Skip to content

April'26 Updates #1704

@shubhirajMsft

Description

@shubhirajMsft

April '26 Updates

Welcome to the April 2026 edition of Azure Container Apps updates.
This month, we’re seeing strong momentum around AI workloads, especially with serverless GPUs, along with emerging real-world deployment patterns across AI-native applications.


🤖 AI & Serverless GPU Workloads

Running multimedia AI workloads on ACA

A new blog demonstrates running ComfyUI (text-to-image and text-to-video) on Azure Container Apps using serverless GPUs.

Highlights:

  • Uses ACA Jobs to download models dynamically
  • Deploys GPU-backed workloads with Terraform
  • Integrates with Azure Monitor for observability
  • Supports A100 and T4 GPUs

🔗 Running multimedia AI models on Container Apps with Serverless GPU


Self-hosted AI models on ACA (Gemma 4 + Ollama)

A new Apps on Azure blog demonstrates deploying Gemma 4 models with Ollama on Azure Container Apps serverless GPUs, enabling fully private, self-hosted AI workloads.

Key patterns:

  • Fully private inference (data stays in customer subscription)
  • OpenAI-compatible API endpoint for easy integration
  • Ability to choose model size and GPU tier (T4 or A100)
  • One-command deployment using Azure Developer CLI (azd)

The deployment runs Ollama + Gemma 4 inside a container app, exposing a secure endpoint while Azure manages GPU infrastructure and scaling.

🔗 Gemma 4 on Azure Container Apps Serverless GPU


⚙️ Platform Trends & Usage Patterns

Across April updates, a few clear patterns emerge:

  • Serverless GPUs driving AI adoption

    • Running custom models (LLMs, multimedia) directly on ACA
    • Leveraging scale-to-zero and per-second billing
  • Shift toward self-hosted AI

    • Models deployed within customer's Azure Subscription
    • Increasing preference over external API-based inference
  • Growing use of ACA Jobs

    • Used for model setup, data prep, and batch workflows
    • Enabling flexible, event-driven execution patterns

Overall, these trends reinforce ACA’s positioning as:

A serverless platform for running AI-native applications with full control over compute, scaling, and data.


🔗 Useful Links

Metadata

Metadata

Assignees

No one assigned

    Labels

    ANNOUNCEMENTAnnouncement from the product group

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions