Foundry Doc intelligence cost

Question

Azure AI Foundry is incurring a relatively high cost of approximately ~ USD 80 per day and ~2000$ only for foundry. Based on our analysis in Cost Management, Azure Document Intelligence appears to be the primary contributor to this cost. This is unexpected, as the total number of documents is fewer than 15, with each document being approximately 115 KB in size and only 2–3 pages long. Could you help us understand why the Foundry/Document Intelligence cost is so high for such a small document set?

Answer 1

Hi Pavan,

The main concern is that the cost is high even though you are processing only a small number of documents. This usually happens because billing in Document Intelligence is not based on file size, but on how the service is used.

Let me explain this in a simple way.

Even if your documents are small, the cost depends on how many pages were processed, which model or feature was used, and how many times the documents were analyzed.

For example, when you submit a document for analysis, the service processes all pages by default. If the same document is processed multiple times during testing, development, or retries, the cost can increase quickly even with a small dataset.

Here are a few things you can check step by step:

First, check how many pages were processed Go to your Document Intelligence resource in Azure portal Open Monitoring and then Metrics Look at the Processed Pages metric to see the actual usage

Second, check which feature or model is generating cost Split the metrics by feature name Some models or features may have higher pricing than others, so this will help you identify the source

Third, check cost breakdown in Cost Management In Cost Analysis, group by Resource and Meter This helps you confirm whether the cost is only from Document Intelligence or if other Foundry resources are also contributing

Fourth, check if documents were processed multiple times If you ran tests, pipelines, or retries, the same document might have been analyzed many times This is one of the most common reasons for unexpected cost increase

Fifth, check custom model usage Training itself is free, but document analysis and layout processing used for labeling can still generate cost.

Based on your scenario, a daily cost around this range usually indicates repeated processing or additional usage behind the scenes rather than actual document size.

For better understanding, you can also refer to these official documents

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/how-to-guides/estimate-cost

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/service-limits
https://learn.microsoft.com/en-us/azure/foundry/concepts/manage-costs

These will help you understand how pricing works and how to monitor and control cost.

If you can share details like which model you are using and how many times you are running analysis, it will be easier to identify the exact reason for the cost.

Thankyou!

Answer 2

Hi,

The cost is most likely not related to the file size of the documents, but to how many pages were analyzed, which Document Intelligence feature/model was used, and how many times the documents were processed.

Azure Document Intelligence billing is based on the model type and the number of pages analyzed. Also, when a document is submitted for analysis, the service processes all pages unless a page range is explicitly specified.

So even if you only have fewer than 15 documents and each one is only 2–3 pages, the cost can still increase if the same files were analyzed repeatedly during testing, Studio usage, pipeline retries, or multiple model runs.

A few things to check:

In the Document Intelligence resource, go to Monitoring / Metrics and review Processed Pages

Split the metric by FeatureName to see which model or feature is generating the usage

In Azure Cost Analysis, group by Resource and Meter to confirm whether the charge is coming only from Document Intelligence or whether other Foundry-related resources/meters are also contributing

If custom models are involved, verify whether layout-based label generation was used, because that can also be billed

One important point: training a custom model itself is free, but document analysis and certain supporting operations such as layout usage for label generation can still incur charges.

Based on the document volume you described, a cost of about USD 80/day would usually suggest repeated processing or additional billable usage, i would say :) ..

Maybe this links would help for clarification :

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/how-to-guides/estimate-cost?view=doc-intel-4.0.0

https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/service-limits?view=doc-intel-4.0.0

https://learn.microsoft.com/en-us/azure/foundry/concepts/manage-costs

Hope this helps.