If you already hold the Azure AI Fundamentals certification, you keep it and have nothing to do. Microsoft retired the AI-900 exam on June 30, 2026, and its AI-901 announcement says that if you hold the certification, "no action is required". If you studied for AI-900 and never sat it, your notes on responsible AI, AI workloads, computer vision, text analysis, speech and generative AI still match objectives on AI-901, and most of the machine learning domain does not. More than half of the skills AI-901 measures are about implementing AI solutions in Microsoft Foundry, and they assume you know some Python, which AI-900 did not require.
The certification
The certification kept its name. Microsoft changed the exam that earns it, and the certification page now lists AI-901 as the only required exam. The same announcement says "taking AI-901 will not change the name of your existing credential."
You do not need to renew it either. Microsoft's renewal page says "Fundamentals certifications do not expire", and its March 2026 retirement post adds that "Retirement does not revoke or invalidate Certifications that were earned."
If you sat AI-900 before the retirement and did not pass, the same post says you cannot retake a retired exam. Your next attempt has to be AI-901.
Whether a holder should sit AI-901
Microsoft says holders "may choose to take AI-901 to showcase your knowledge of the latest AI technologies and a broader set of AI skills." Passing it does not change the certification you hold or its name. Sitting it makes most sense if your work is moving toward building AI apps on Azure, because the preparation covers deploying a model, calling it from Python and building a simple agent in Foundry. If you have no plans to build on Foundry, it stays optional.
Exam format
Both exams sit at the Fundamentals level and share a passing score of 700 and 45 minutes of exam time (65 minutes in the seat).
| Detail | AI-900 | AI-901 |
|---|---|---|
| Status and US price | Retired June 30, 2026 | $99 as of October 7, 2026 (set by the country where you test), no retirement date listed |
| Domains | Five (AI workloads, machine learning, computer vision, natural language processing, generative AI), per the AI-900 study guide, skills measured as of May 2, 2025 | Two (Identify AI concepts and capabilities, Implement AI solutions by using Microsoft Foundry), per the AI-901 study guide, skills measured as of April 15, 2026 |
| Weights | 15-20% each for the first four, 20-25% for generative AI | 40-45% concepts, 55-60% Foundry implementation |
| Coding knowledge, from Microsoft's comparison in the AI-901 announcement | None | Basic Python syntax and programming concepts |
| Audience | Technical and non-technical backgrounds, no data science or software engineering experience needed | People starting out in AI solution development, familiar with Azure resources |
| What it means for you | Its concept topics still help | Plan new study for the Foundry domain |
How AI-900 topics map to AI-901
| AI-900 topic | Closest AI-901 objective | What to do with your notes |
|---|---|---|
| Six responsible AI principles (fairness, reliability and safety, privacy and security, inclusiveness, transparency, accountability) | "Describe principles of responsible AI", the same six principles | Keep them as they are |
| Identify computer vision, NLP, document processing and generative AI workloads | "Identify scenarios for common AI workloads, including generative and agentic AI, text analysis, speech, computer vision, and information extraction" | Keep them, and add agentic AI |
| Key phrase extraction, entity recognition, sentiment analysis, Azure AI Language service | "keyword extraction, entity detection, sentiment analysis, and summarization", plus building an app that includes text analysis | Keep them, and add summarization and the build task |
| Language modeling, the Transformer architecture | "Describe how generative AI models work". Microsoft's training for it, the Large language models unit, teaches tokenization, transformers and embeddings | Keep them at the depth of that unit |
| Translation | No objective names it | Low priority |
| Speech recognition and synthesis, Azure AI Speech service | "speech recognition and speech synthesis", plus building an app with Azure Speech in Foundry Tools | Keep the concepts, and add the build task |
| Image classification, object detection, OCR, facial detection and analysis, Azure AI Vision and Face services | "computer vision and image-generation models", plus visual input to a multimodal model and building an app with vision | Keep the concepts, and add image generation; no objective names the Face service |
| Generative AI models, scenarios and responsible generative AI; Azure AI Foundry, Azure OpenAI, model catalog | How generative AI models work, choosing a model by its capabilities, deployment options and configuration parameters, prompts | Keep the model concepts, and add model choice and deployment settings |
| Regression, classification, clustering, deep learning | No objective names them | Low priority |
| Features and labels, training and validation datasets | No objective names them | Low priority |
| Azure Machine Learning (automated ML, data and compute services, model management) | No objective names them | Low priority |
A low-priority topic can still turn up, because the AI-901 study guide says "Related topics may be covered in the exam." Study those after everything that has its own objective.
Some of the change is renaming. Microsoft's Foundry overview maps the old "Azure AI Studio / Azure AI Foundry" brand to Microsoft Foundry and "Azure AI Services" to Foundry Tools. So the Azure AI Speech service in your AI-900 notes appears on AI-901 as Azure Speech in Foundry Tools. Document processing became its own skill group. AI-900 asked you to identify document processing workloads, and AI-901 asks you to pull information out of documents, forms, images, audio and video with Azure Content Understanding, then build a small app that does it.
The Python and Foundry work
In the AI-901 study guide, 13 of the Foundry domain's 15 objectives are worded as tasks starting with create, deploy, build or extract. Microsoft doesn't identify question types before an exam, so read those tasks as the skills you need. The study guide lists tasks such as these:
- Deploy a model and interact with it in the Foundry portal.
- Create a lightweight chat client application by using the Foundry SDK.
- Create and test a single-agent solution in the Foundry portal, then a client application for the agent.
- Build lightweight applications for text analysis, speech, vision and information extraction.
Microsoft's AI-901 exam page says you need "knowledge of Python coding syntax and programming techniques", and the study guide adds that you should be familiar with REST APIs, SDKs and CLIs. Microsoft's learning path for this domain, Get started with AI applications and agents on Azure, expects "a basic understanding of computing concepts and Python" before you start, and one of its modules has you "build a lightweight Python client application that analyzes text."
If you have never written Python, learn enough to write and run a short script that imports a library and calls it before you start the Foundry modules. The "build a lightweight application" objectives depend on it. If you already code, the new work is the Foundry portal and SDK, and your AI-900 concept knowledge still explains what each app does.
Studying with AI-900 material
Treat the AI-901 "Skills measured" list as your checklist, and judge every resource against it, including Microsoft's own pages. On October 7, 2026, the AI-901 study guide's "Find documentation" row still linked Anomaly Detector, Language Understanding, Azure Machine Learning and Azure Bot Service, and none of those has an AI-901 objective. The scheduling card on the AI-901 exam page still described the five AI-900 areas, machine learning included. The skills list is the part Microsoft updated on April 15, 2026.
A workable order for someone with AI-900 notes is to refresh the concepts domain from those notes first, then put most of your new study time into the Foundry domain, which carries more weight and is new to you. The AI-901 exam page lists two self-paced learning paths of seven modules each. AI concepts for developers and technology professionals teaches "core AI concepts and terminology", and Get started with AI applications and agents on Azure covers the Foundry work. Microsoft's free practice assessment for AI-901 is on AI Skills Navigator, and you have to sign in there to start it.
To see where you stand on AI-901 questions after studying for AI-900, take CramHQ's free AI-901 assessment. It takes about 15 minutes, needs no credit card and shows which areas are weakest when you finish. The AI-901 Pass Plan costs $19.99 once for 12 months of access, orders repair tasks by how many exam points each gap costs, and includes five practice tests and a 40-question timed simulation.