
Microsoft retired several popular Azure certifications in 2026 and replaced them with new exams built around AI and Microsoft Foundry. If a course page, a YouTube playlist or a friend tells you to "do AI-102" or "do AZ-204", that advice is now out of date. This guide lists every change confirmed on Microsoft Learn: which exam replaced which, what the new exams actually test, and what to do if you were halfway through preparing or already hold an old certification.
All facts here come from Microsoft Learn certification pages, study guides and Microsoft's Skills Hub announcement, checked in November 2026. Microsoft changes exams often, so always open the official exam page before booking.
Exam guides for the new exams: AI-103 (replaces AI-102) · AI-901 (replaces AI-900) · AI-200 (replaces AZ-204).
The changes at a glance
| Retired certification (exam) | Retired | Replacement (exam) |
|---|---|---|
| Azure Data Scientist Associate (DP-100) | 1 June 2026 | Machine Learning Operations Engineer Associate (AI-300) |
| Azure AI Fundamentals (AI-900) | 30 June 2026 | Azure AI Fundamentals (AI-901) |
| Azure AI Engineer Associate (AI-102) | 30 June 2026 | Azure AI Apps and Agents Developer Associate (AI-103) |
| Azure Developer Associate (AZ-204) | 31 July 2026 | Azure AI Cloud Developer Associate (AI-200) |
| Azure Security Engineer Associate (AZ-500) | 31 August 2026 | Cloud and AI Security Engineer Associate (SC-500) |
| Windows Server Hybrid Administrator Associate (AZ-800 + AZ-801) | Planned for September 2026 | Windows Server Administrator Associate (AZ-802) |
| Azure Data Engineer Associate (DP-203) | 31 March 2025 (earlier change) | Fabric Data Engineer Associate (DP-700) |
Not affected: AZ-900 Azure Fundamentals, AZ-104 Azure Administrator and AZ-400 DevOps Engineer Expert are all still active. AZ-104 remains the main certification for Azure administrator jobs.
I already have an old certification. Is it still valid?
Yes. Microsoft says earned credentials stay valid until they expire; a retirement does not revoke them. Keep these points in mind:
- Associate certifications expire every 12 months. You can renew one only while the certification is still active, so once it retires there is no renewal. After it expires, take the replacement exam.
- Fundamentals certifications (AZ-900, AI-900) never expire. An AI-900 you passed stays on your profile.
- In interviews, say both names: "I hold AI-102 (Azure AI Engineer) and I'm preparing for AI-103." That shows you are keeping up.
I was preparing for AI-102 or AZ-204. What should I do?
Your preparation is not wasted. Most of the topics carry over to the new exams, with more weight on generative AI and agents:
- From AI-102 to AI-103: Azure AI Search, Document Intelligence, Vision, Language and Speech still appear. The big shift is that 30–35% of AI-103 is generative AI and agents built on Microsoft Foundry: RAG, tool calling, multi-agent orchestration, evaluation and tracing. Start with our Microsoft Foundry tutorial and Foundry RAG with Azure AI Search.
- From AZ-204 to AI-200: containers, Functions, Key Vault, messaging and monitoring stay. New areas are vector search in Cosmos DB, PostgreSQL (pgvector) and Azure Managed Redis, which together make up 25–30% of the exam. Our pgvector RAG tutorial covers the core idea.
What the new exams test
AI-103: Developing AI Apps and Agents on Azure
Microsoft describes the candidate as an Azure AI engineer who builds and deploys agents and AI solutions on Microsoft Foundry, using Python.
| Skill area | Weight |
|---|---|
| Plan and manage an Azure AI solution | 25–30% |
| Implement generative AI and agentic solutions | 30–35% |
| Implement computer vision solutions | 10–15% |
| Implement text analysis solutions | 10–15% |
| Implement information extraction solutions | 10–15% |
Topics to focus on:
- Choosing between models and Foundry services.
- Keyless (managed identity) security and private networking.
- Guardrails and safety evaluations.
- Building agents with retrieval, function calling and memory.
- Multi-agent orchestration with approval steps.
- Tracing and token analytics.
- Hybrid and vector search.
- Azure Content Understanding for extracting data from documents, images and video.
Related reading:
AI-901: Microsoft Azure AI Fundamentals
AI-901 is a far more hands-on exam than AI-900. You are expected to know basic Python syntax.
| Skill area | Weight |
|---|---|
| Identify AI concepts and capabilities | 40–45% |
| Implement AI solutions by using Microsoft Foundry | 55–60% |
You'll be asked to:
- deploy a model in the Foundry portal;
- build a small chat client with the Foundry SDK;
- create and test a single agent;
- use Speech and Content Understanding.
The responsible AI principles (fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability) are still tested.
AI-200: Developing AI Cloud Solutions on Azure
This is the back-end developer exam: AZ-204, updated for AI.
| Skill area | Weight |
|---|---|
| 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% |
| Secure, monitor and troubleshoot Azure solutions | 20–25% |
Topics include:
- Containers and hosting: Azure Container Registry, Container Apps with KEDA scaling, AKS manifests.
