Azure document intelligence production server specs

Question

Azure document intelligence production server specs

moaz alla 20

Hello
i was trying to estimate the specs need to run the azure disconnected containers (layout ,read and custom template ) on a production environment that will serve 50 concurrent users , we assume that each user will send about 10~15 pages (dense pages with detailed tables or figures ), i know that on the documentation page it mentions the specs for each container but it doesn't say any details about how much users each container will handle with these specs , i need to know an the estimated specs to run a production server in environment mentioned above ?

Thanks in advanced .

moaz alla 20 Reputation points

2025-12-02T08:40:25.0066667+00:00
thank you for the answer but i have 2 more question about what you just said

For your scenario with about 50 concurrent users uploading 10–15 dense pages each, a stronger setup is recommended. A good starting point would be a server with 8–12 CPU cores, 32–48 GB RAM, and SSD storage with at least 20–30 GB free.

do you mean per container or for my entire setup (layout , read and custom template )?

about the specs need for custom template are those for the custom template container only or for the layout and studio containers needed by the custom template container to function too ?
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-02T12:38:38+00:00
Hi moaz alla,

The numbers I gave (8–12 CPU cores, 32–48 GB RAM, and sufficient SSD space) refer to the overall machine/VM hosting all your Document Intelligence containers together, not per container. In most production environments, customers run Layout, Read, and Custom Template containers on the same server, so the capacity needs to support their combined load.

The heavier requirements mainly come from the Custom Template container, because it handles training and inference of more complex models. However, Layout and Read are still required for Custom Template to function, and they also consume CPU/RAM just less. So, the recommended specs cover all three containers running together on the same machine.

If you plan to run each container on a separate server, then yes the Custom Template server would need the larger resources, while Layout and Read could run on smaller machines. But if everything will be deployed on a single machine (which is the most common setup), use the combined sizing I mentioned earlier.

Let me know if you want a per-container sizing breakdown or any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!
moaz alla 20 Reputation points

2025-12-02T12:57:38.9633333+00:00

ok thank you very much my question is answered
but one extra thing about the studio container dose these spec (the one in the official docs and the one you said ) account for it ?
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-02T14:52:17.42+00:00

Hi moaz alla ,

yes, the specifications listed in the official documentation (and the ones I mentioned earlier) apply to the Document Intelligence Studio container as well. The container uses the same underlying compute requirements as the production service because it runs the same model stack locally.
Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Answer accepted by question author

0 additional answers

Your answer

moaz alla 20 Reputation points

2025-12-02T08:40:25.0066667+00:00

thank you for the answer but i have 2 more question about what you just said

For your scenario with about 50 concurrent users uploading 10–15 dense pages each, a stronger setup is recommended. A good starting point would be a server with 8–12 CPU cores, 32–48 GB RAM, and SSD storage with at least 20–30 GB free.

do you mean per container or for my entire setup (layout , read and custom template )?

about the specs need for custom template are those for the custom template container only or for the layout and studio containers needed by the custom template container to function too ?
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-02T12:38:38+00:00

Hi moaz alla,

The numbers I gave (8–12 CPU cores, 32–48 GB RAM, and sufficient SSD space) refer to the overall machine/VM hosting all your Document Intelligence containers together, not per container. In most production environments, customers run Layout, Read, and Custom Template containers on the same server, so the capacity needs to support their combined load.

The heavier requirements mainly come from the Custom Template container, because it handles training and inference of more complex models. However, Layout and Read are still required for Custom Template to function, and they also consume CPU/RAM just less. So, the recommended specs cover all three containers running together on the same machine.

If you plan to run each container on a separate server, then yes the Custom Template server would need the larger resources, while Layout and Read could run on smaller machines. But if everything will be deployed on a single machine (which is the most common setup), use the combined sizing I mentioned earlier.

Let me know if you want a per-container sizing breakdown or any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!
moaz alla 20 Reputation points

2025-12-02T12:57:38.9633333+00:00

ok thank you very much my question is answered
but one extra thing about the studio container dose these spec (the one in the official docs and the one you said ) account for it ?
Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-02T14:52:17.42+00:00

Hi moaz alla ,

yes, the specifications listed in the official documentation (and the ones I mentioned earlier) apply to the Document Intelligence Studio container as well. The container uses the same underlying compute requirements as the production service because it runs the same model stack locally.
Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Answer 1

Hi moaz alla,

Thank you for reaching out on the Microsoft Q&A.

There is no official Microsoft sizing guide for how much server power you need for a certain number of users in Azure Document Intelligence. The documentation only provides minimum requirements for running the containers, such as 8 CPU cores and around 16–24 GB of RAM depending on the model. However, community experience shows that these minimum specs are usually not enough for real production workloads, especially when documents contain many pages, tables, images, or scanned content. Most users report that performance becomes smoother and more stable when the server has at least 32 GB of RAM and fast SSD storage for temporary processing.

For your scenario with about 50 concurrent users uploading 10–15 dense pages each, a stronger setup is recommended. A good starting point would be a server with 8–12 CPU cores, 32–48 GB RAM, and SSD storage with at least 20–30 GB free. This setup gives the service enough headroom to handle multiple heavy documents at the same time. Since actual performance depends on the complexity and quality of your documents, the best next step is to run a load test with your real files and monitor CPU, memory, and processing time.

This testing will help you adjust your resources as needed. If you see the server getting overloaded, you can either increase the machine size or run multiple container instances behind a load balancer. This approach ensures you get reliable performance without over- or under-provisioning your system.

References Links:

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

Thankyou!

Anshika Varshney 3,795 Reputation points Microsoft External Staff Moderator

2025-12-02T14:57:24.2833333+00:00

Hi moaz alla,

I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted? This helps others in the community with the same question find the solution more easily.

Share via

Azure document intelligence production server specs

0 additional answers

Your answer