Azure Data Factory Pricing UK 2026: DIUs, Data Flow and Hidden Costs

Azure Data Factory's billing is deliberately opaque. Microsoft's own engineering blog admitted the charges "lack the clarity and transparency that we thrive to provide." The core confusion is the DIU — the Data Integration Unit — a compute abstraction that scales with data volume and parallelism in ways that aren't immediately obvious from the Azure portal.

The "Auto" DIU setting defaults to 4 DIUs minimum. This means that even a simple copy of 1 MB of data incurs the same minimum cost as copying 4× as much data. Engineers deploying dev/test pipelines with the default settings can easily accumulate charges they didn't expect.

What DIUs Actually Are

A Data Integration Unit (DIU) is a combination of CPU, memory, and network resources allocated to your Copy Activity. It's not a fixed size — it's a unit of parallelism and compute that Azure assigns based on the DIU count you specify.

The "Auto" default is 4 DIUs minimum

When you set DIU count to "Auto" (the default), Azure selects between 2 and 256 DIUs based on data size and partitioning. The key detail: the minimum is always 4 DIUs, regardless of how small the data transfer is. You cannot use 1 or 2 DIUs with Auto mode.

At UK South rates, Copy Activity costs approximately £0.1881/DIU-hour(sourced from the Azure Retail Prices API). This means:

• A 10-minute pipeline run at 4 DIUs = 4 × (10/60) × £0.1881 = £0.125 per run
• The same pipeline at 2 DIUs = 2 × (10/60) × £0.1881 = £0.063 per run
• At 100 runs/month: 4 DIUs costs £12.54/month, 2 DIUs costs £6.27/month

How DIU-hours are calculated

DIU-hours = number of runs × (duration in hours) × DIU count. This is billed at the end of the month. If your pipeline runs 500 times per month, for 5 minutes each, at 4 DIUs:

500 × (5/60) × 4 × £0.1881 = £31.35/month

Copy Activity vs Data Flow: When to Use Each

Copy Activity

Copy Activity is optimised for data movement — extracting from a source, optionally transforming lightly, and loading to a destination. It uses DIU-hours and is the cheapest option for straightforward ELT workloads.

Data Flow

Data Flow runs Spark-based transformations on a managed cluster. It's appropriate when you need complex transformations (joins, aggregations, pivots) that would be awkward in Copy Activity mapping. The cost model is fundamentally different:

• General Purpose cluster: approximately £0.216/vCore-hour
• Memory Optimised cluster: approximately £0.270/vCore-hour
• Compute Optimised cluster: approximately £0.172/vCore-hour

The minimum cluster size is 8 vCores. A General Purpose cluster running for 10 hours/month: 8 × 10 × £0.216 = £17.28/month. This is the minimum possible Data Flow charge if you run any jobs.

The cluster startup trap

Data Flow clusters take 1–4 minutes to start. Every time a pipeline triggers a Data Flow activity after the cluster has shut down, you pay for that startup time. The Time-To-Live (TTL) setting keeps the cluster warm between activities — typically worth setting to 10 minutes if you trigger Data Flows more than once per hour.

Self-hosted IR: The Hidden Hourly Charge

A Self-hosted Integration Runtime (IR) runs on your own infrastructure — typically a Windows VM — and enables ADF to connect to on-premises data sources or private networks. It's charged at approximately £0.0752/hour per license, separate from the VM compute costs you're already paying.

This is billed as long as the IR is registered and active — not just when it's processing data. An always-on Self-hosted IR costs approximately £55.9/month (£0.0752 × 744) in licensing fees, on top of whatever VM you're running it on.

Pipeline Orchestration: Free Tier and Beyond

Azure Data Factory includes 1,000 free pipeline activity runs per monthfor Azure-hosted Integration Runtime. This covers trigger runs, pipeline runs, and individual activity executions within a pipeline.

Above the free tier, activity runs are charged per 1,000 runs. At UK South rates:

• 0–1,000 runs/month: Free
• 1,001–100,000 runs/month: approximately £0.45 per 1,000 runs
• 100,000+ runs/month: approximately £0.36 per 1,000 runs

A pipeline that triggers 100 activities per run and runs 50 times per day generates 150,000 activity executions per month — 149,000 of which are paid. At the tiered rates, that's approximately £63/month in orchestration costs.

Finding Your Actual ADF Costs in Azure Cost Management

In Azure Cost Management, filter by:

• Service name: Azure Data Factory v2
• Meter: Select individual meters (Cloud Data Movement, On Premises Data Movement, etc.)

The breakdown you'll see:

• Cloud Data Movement — Copy Activity DIU-hours (Azure-hosted IR)
• On Premises Data Movement — Self-hosted IR hourly licensing
• Cloud Low Frequency Activity — Pipeline activity run charges
• Cloud High Frequency Activity — Trigger run charges

Note that Data Flow charges appear under a separate resource within the same ADF instance. If you're using Data Flow, look for vCore Hours meters in the Cost Management breakdown.

Estimate Your Data Factory Costs

Built and verified by an independent Azure engineer. See methodology →