Connected containers on AWS/GCP

Yes. Document Intelligence connected containers can be deployed on non-Azure environments such as AWS or GCP, as long as the runtime requirements are met and the containers can reach the Azure billing endpoint.

From the container perspective, the key requirements are:

1. Docker-compatible host
   • A Linux-capable Docker Engine must be installed on the host. This can be on macOS, Windows, or Linux, and Docker containers can also be deployed to Kubernetes or other orchestrators.
   • The platform can be any cloud or on-premises environment that supports Docker (for example, AWS EC2, GCP Compute Engine, or Kubernetes clusters in those clouds).
2. Azure resource for billing
   • A Document Intelligence (or multi-service AI) resource in Azure is required, with its ApiKey and Billing (endpoint URI) configured for the container.
   • The container must be able to connect to Azure to send billing data. Docker must be configured to allow outbound connectivity so the container can reach the Azure metering service.
3. Continuous connectivity for connected containers
   • Connected containers are not licensed to run without being connected to Azure for metering. They periodically send billing information (about every 10–15 minutes). If they cannot reach the billing endpoint for an extended period, they stop serving requests.

There are no restrictions in the documentation that limit connected containers to Azure-hosted compute only; the documented constraints are about Docker support and network connectivity to Azure for billing, not the underlying cloud provider.

If a fully offline scenario is required (no connectivity to Azure), disconnected containers and a commitment plan must be used instead of connected containers.

References:

• What are Azure containers?
• Install and run containers
• Configure Document Intelligence containers
• Frequently asked questions
• Containers in disconnected (offline) environments