Clarification on India-Only Data Residency for Azure OpenAI, Embeddings, and Azure AI Document Intelligence

Hello Azure Community,

We are currently evaluating Azure services for an enterprise insurance platform that serves Indian customers with strict data residency requirements. Our customer requires a clear confirmation that data must not leave India at any stage of processing, including document ingestion, embeddings, and LLM inference. We would appreciate your clarification on the following points so we can finalize and update our architecture.1. Azure OpenAI – LLM Inference in India We plan to use Azure OpenAI models (for example, GPT-4.1 and GPT-4.1-mini) for policy-related question answering. Could you please confirm:

• Whether Azure OpenAI supports LLM inference fully within India, when an Azure OpenAI resource is deployed in an India region.
• Whether there are any scenarios where:
  • Requests are routed to another region, or
    • Inference is executed outside India due to capacity, fallback, or internal routing.
    • If India-only inference is possible, do you provide:
      • A written guarantee or contractual clause confirming that prompts, retrieved context, and outputs remain within India?
        • Any documentation that explicitly states this behavior?

2. Embedding Models – India-Only Processing We also generate embeddings for:

• Policy document chunks
• User queries

Our requirement is that text used for embeddings must not leave India. Could you please clarify:

• Whether Azure OpenAI embedding models (or any Azure embedding services) can guarantee that:
  • Embedding generation happens entirely within India, and
    • No input text is processed outside the India region.
    • If not guaranteed by default:
      • Are there specific configurations, SKUs, or contracts that enforce India-only embedding inference?

3. Azure AI Document Intelligence – Data Residency We plan to use Azure AI Document Intelligence (formerly Form Recognizer) for document ingestion and extraction (PDFs, scanned documents, tables). Please confirm:

• If an Azure AI Document Intelligence resource is created in an India region, does Microsoft guarantee that:
  • Document content,
    • Extracted text,
      • Intermediate processing, and
        • Temporary storage

          all remain within the India region.

• Whether there is any cross-region processing or storage involved under normal or exceptional conditions.
• If there is any official documentation or contractual assurance covering this behavior.

We are looking for explicit confirmation or documentation on the above points to finalize our architecture.

Thank you in advance for your support.
Best regards.