriptionOperation unit of measureFabric consumption rateCalculation
Eventstream Per HourFlat chargePer hour0.222 CU hour0.22 per hour
Eventstream Data Traffic per GBData ingress & egress volume in default and derived streams
(Includes 24-hour retention)
Per GB0.342 CU hour2 GB (1 in + 1 out) × 0.342 = 0.68 per hour
Eventstream Processor Per HourComputing resources consumed by the processorPer hourStarts at 0.778 CU hour and autoscale per throughput1 × 0.778 = 0.77 CU per hour
Eventstream Connectors Per vCore HourComputing resources consumed by the connectorsPer hour0.611 CU hour per vCore  1 × 0.611 = 0.61 per hour
    SUM = 2.28 CU per hour  
Table showing Eventstream capacity metrics with operation types, descriptions, units, consumption rates, and example calculations. Includes hourly Eventstream charge, data traffic per GB, processor cost, connector vCore cost, and a total of 2.28 CU per hour.

Recommended SKU for Example 3A: F4

Example 3B: 10 GB per hour

Operation in Capacity Metrics AppDescriptionOperation unit of measureFabric consumption rateCalculation
Eventstream Per HourFlat chargePer hour0.222 CU hour0.22 per hour
Eventstream Data Traffic per GBData ingress & egress volume in default and derived streams
(Includes 24-hour retention)
Per GB0.342 CU hour20 GB (10 in + 10 out) × 0.342 = 6.84 per hour
Eventstream Processor Per HourComputing resources consumed by the processorPer hourStarts at 0.778 CU hour and autoscale per throughput1 × (0.778 x 3*) = 2.33 CU per hour
*1 base rate (2.33 CU hours)
Eventstream Connectors Per vCore HourComputing resources consumed by the connectorsPer hour0.611 CU hour per vCore  1 × 0.611 = 0.61 per hour
    SUM = 10.00 CU per hour  

Table showing Eventstream capacity metrics with operation types, descriptions, units, consumption rates, and example calculations. Includes hourly Eventstream charge, data traffic for 20 GB per hour, processor autoscale cost, connector vCore cost, and a total of 10.00 CU per hour.

Recommended SKU for Example 3B: F16


 To Summarize:

The cost of operating Eventstream is determined by several factors:

  • Type and number of input sources
  • Type and number of destinations
  • Compute required for processing
  • Data volume (ingress and egress) from streams
  • Eventstream uptime

While this blog post helps calculate CU consumption, we recommend running your Eventstream in a realistic, steady-state configuration for a period of time—ideally 24 hours—to observe actual consumption patterns.

The Microsoft Fabric Capacity Estimator can also be used to assist in cost planning, as it estimates the capacity requirements of an Eventstream based on your input parameters. It is important to consider any additional Fabric workloads, such as Eventhouse, Lakehouse, or Notebooks, since all services draw from the same capacity pool.

We’d Love Your Feedback!

If you found this blog post helpful, please give it a thumbs up and share your feedback with us.

Questions? Reach out via email at askeventstreams@microsoft.com.

You can also submit feedback or feature requests on Fabric ideas and join the conversation with fellow users in the Fabric Community.


Пов’язані дописи в блозі

Understanding Fabric Eventstream Pricing

лютого 17, 2026 від Virginia Roman

We’re introducing billing reporting updates that make it easier to track AI-related usage in Microsoft Fabric. New AI Functions operation Until now, Fabric AI functions usage was reported under other operations, such as Spark-related operations, or Dataflows Gen2-related operations, depending on where the functions were used. To provide more transparency, Fabric AI functions will have … Continue reading “Billing updates: new operations for Fabric AI functions and AI services”

лютого 17, 2026 від Penny Zhou

Coauthor: Abhishek Narain Ensuring secure connectivity to your data sources is critical for modern data estates. We have released the Key-Pair authentication for Snowflake connections Preview in October and are happy to announce it is now Generally Available (GA). This release offers an enhanced security alternative to basic authentication methods, such as username and password, … Continue reading “Snowflake Key-Pair Authentication (Generally Available)”