Use Azure AI Services in Containers

Did you know you can run Azure AI services on your own infrastructure using containers? Microsoft provides container images for individual AI service APIs through the Microsoft Container Registry (MCR). This allows you to deploy AI closer to your data on-premises, in Azure, or even at the edge.

How It Works

Using an Azure AI services container involves three main steps

1. Deploy the container: Pull the image for your desired AI service and run it on Docker, Azure Container Instance (ACI), or Azure Kubernetes Service (AKS)

2. Consume the service: Client applications send data to the container endpoint, just like a cloud-based Azure AI API. Sensitive data stays local

3. Report usage for billing: The container periodically sends usage metrics to your Azure AI service for billing. You still need an Azure resource provisioned for billing even if all processing happens locally

Visual workflow

Client Application --> Container Endpoint --> Local Processing
                                       \
                                        --> Azure AI Resource (for billing only)

Available Container Images

Each container provides a subset of a service’s functionality.

For example:

• Language Services: Key Phrase Extraction, Language Detection, Sentiment Analysis, Translation, Summarization
• Speech Services: Speech to Text, Custom Speech, Neural Text to Speech, Language Detection
• Vision Services: Read OCR, Spatial Analysis

Some containers are in gated public preview and may require access approval

Configuration

When deploying a container you must provide

• ApiKey from your Azure AI service for billing
• Billing endpoint URI of your Azure resource
• Eula set to accept

Once deployed, your application communicates with the container endpoint. No subscription key is required, and you can implement your own authentication and network security policies.

Visual configuration workflow

Container Host
  |--> ApiKey and Billing Endpoint
  |--> Accept Eula
  |--> Expose Container Endpoint for Client Apps

Why Use Containers

• Keep sensitive data local and secure
• Process data closer to the source for low latency
• Scale easily with container orchestration
• Maintain the same API as Azure cloud services

Containers make Azure AI services flexible, secure, and fast while letting you process data locally. They are ideal for scenarios such as healthcare, manufacturing, retail, or any environment where privacy and low latency are important