CertQuests
Why Microsoft is retiring AZ-204 now
AZ-204 has been the defining developer credential in the Azure ecosystem since 2020. The exam tested the ability to build cloud solutions using Azure Compute (VMs, App Service, Azure Functions), Azure Storage, Azure Cosmos DB, Azure Cache for Redis, API Management, and authentication via Microsoft Identity Platform. For five years, this scope accurately described the core competency set of an Azure developer.
That scope is no longer sufficient. AI integration has gone from an optional add-on to a core expectation of cloud development work. Azure OpenAI Service, Azure AI Search with vector indexing, Azure Container Apps running model inference workloads, and event-driven pipelines consuming AI model output have become standard components in production applications. The old AZ-204 curriculum did not test any of these at meaningful depth — not because the exam was poorly designed, but because the technology landscape moved faster than a certification update cycle could track.
Rather than patch the existing exam with AI modules, Microsoft made a structural decision: retire AZ-204 and replace it with a credential designed from the ground up around AI-native development patterns. The result is AI-200, an exam where the AI integration is not a single domain bolted onto existing content but the lens through which every technical domain is examined.
The AZ-204 → AI-200 transition timeline
What AI-200 covers: the four exam domains
The AI-200 exam is structured around four domains that reflect the full development lifecycle of an AI-integrated cloud application — from the containerized compute layer through data management and connectivity to security and observability.
Domain 1 — Develop Containerized Solutions on Azure
Container-native development is central to AI-200 in a way it was peripheral to AZ-204. This domain tests the ability to build and deploy containerized workloads on Azure Container Apps and Azure Kubernetes Service, with an emphasis on applications that serve or consume AI models. Key areas include writing multi-stage container builds optimised for inference workloads, configuring autoscaling rules based on AI pipeline queue depth, managing container registries with Azure Container Registry, and orchestrating model-serving containers with environment variable injection for API keys and endpoint configuration.
Candidates must understand the difference between Azure Container Apps (event-driven, serverless container hosting suited to stateless AI API wrappers) and AKS (full Kubernetes for complex, stateful, or latency-sensitive model deployments). The exam tests selecting the appropriate compute target based on workload characteristics rather than configuring either service in exhaustive depth.
Domain 2 — Develop AI Solutions Using Azure Data Management Services
The data management domain covers the infrastructure that AI applications use to store, retrieve, and process the information they work with. This is where AI-200 diverges most sharply from AZ-204. Rather than testing general storage pattern knowledge (Blob vs Table vs Queue selection), AI-200 tests vector-specific data management — configuring Azure AI Search with vector indexes for semantic retrieval, managing embeddings in Azure Cosmos DB for NoSQL with vector search enabled, and designing data pipelines that transform raw content into searchable vector representations.
Retrieval-Augmented Generation (RAG) architecture is explicitly in scope: the exam tests building the pipeline from document ingestion through chunking, embedding generation via Azure OpenAI, vector storage, and similarity search at query time. Candidates who have built RAG applications against Azure AI Search will find this domain familiar; those who have only worked with relational databases or blob storage will need dedicated preparation.
Domain 3 — Connect to and Consume Azure AI Services
This domain covers the integration layer: how Azure developers wire AI capabilities from Azure AI services into application code. The scope includes Azure OpenAI Service (chat completions, embeddings, fine-tuned model deployment, prompt engineering, and token budget management), Azure AI Content Safety (moderation filters in the request/response path), Azure AI Document Intelligence (form extraction and structured data output), and the Azure AI Foundry SDK for multi-service orchestration.
A significant portion of this domain tests responsible AI integration practices: implementing content filters, logging prompt and completion pairs for audit, rate-limiting to stay within provisioned throughput units (PTUs), and handling model-level refusals gracefully at the application layer. These concerns did not appear in AZ-204 and reflect Microsoft’s position that AI safety is a developer responsibility, not an optional operations concern.
Domain 4 — Secure, Monitor, and Troubleshoot Azure Solutions
The final domain covers the cross-cutting concerns that keep AI applications secure and observable in production. Key areas include Azure Key Vault integration for managing API keys and model endpoint secrets (the exam specifically tests avoiding hardcoded credentials in container images or environment variables in source control), Managed Identity for zero-credential access to Azure AI services, and Azure Monitor with Application Insights configured to capture AI-specific telemetry — latency per model call, token consumption rates, embedding generation throughput, and vector search query performance.
Distributed tracing of AI pipelines receives dedicated attention. Because an AI-enabled request often traverses an event trigger, a container app, a vector search call, and a model completion call before returning a response, the exam tests configuring correlation IDs that survive the entire chain and surfacing the resulting trace in Application Insights. This is meaningfully more complex than the single-hop request tracing that AZ-204 covered.
