AI-200 Replaces AZ-204: What Changed in 2026

AI-200 replaces AZ-204 as the Microsoft associate exam for Azure developers, and the focus has moved to the back end of AI applications. According to Microsoft Learn, AZ-204 retired on July 31, 2026, and the current AI-200 skills outline is dated October 7, 2026.

AI-200 is called Developing AI Cloud Solutions on Azure. It earns a new credential, Microsoft Certified: Azure AI Cloud Developer Associate. This guide shows what moved, what was dropped and how to prepare.

Does AI-200 replace AZ-204?

For Azure back-end developers, yes. Microsoft retired AZ-204 and the Azure Developer Associate certification on July 31, 2026, and AI-200 now tests the same layer of the stack: hosting, data, messaging, security and monitoring.

The credential name changed. Passing AI-200 earns the Azure AI Cloud Developer Associate, not the Azure Developer Associate.

The emphasis changed too. Microsoft’s AI-200 profile describes “an emphasis on back-end services and components” for AI solutions, and lists vector databases among the skills you should be proficient in.

What changed between AZ-204 and AI-200?

The outline went from five domains to four, and data services became the largest one. In AZ-204, storage was 15-20% of the exam. In AI-200, data management services for AI carry 30-35%.

A vector database stores embeddings, numeric representations of text or images, so an application can find the most similar items to a query. AI-200 tests this in Cosmos DB, PostgreSQL and Redis.

AZ-204 (retired July 31, 2026)	Weight	AI-200 (skills as of October 7, 2026)	Weight
Develop Azure compute solutions	25-30%	Develop containerized solutions on Azure	20-25%
Develop for Azure storage	15-20%	Develop AI solutions by using Azure data management services	30-35%
Implement Azure security	15-20%	Secure, monitor, and troubleshoot Azure solutions	20-25%
Monitor, troubleshoot, and optimize Azure solutions	5-10%	(merged into the domain above)
Connect to and consume Azure services and third-party services	20-25%	Connect to and consume Azure services	20-25%

The study guide’s section heading for the data domain still shows 25-30%, while its skills-at-a-glance list shows 30-35%. The change log says that domain’s share increased on October 7, 2026, so 30-35% is the current figure.

What was removed from the exam?

Several AZ-204 services do not appear in the AI-200 skills list. Azure Blob Storage, Azure API Management, Event Hubs, Queue Storage and Azure Container Instances are gone.

The identity work has narrowed as well. AZ-204 asked for authentication with the Microsoft identity platform and Microsoft Entra ID, shared access signatures and Microsoft Graph. AI-200 keeps Azure Key Vault and Azure App Configuration for secrets and settings.

Microsoft notes that related topics may still appear. Treat this as a change of weight, not a promise that a service never shows up.

What is new on the exam?

According to the AI-200 study guide, the new skills include:

• event-driven scaling with Kubernetes Event-driven Autoscaling (KEDA) in Azure Container Apps;
• deploying and troubleshooting applications on Azure Kubernetes Service (AKS) with manifest files;
• storing embeddings and running vector similarity search in Azure Cosmos DB for NoSQL;
• modelling schemas, indexing and RAG patterns in Azure Database for PostgreSQL with pgvector;
• caching and vector indexing in Azure Managed Redis;
• tracing with OpenTelemetry SDKs and writing KQL queries over logs and metrics.

KQL (Kusto Query Language) is the query language used to search Azure Monitor logs and metrics. pgvector is the PostgreSQL extension that adds a vector data type and similarity search.

What stayed the same from AZ-204?

Containers, Functions, Cosmos DB basics, messaging and secrets carried over. Both outlines ask you to build and publish images to Azure Container Registry and deploy to Azure App Service and Azure Container Apps.

Cosmos DB still covers the SDK, consistency levels and the change feed. Service Bus and Event Grid remain for messaging, now with dead-letter handling and event filters named explicitly.

Azure Functions triggers and bindings are still tested, and Key Vault and App Configuration still guard secrets and settings.

How long is the AI-200 exam?

Microsoft gives AI-200 120 minutes. The AZ-204 certification page does not show a duration now that it is retired, so plan on the full 120 for AI-200.

Microsoft does not publish a question count. Its exam-experience page says most exams have 40-60 questions, and associate exams that may contain labs run 120 minutes with 140 minutes of seat time.

A score of 700 or greater is required to pass. Microsoft Learn is available in a split screen, but the clock keeps running while you use it. Practise the 120-minute pace with timed AI-200 practice tests.

Who is AI-200 for, and who is it not for?

AI-200 is for developers who build and run the Azure services that AI applications depend on. Microsoft’s profile lists Azure SDKs, data management services, messaging, monitoring, vector databases, Python and containers.

It suits:

• former AZ-204 candidates who need a current, renewable associate credential;
• back-end developers adding vector search and retrieval to existing applications;
• engineers who operate containers on AKS or Container Apps and troubleshoot them.

It is a weaker fit for:

• developers whose work is prompts, models and agents, who should look at AI-103;
• newcomers to Azure AI, who may prefer to start with AI-901.

How should you study for AI-200? A 6-week plan

Put the most time into data services, then containers and operations. The plan assumes six to eight hours a week of hands-on work.

1. Week 1: containers. Build, version and push images to Azure Container Registry, run ACR Tasks, and deploy to App Service with environment variables and secrets.
2. Week 2: orchestration. Deploy to Container Apps with revisions and KEDA scaling, then to AKS with manifests, and practise reading logs and events.
3. Week 3: Cosmos DB. Query with the SDK, tune indexing and consistency, store embeddings, run vector search, and process the change feed.
4. Week 4: PostgreSQL and Redis. Design a pgvector table with indexes, run filtered similarity search, and add caching and vector indexing in Azure Managed Redis.
5. Week 5: messaging, Functions and security. Build Service Bus queues and topics with dead-letter handling, Event Grid filters, a Function with bindings, and Key Vault secret rotation.
6. Week 6: observe and practise. Instrument an app with OpenTelemetry, write KQL over its logs, then sit full timed tests with a case study and review misses by domain.

Mixing domains in each review session works better than finishing one topic at a time, as our article on three ways to study smarter explains.

What happens to your Azure Developer Associate certification?

It stays on your Microsoft Learn transcript but can no longer be renewed. Microsoft’s certification page states that the certification and its renewal assessment are retired.

Associate certifications normally expire annually and renew through a free online assessment. AI-200 is now the way to hold a current associate credential for Azure development.

Is AI-200 harder than AZ-204?

It is broader in data and operations and narrower in identity and integration services. The largest domain, data management for AI, asks for design choices such as indexing strategy and connection optimisation.

Candidates who mainly memorised AZ-204 service features may find AI-200 more practical. The monitoring and troubleshooting work, now 20-25% shared with security, rewards time spent reading real logs.

Before you book, check the AI-200 study guide for the current skills date.

Related reading

• AI-103 replaces AI-102: what changed for Azure AI engineers
• AI-901 replaces AI-900: what changed in 2026
• What 8,600 practice questions reveal about exams
• All CertCrafter articles