AZ-204 vs AI-200: How the Azure Developer Exam Became an AI Back-End Exam

Exam AZ-204 retired on 31 July 2026, and the Microsoft Certified: Azure Developer Associate credential retired with it. Microsoft lists Exam AI-200, which awards Microsoft Certified: Azure AI Cloud Developer Associate, as the replacement.

The new exam is not AZ-204 with a new number. Storage was 15–20% of AZ-204. Its successor domain, AI solutions on Azure data management services, is the largest on AI-200 at 25–30%, and it covers Cosmos DB vector search, PostgreSQL with pgvector and Azure Managed Redis. Azure Blob Storage is no longer in the skills measured.

Below is how the domains moved, what left the syllabus, what is new and what still carries over. Gnoseed does not sell, proctor or guarantee this exam, and it is not a course: it is a free study companion. Confirm every detail on Microsoft Learn before you book, since Microsoft updates the skills measured periodically. The same replacement pattern hit the AI engineer track, covered in AI-102 vs AI-103.

AZ-204 published five domains: Develop Azure compute solutions (25–30%), Develop for Azure storage (15–20%), Implement Azure security (15–20%), Monitor and troubleshoot Azure solutions (5–10%), and Connect to and consume Azure services and third-party services (20–25%).

AI-200 publishes four: Develop containerized solutions on Azure (20–25%), Develop AI solutions by using Azure data management services (25–30%), Connect to and consume Azure services (20–25%), and Secure, monitor, and troubleshoot Azure solutions (20–25%).

Two moves matter. Containers were one part of the compute domain and now have a domain of their own, while Azure Functions sits with messaging and eventing under the services domain. Security and monitoring were separate domains and are now one, so do not read its 20–25% as growth: added together, they were already 20–30% on AZ-204.

AZ-204 asked you to perform operations on Cosmos DB and Blob Storage through the SDK. AI-200 goes much further with three services. On Azure Cosmos DB for NoSQL you optimise Request Unit consumption with indexing policies and consistency levels, store embeddings and run vector similarity search.

On Azure Database for PostgreSQL you design schemas and indexes, reduce pgvector compute overhead, size compute, memory and storage for vector workloads, and implement retrieval-augmented generation with metadata filters. On Azure Managed Redis you handle caching, expiration and invalidation as well as vector indexing.

This is the section that a study plan carried over from AZ-204 will miss. PostgreSQL and Redis were not in the AZ-204 skills measured at all, and Microsoft now names vector databases and Python programming in the candidate profile. Neither appeared in the AZ-204 study guide.

Comparing the two study guides, these AZ-204 skills have no counterpart in the AI-200 skills measured: Azure API Management, Microsoft Graph, user authentication through the Microsoft identity platform and Microsoft Entra ID, shared access signatures, managed identities, Azure Blob Storage, Azure Event Hubs, Azure Queue Storage, Application Insights and App Service deployment slots.

Treat that list as a guide to your study time, not as advice about real projects. All of these services still exist and most Azure developers still need several of them. The AI-200 study guide also keeps some of them in its general documentation links, so an isolated question touching one is possible. They are simply no longer listed skills.

The containers domain goes beyond building and publishing images. It now includes Azure Container Apps environment configuration and revision management, event-driven scaling with KEDA, deploying to Azure Kubernetes Service with manifest files, and troubleshooting AKS and Container Apps through logs, events and connectivity.

Monitoring changed character too. AZ-204 centred on Application Insights. AI-200 lists tracing distributed systems with OpenTelemetry SDKs and writing KQL queries to analyse logs and metrics.

Several skills appear in both exams: Cosmos DB SDK operations, consistency levels and the change feed; Azure Container Registry; deploying containers to App Service; Azure Functions triggers and bindings; Azure Event Grid and Azure Service Bus; Azure Key Vault secrets; and Azure App Configuration. The messaging skills also went deeper, with dead-letter handling, topics and subscriptions, and Event Grid filters and retries named directly.

Gnoseed has a free Azure AI Cloud Developer (AI-200) deck weighted to the four published domains, with cards on Container Apps, Cosmos DB, pgvector, Azure Managed Redis, messaging and monitoring. It is a study companion rather than a course, so pair it with Microsoft Learn and hands-on practice. The other decks linked on this page cover the vendor-neutral ideas the exam builds on: embeddings, vector search and RAG, dead-letter queues and event-driven design, container images, OpenTelemetry tracing, and secret rotation.