AWS
Microsoft Azure
Google Cloud
Key takeaways
1
ADF charges separately for workflow coordination, execution resources and operations, so data volume alone cannot predict the monthly bill.
2
Reliable estimates need run counts, allocated resources and billable time, with connected-service costs added separately.
3
Data Flow reservations can reduce eligible compute costs, but savings depend on matching usage within each hour.
Azure Data Factory pricing starts to make sense when you follow a workflow from its trigger to its last task. Copying a file, checking a table and transforming records can activate different meters. The examples below connect those steps to monthly spending, then show which costs to control before considering reservations.
Pipelines run manually, on schedules or through events. The integration runtime supplies execution infrastructure. Azure IR is Microsoft-managed and serverless; self-hosted IR runs on your machine. Managed VNet is an Azure IR networking configuration.
These Azure Data Factory pricing tables use USD retail rates checked November 12, 2026, against Microsoft’s billing definitions and East US retail meters. Your agreement, currency and region can change prices.
Azure Data Factory pricing rates
ADF is a managed data integration service. A pipeline groups tasks called activities. It can copy data from storage into a database, transform it through Mapping Data Flows, or dispatch work to Databricks. Storage keeps the data; ADF coordinates and performs the configured work.Pipelines run manually, on schedules or through events. The integration runtime supplies execution infrastructure. Azure IR is Microsoft-managed and serverless; self-hosted IR runs on your machine. Managed VNet is an Azure IR networking configuration.
These Azure Data Factory pricing tables use USD retail rates checked November 12, 2026, against Microsoft’s billing definitions and East US retail meters. Your agreement, currency and region can change prices.
| Meter | Azure IR | Managed VNet | Self-hosted |
|---|---|---|---|
| Orchestration per 1,000 runs | $1 | $1 | $1.50 |
| Copy execution | $0.25/DIU-hour | $0.25/DIU-hour | $0.10/activity-hour |
| Pipeline activity execution | $0.005/hour | $1/hour | $0.002/hour |
| External activity execution | $0.00025/hour | $1/hour | $0.0001/hour |
Orchestration includes activity runs, trigger executions and debug runs. Pipeline activities include Lookup and Get Metadata; external activities dispatch work to services such as Databricks. Mapping Data Flow compute has its own rates below.
Managed VNet execution prices cover up to 50 concurrent pipeline activities or 800 external activities per node, respectively. Do not multiply the hourly rate by every overlapping activity.
Managed VNet compute capacity and time to live (TTL) affect billing. Pipeline and external activity TTL cannot be disabled; retained compute can remain billable between tasks. This is separate from long-term Data Flow reservations.
A DIU-hour measures allocated copying resources over time. Four DIUs for 15 minutes consume one DIU-hour. Mapping Data Flows use managed Spark clusters. A vCore-hour measures one virtual core for one hour of Data Flow compute.
Managed VNet execution prices cover up to 50 concurrent pipeline activities or 800 external activities per node, respectively. Do not multiply the hourly rate by every overlapping activity.
Managed VNet compute capacity and time to live (TTL) affect billing. Pipeline and external activity TTL cannot be disabled; retained compute can remain billable between tasks. This is separate from long-term Data Flow reservations.
A DIU-hour measures allocated copying resources over time. Four DIUs for 15 minutes consume one DIU-hour. Mapping Data Flows use managed Spark clusters. A vCore-hour measures one virtual core for one hour of Data Flow compute.
| Region | General Purpose per vCore-hour | Memory Optimized per vCore-hour |
|---|---|---|
| East US | $0.27325 | $0.34275 |
| West Europe | $0.26800 | $0.34475 |
| Central India | $0.28425 | $0.34475 |
Ordinary Mapping Data Flow execution starts at eight vCores. IR execution and Data Flow compute are prorated by the minute and rounded up. Two minutes and twenty seconds therefore bills three minutes.
ADF operations add $0.50 per 50,000 modified/referenced entities for read/write and $0.25 per 50,000 retrieved run records for monitoring. These measure management work, not business-data rows copied.
Azure-SSIS pricing separately covers existing SSIS jobs, with configuration and licensing affecting runtime costs. This guide covers standalone ADF V2 pipelines; Fabric uses a different billing model. Our Azure pricing guide explains the broader billing context.
Microsoft’s current free-account offer advertises five low-frequency Data Factory activities as always free. The offer does not specify which V2 billing meters it covers, so do not interpret it as five entirely free V2 pipelines. Confirm your subscription’s allowance before deducting it from an estimate.
Similarly, do not apply Synapse’s first million free monthly pipeline operations to standalone ADF V2.
Trial credits can cover experimentation, but production estimates should show consumption before credits. A workflow can also create charges in connected storage or databases even when a credit offsets its ADF charges.
ADF operations add $0.50 per 50,000 modified/referenced entities for read/write and $0.25 per 50,000 retrieved run records for monitoring. These measure management work, not business-data rows copied.
Azure-SSIS pricing separately covers existing SSIS jobs, with configuration and licensing affecting runtime costs. This guide covers standalone ADF V2 pipelines; Fabric uses a different billing model. Our Azure pricing guide explains the broader billing context.
Is Azure Data Factory free
Assume production usage is billable. Eligible new customers can use Azure’s $200 credit for 30 days, shared across qualifying services.Microsoft’s current free-account offer advertises five low-frequency Data Factory activities as always free. The offer does not specify which V2 billing meters it covers, so do not interpret it as five entirely free V2 pipelines. Confirm your subscription’s allowance before deducting it from an estimate.
Similarly, do not apply Synapse’s first million free monthly pipeline operations to standalone ADF V2.
Trial credits can cover experimentation, but production estimates should show consumption before credits. A workflow can also create charges in connected storage or databases even when a credit offsets its ADF charges.
Azure Data Factory pricing examples
For each ADF meter, multiply consumption by the rate for its runtime and billing unit, then add connected-service and self-hosted infrastructure costs. For example, divide orchestration runs by 1,000 before multiplying by the listed rate. Use billable resource-time rather than total workflow duration.
These Azure Data Factory pricing examples are modeled execution subtotals, not customer invoices. Each assumes 30 days, one scheduled trigger execution per workflow, successful runs, and no retries, extra activities, debugging or retained Data Flow compute. Durations are assumed billable time, not measured throughput.
A daily cloud copy
One Azure IR Copy activity runs daily, using four DIUs for ten minutes.| Component | Calculation | Monthly cost |
|---|---|---|
| Orchestration | 30 × (one trigger + one Copy) ÷ 1,000 × $1 | $0.06 |
| Copy compute | 30 × 4 DIUs × 10/60 hours × $0.25 | $5.00 |
| Execution subtotal | $5.06 |
Operations, source/destination services and applicable networking are additional.
Hourly copying and transformations
A workflow runs 720 times. Each execution copies data using eight DIUs for ten minutes, then transforms it using eight General Purpose vCores for fifteen minutes.| Component | Calculation | Monthly cost |
|---|---|---|
| Orchestration | 720 × (trigger + Copy + Data Flow) ÷ 1,000 × $1 | $2.16 |
| Copy compute | 720 × 8 × 10/60 × $0.25 | $240.00 |
| Data Flow compute | 720 × 8 × 15/60 × $0.27325 | $393.48 |
| Execution subtotal | $635.64 |
Operations, managed Data Flow disk/blob storage, connected services and networking remain additional. Transformations dominate this example.
A daily hybrid copy
One self-hosted Copy activity runs daily for one billable hour.| Component | Calculation | Monthly cost |
|---|---|---|
| Orchestration | 30 × (one trigger + one Copy) ÷ 1,000 × $1.50 | $0.09 |
| Copy execution | 30 × 1 hour × $0.10 | $3.00 |
| Execution subtotal | $3.09 |
The host machine, maintenance, operations, storage and networking are excluded. This subtotal does not establish that hybrid execution is cheaper; the configurations are not performance-equivalent.
What adds to your bill
An Azure Data Factory pricing estimate should include costs outside ADF execution:Storage and transactions at the source, destination and staging locations.
Processing services invoked by ADF. Our Azure Databricks pricing guide separates those compute charges.
Networking and infrastructure, including applicable transfer, private connectivity and self-hosted machines.
Development resources, including active debug sessions and managed Data Flow disk/blob storage.
Optional exported logs can incur Azure Monitor charges, separate from ADF’s monitoring-operation meter.
Microsoft’s inactive-pipeline FAQ defines inactivity as no associated trigger and zero runs within a month; charging begins after one month of zero runs. The East US retail inactive-pipeline rate is $0.80/month.