- Data and vectors:
- Cosmos DB for NoSQL, including vector search and the change feed.
- PostgreSQL with pgvector.
- Azure Managed Redis vector indexing.
- Messaging and serverless: Service Bus and Event Grid, Azure Functions.
- Security and monitoring: Key Vault, App Configuration, OpenTelemetry tracing and KQL.
SC-500: Implementing End-to-End Security Controls for Cloud and AI Workloads
This replaces AZ-500 and adds AI security.
| Skill area | Weight |
|---|---|
| Manage identity, access and governance | 20–25% |
| Secure storage, databases and networking | 25–30% |
| Secure compute | 20–25% |
| Manage and monitor security posture | 20–25% |
The "secure compute" area now includes securing AI:
- Microsoft Entra Agent ID and Conditional Access for agents.
- Defender for AI Services.
- An AI gateway in API Management for Foundry.
- Agent guardrails in Foundry.
- Purview DSPM for Copilot risks.
It also covers PIM, Key Vault, Azure Firewall, private endpoints, Defender for Cloud, Sentinel and Security Copilot.
AI-300: Operationalizing Machine Learning and Generative AI Solutions
This replaces DP-100 and covers both MLOps and GenAIOps.
| Skill area | Weight |
|---|---|
| Design and implement an MLOps infrastructure | 15–20% |
| Implement machine learning model lifecycle and operations | 25–30% |
| Design and implement a GenAIOps infrastructure | 20–25% |
| Implement generative AI quality assurance and observability | 10–15% |
| Optimize generative AI systems and model performance | 10–15% |
Expect:
- Azure Machine Learning with MLflow.
- Bicep and GitHub Actions.
- Model deployment with safe rollback.
- Data drift.
- Foundry model deployments and provisioned throughput.
- Prompt versioning in Git.
- Groundedness and safety evaluations.
- RAG tuning and fine-tuning.
See our guides on LLMOps, LLM evaluation and fine-tuning LLMs.
Which certification should you take now?
| Your goal | Suggested path |
|---|---|
| Fresher, first cloud certification | AZ-900 → AZ-104 |
| Fresher interested in AI | AI-901 → AI-103 |
| Azure administrator / cloud support | AZ-104 → AZ-400 or SC-500 |
| .NET, Java or Python back-end developer on Azure | AI-200 (plus AZ-104 if you also manage infrastructure) |
| AI engineer / GenAI developer | AI-103 (and AWS Bedrock knowledge if your employer is multi-cloud) |
| Data scientist / ML engineer | AI-300 |
| Security engineer / SOC | SC-200 → SC-500 |
| Data engineer on Microsoft Fabric | DP-700 |
| Multi-cloud architect | AZ-104 + AWS Solutions Architect Associate |
Common questions
Are the new exams harder? They are more hands-on. AI-103, AI-901 and AI-200 all expect Python and working knowledge of Foundry or vector databases. Hands-on practice in a real subscription matters more than memorising dumps (which also break Microsoft's exam rules).
Do I still need AZ-900 before AZ-104? No, there are no prerequisites. AZ-900 remains a useful first step for freshers.
Will employers still accept AI-102 or AZ-204? Yes, until the certification expires. Job descriptions change slowly, so many will still list AI-102 or AZ-204 in 2026–27. Recruiters increasingly recognise the new names too.
Does the AZ-400 prerequisite change? Microsoft Learn currently lists Azure Administrator Associate *or* Azure Developer Associate as the prerequisite for DevOps Engineer Expert. Check the AZ-400 page before you plan around it.
Next steps
- Read which certifications freshers should take first.
- Browse live IT jobs to see which certifications employers actually ask for this month.
Cloud Soft Solutions in Ameerpet, Hyderabad teaches Azure, Azure AI and Microsoft Foundry, AWS and DevOps with real-time projects and placement support. See all courses.
📞 Call +91 96660 19191 · WhatsApp · Flat No. 513, 5th Floor, Nilagiri Block, Aditya Enclave, Ameerpet, Hyderabad
Frequently Asked Questions
What replaced AI-102?
AI-103: Developing AI Apps and Agents on Azure, for the Azure AI Apps and Agents Developer Associate certification. AI-102 retired on 30 June 2026.
What replaced AZ-204?
AI-200: Developing AI Cloud Solutions on Azure, for the Azure AI Cloud Developer Associate certification. AZ-204 retired on 31 July 2026.
What replaced AZ-500?
SC-500, for the Cloud and AI Security Engineer Associate certification. AZ-500 retired on 31 August 2026.
What replaced DP-100?
AI-300, for the Machine Learning Operations Engineer Associate certification. DP-100 retired on 1 June 2026.
Is AI-900 still available?
No. AI-900 retired on 30 June 2026 and was replaced by AI-901, which keeps the Azure AI Fundamentals name but is more hands-on with Microsoft Foundry.
Is my AI-102 or AZ-204 certification still valid?
Yes, until it expires. Retirement does not revoke earned credentials, but a retired associate certification cannot be renewed.
Are AZ-900 and AZ-104 retired?
No. AZ-900, AZ-104 and AZ-400 are still active.



