Clarification Needed: Is GPT-4 Model Inference Supported in Azure UAE North Region?

Hi Muhammad Akther

At present, while Azure OpenAI resources can be provisioned in UAE North, model inference (e.g., GPT-3.5, GPT-4) is not executed within UAE North data centers. Instead, the actual processing is routed to other supported regions such as West Europe or France Central, where Azure OpenAI infrastructure is currently available. This means that prompt and response data may leave the UAE for model execution, which is an important consideration for organizations with strict data residency or compliance requirements.

Microsoft does not currently guarantee that prompt and response data will remain within the UAE if the region does not support model execution. Although UAE North is paired with UAE Central for data resiliency, this does not imply support for local inference capabilities. Customers in regulated industries are advised to avoid using real or sensitive data until model execution is fully supported within the region.

As of now, there is no publicly confirmed timeline for when full model execution support will be available in UAE North. Microsoft continues to expand regional support based on infrastructure readiness and regulatory alignment.

For further details and updates, please refer to the following official resources:

• https://learn.microsoft.com/en-us/legal/cognitive-services/openai/data-privacy
• https://azure.microsoft.com/en-us/global-infrastructure/geographies/
• https://www.microsoft.com/en-us/trust-center/privacy/data-location

Please let us know if you need assistance with alternative deployment strategies or escalation pathways for compliance assurance.

Hope it helps!

Thank you

Thanks Ravada .

As a follow up, could you please advise if any other LLM models available in Azure (e.g., smaller models or partner hosted options) currently support regional inference within the UAE, ensuring that prompt and response data do not leave the country? We're exploring compliant alternatives and would appreciate guidance on any models or configurations that meet strict data residency requirements.

Appreciate your continued support and insights.

Thanks

Hi Muhammad Akther

Sorry for Delayed Response

As of now, Azure OpenAI Service does not support local model inference within the UAE North region. Even if resources are provisioned in UAE North, the actual inference for models such as GPT-3.5 and GPT-4 is routed to other regions like West Europe or France Central. This means that prompt and response data may leave the UAE, which does not meet strict data residency requirements.

For organizations exploring compliant alternatives, Azure AI Foundry offers a broader catalog of models, including partner-hosted options such as Cohere, Meta Llama, Falcon/TII, Snowflake Arctic, and SDAIA (Arabic model). Some of these models may be hosted on third-party infrastructure or deployed via Azure Arc, which could allow regional control if configured appropriately. However, none of these models are currently confirmed to support full inference within UAE borders, and infrastructure readiness remains a key constraint.

Azure’s roadmap includes the AI Extended Zones (AEZ) Strategy, which aims to expand localized AI infrastructure, including GPU deployments and private links. While UAE is not yet listed among the deployed AEZ zones, future expansion is planned for regions near Azure Hero zones, which could eventually support regional inference.

For teams considering migration to compliant alternatives, Microsoft provides detailed guidance through resources like the Azure OpenAI Model Migration Guidance and the LLMOps maturity model, which outline phased rollout strategies, traffic routing via Azure Front Door, and rollback mechanisms.

Reference links: Data, privacy, and security for Azure OpenAI Service

Hope it helps!

Thank you

Hi Ravada,

Thanks for the detailed explanation. Could you please clarify in Azure AI Foundry, when we select Regional Provisioned Throughput (regional hosting model), does this guarantee that inference stays fully within the selected region (e.g., UAE North), or is it still routed to other Azure regions?

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/provisioned-throughput?tabs=global-ptum

@Ravada Shivaprasad kindly see the above query