Microsoft’s inactive-pipeline FAQ defines inactivity as no associated trigger and zero runs within a month; charging begins after one month of zero runs. The East US retail inactive-pipeline rate is $0.80/month.
How to reduce ADF costs
Understanding Azure Data Factory pricing helps you target unnecessary work without shrinking resources indiscriminately.| Action | Consumption affected | Condition to check |
|---|---|---|
| Remove unnecessary runs and retries | Orchestration and execution | Preserve required data freshness and recovery |
| Copy changed data where supported | Copy resource-hours | Validate change tracking and completeness |
| Benchmark resource allocations | DIU-hours or vCore-hours | Fewer resources can take longer |
| Close unused debug sessions | Data Flow compute | Separate development from production usage |
| Review TTL settings | Retained Data Flow compute | Weigh startup delays against warm capacity |
Unused debug time can add up. In an illustrative East US pay-as-you-go scenario, an eight-vCore General Purpose debug cluster remaining billable for two extra hours on each of 20 working days adds 8 × 2 × 20 × $0.27325 = $87.44 in compute, excluding storage and before discounts or credits. Count actual billable cluster time: debug sessions incur charges while running, including retained TTL time.
Microsoft’s runtime performance guidance explains why larger clusters can finish faster without necessarily costing less. Compare resources multiplied by billable duration for the same work.
Time to live, or TTL, keeps a Data Flow cluster available between jobs. It can reduce startup delays for sequential flows; parallel flows can require separate clusters. Avoid treating TTL as an automatic saving.
Use ADF’s consumption view to find expensive runs. Microsoft’s per-pipeline billing guidance explains the required opt-in and exclusions: operations, SSIS nodes and Data Flows on TTL-configured runtimes can remain at factory level.
Consumption reports show quantities; actual charges depend on your billing terms. Budget alerts help detect overspending but do not themselves stop execution.
Unused capacity in an hour cannot roll forward under the hourly reservation rules.
Consider an illustrative eight-vCore, three-year General Purpose reservation in the East US. The reservation meter lists $4,668 per vCore for the term. Normalizing across three 365-day years gives $4,668 ÷ 26,280 = approximately $0.177626/vCore-hour.
Microsoft’s runtime performance guidance explains why larger clusters can finish faster without necessarily costing less. Compare resources multiplied by billable duration for the same work.
Time to live, or TTL, keeps a Data Flow cluster available between jobs. It can reduce startup delays for sequential flows; parallel flows can require separate clusters. Avoid treating TTL as an automatic saving.
Use ADF’s consumption view to find expensive runs. Microsoft’s per-pipeline billing guidance explains the required opt-in and exclusions: operations, SSIS nodes and Data Flows on TTL-configured runtimes can remain at factory level.
Consumption reports show quantities; actual charges depend on your billing terms. Budget alerts help detect overspending but do not themselves stop execution.
When Data Flow reservations pay off
Reservations reduce the eligible compute portion of Azure Data Factory pricing. Microsoft’s reserved-capacity options offer one- or three-year terms, paid upfront or monthly. Region, compute type, quantity and scope determine matching usage.Unused capacity in an hour cannot roll forward under the hourly reservation rules.
Consider an illustrative eight-vCore, three-year General Purpose reservation in the East US. The reservation meter lists $4,668 per vCore for the term. Normalizing across three 365-day years gives $4,668 ÷ 26,280 = approximately $0.177626/vCore-hour.
| Thirty-day model | Concentrated usage | Continuous usage |
|---|---|---|
| Daily profile | 32 vCores × 6 hours | 8 vCores × 24 hours |
| Monthly vCore-hours | 5,760 | 5,760 |
| Reservation utilization | 25% | 100% |
| All-PAYG compute | $1,573.92 | $1,573.92 |
| Normalized reservation cost | $1,023.12 | $1,023.12 |
| Uncovered PAYG compute | $1,180.44 | $0 |
| Combined compute cost | $2,203.56 | $1,023.12 |
Reservation utilization is the share of purchased reservation vCore-hours used by matching workloads within their respective hours over this 30-day model.
Reservation cost is 8 × 720 × the normalized rate. Concentrated usage leaves 24 vCores uncovered for six hours daily: 24 × 6 × 30 × $0.27325 = $1,180.44.