AZ-204 vs AI-200: what actually changes
The shift from AZ-204 to AI-200 is not cosmetic. It represents a genuine change in what Microsoft considers core competency for an Azure developer in 2026. Several AZ-204 topics carry forward — Azure Functions, App Service, authentication with Microsoft Identity Platform, and Azure Key Vault appear in both exams — but the context in which they are tested changes completely. In AZ-204, Azure Functions was tested as a general compute trigger mechanism. In AI-200, Azure Functions is tested as a serverless event handler that invokes AI model endpoints and routes the output to downstream services.
Topics that are present in AZ-204 but absent or greatly reduced in AI-200 include Azure Cache for Redis, Azure Service Bus deep configuration, API Management policy authoring, and CDN management. Topics that are new to AI-200 and absent from AZ-204 include vector databases, embedding pipelines, RAG architecture, Azure OpenAI Service SDK usage, AI Content Safety integration, and AI-specific observability patterns.
Candidates who have passed AZ-204 will not find AI-200 trivially familiar. The compute and identity foundations overlap, but the data layer and AI integration domains require dedicated study. Plan for 60–80 hours of preparation if you hold AZ-204 and are transitioning to AI-200.
What happens to your AZ-204 credential after July 31?
Retired Microsoft certifications remain permanently visible on your Microsoft Learn transcript and are always verifiable through the Credly badge system. The badge and transcript entry show the certification as “Retired” rather than expired, which distinguishes it from certifications that lapsed due to non-renewal. Employers who verify credentials through Credly can see the full history including the retirement status. The credential does not disappear — it becomes a historical record rather than an active one.
There is no automatic migration or credit transfer. Passing AZ-204 does not reduce the requirements for AI-200, and Microsoft has not announced a discount or voucher for AZ-204 holders transitioning to AI-200. The two certifications are separate credentials that test separate competencies. This is consistent with how Microsoft handled the AZ-103 to AZ-104 transition in 2020 and the AZ-900 update in 2022: existing holders keep their old credential, and earning the new one requires passing the new exam.
Who should still take AZ-204 before July 31?
If you are mid-preparation for AZ-204 and the retirement date is weeks away, the calculus depends on how far along you are. AZ-204 material that transfers cleanly to AI-200 — Azure Functions, App Service, authentication patterns, Key Vault — represents roughly 35–40% of AZ-204’s measured skills. The other 60–65% (Cache for Redis, Service Bus, API Management, CDN, classic storage patterns) has minimal overlap with AI-200.
If you are within three weeks of exam-ready for AZ-204, finishing and passing it before July 31 is a reasonable decision: you earn a credential that remains on your transcript, and the foundational Azure knowledge you have built accelerates your AI-200 preparation. If you are at the beginning of preparation with no strong preference, studying directly for AI-200 is the more strategically sound path — every hour spent on vector databases and Azure OpenAI SDK is preparation for AI-200 that you would have to add anyway.
Microsoft Learn hosts the official AI-200 study guide with the full list of measured skills and recommended training paths. The official course is AI-200T00 (Develop AI cloud solutions on Azure). For hands-on practice, configure Azure AI Search with vector indexing, build a minimal RAG pipeline against Azure OpenAI, and deploy a containerized AI API to Azure Container Apps — those three exercises cover the highest-weighted skill areas in the exam.
The broader pattern: Microsoft pivots every developer cert toward AI
AZ-204’s retirement is not an isolated event. Microsoft has been systematically repositioning its developer-track certifications around AI fluency since early 2026. AZ-802 (a new AI agent development credential currently in beta) is expected to go live in August 2026, targeting developers who build autonomous AI agents beyond out-of-the-box scenarios. AI-900 (Azure AI Fundamentals) was refreshed in March 2026 to cover generative AI and responsible AI principles at the foundational level. The pattern is clear: Microsoft’s certification program now treats AI integration as a baseline expectation for developers, not a specialist track.
For IT professionals building their Azure credential stack, this has a practical implication: prioritise AI-200 over AZ-204 in any plan started from today. The returning AZ-204 renewal candidates — those who passed in 2023 and are approaching their three-year renewal window — will not be able to renew through the existing AZ-204 path after July 31. Microsoft has confirmed that the AI-200 beta exam is accepted as a renewal mechanism for AZ-204 holders in the six-month window before the retirement date. After retirement, renewal is no longer an option; AZ-204 holders who want an active associate-level Azure developer credential must earn AI-200.
Practice Azure concepts and prepare for your next Microsoft certification on CertQuests.
Browse Azure Certifications →