No other matching workload consumes unused hours. Equal consumption does not imply equal performance. Figures allocate term cost to 720 hours, not monthly installments, and exclude other meters, services, fees and Cashback. Actual reservation dates govern the term.
The lesson is to reserve sustainable hourly demand, rather than infer savings from monthly averages.
Reservation cost is 8 × 720 × the normalized rate. Concentrated usage leaves 24 vCores uncovered for six hours daily: 24 × 6 × 30 × $0.27325 = $1,180.44.
No other matching workload consumes unused hours. Equal consumption does not imply equal performance. Figures allocate term cost to 720 hours, not monthly installments, and exclude other meters, services, fees and Cashback. Actual reservation dates govern the term.
The lesson is to reserve sustainable hourly demand, rather than infer savings from monthly averages.
Build your ADF budget
Build your Azure Data Factory pricing estimate from a representative schedule, then replace assumptions with consumption evidence.
1
Choose region and runtime. Use the execution region and the correct Azure, managed VNet or self-hosted meters.
2
Count executions. Include activities inside loops, trigger executions, expected retries and development runs. Pipeline count alone is insufficient.
3
Collect billable resource-time. Record copying DIUs and duration, Data Flow cores and duration, and other execution hours. Use billable quantities rather than elapsed workflow time.
4
Add surrounding costs. Include operation volumes, storage, databases, networking and host infrastructure. Keep one-time testing separate from recurring production.
5
Calculate and reconcile. Enter these inputs in the Azure Pricing Calculator. Microsoft's sample-run estimation method recommends representative datasets before projecting the full schedule.
For the daily cloud-copy example, use East US, USD and Azure IR with these monthly execution quantities:
| Input | Monthly quantity |
|---|---|
| Orchestration runs | 60, including 30 triggers and 30 Copy activity runs |
| Copy consumption | 20 DIU-hours |
| Data Flow consumption | 0 vCore-hours |
Compare estimates with Cost Management charges over the same period. Investigate which meter changed before changing resources or buying more commitments. A useful budget explains both its total and the operational assumptions behind it.
With our Flex Insured Commitments, teams can get up to 35% savings on eligible Data Flow compute through three-year reserved capacity with none of the commitment risk.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide Cashback protection for the qualifying difference. Our protection does not cancel the underlying commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.
We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.
Once you have identified your eligible Data Flow baseline, our read-only Savings Test can assess existing coverage and additional commitment opportunities.
How Usage.ai manages Data Flow commitments
At Usage.ai, we help you review Data Flow reservation coverage as demand changes. Through our Flex Insured Commitment Program, we purchase and manage supported commitments on your behalf, including Data Factory Data Flows. Your pipelines and Azure resources remain unchanged.With our Flex Insured Commitments, teams can get up to 35% savings on eligible Data Flow compute through three-year reserved capacity with none of the commitment risk.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide Cashback protection for the qualifying difference. Our protection does not cancel the underlying commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.
We charge an agreed percentage of realized savings, so the fee applies when verified savings are generated.
Once you have identified your eligible Data Flow baseline, our read-only Savings Test can assess existing coverage and additional commitment opportunities.
Frequently asked questions
Is Azure Data Factory free ?
ADF is not entirely free. Microsoft advertises five low-frequency activities, and qualifying new customers can use trial credits. These offers do not make every V2 pipeline charge free; budget other consumption and connected services separately.
How much does ADF cost monthly ?
Azure Data Factory pricing combines orchestration, execution, Data Flow compute and operations. Our daily cloud-copy model is $5.06/month in execution charges; storage, networking and other excluded costs increase the overall total.
Does ADF charge per terabyte ?
Azure Data Factory pricing is not a flat charge per terabyte. Copy costs depend on runtime resources and billable duration; orchestration and transformations add their own consumption.
Does idle ADF still cost money ?
It can. Inactive pipelines meeting Microsoft's stated conditions incur charges. Debug compute, retained resources, self-hosted infrastructure and reservation obligations can also continue costing money without production runs.
When do Data Flow reservations save ?
When matching eligible hourly usage makes the reservation cheaper than equivalent PAYG consumption. Unused hours cannot roll forward, so a monthly average can conceal poor utilization.
Share
Facebook
X
LinkedIn
Reddit
