AI-901 is Microsoft's Azure AI Fundamentals exam — the replacement for the long-running AI-900 — and it validates entry-level skills in AI concepts, responsible AI, and building AI solutions with Microsoft Foundry. The key change in 2026: where AI-900 asked you to describe Azure AI services, AI-901 asks you to build with them. AI-900 retires on June 30, 2026, so AI-901 is now the path to the Azure AI Fundamentals credential.
If you're starting an AI career or adding an Azure AI credential to a cloud résumé, this guide covers the official skills measured (as published April 15, 2026), exam logistics, the AI-900-to-AI-901 jump, and how to actually pass it. The facts here come from Microsoft's official AI-901 study guide — verify against it before you book, since this exam is brand new.
What is the AI-901 certification?
AI-901 is the exam that earns the Microsoft Certified: Azure AI Fundamentals certification. It's aimed at candidates at the beginning of an AI development career who need conceptual knowledge of Azure AI plus the foundational technical skills to work with it. There are no formal prerequisites, but the audience profile now expects Python coding syntax knowledge and familiarity with Azure resources.
Importantly, AI-901 is a replacement exam code, not a new credential. If you pass AI-901 you get the same Azure AI Fundamentals badge AI-900 awarded. The difference is entirely in what the exam measures — and that difference is significant.
AI-900 vs AI-901: what changed
The short answer: AI-901 moved from recognition to application. AI-900 tested whether you could identify Azure AI services and concepts. AI-901 tests whether you can deploy a model, write a prompt, build a chat app, and stand up an agent in Microsoft Foundry.
| AI-900 (retiring) | AI-901 (current) | |
|---|---|---|
| Approach | Describe services | Build solutions |
| Platform focus | Individual cognitive services | Unified Microsoft Foundry |
| Python required | No | Yes (basic) |
| Agentic AI | Not covered | Covered (build + test agents) |
| Content Understanding | Not covered | Covered (docs, images, audio, video) |
| Retirement | June 30, 2026 | — |
This mirrors a pattern I keep seeing across the cert landscape in 2026: fundamentals exams are getting harder and more practical because "I can name the service" no longer differentiates anyone. The vendors know employers want people who've touched the console. AI-901 reflects where enterprise AI actually sits now — generative models, Copilot-style apps, and agents — not the 2021-era "here are the five cognitive services" framing.
If you're weighing where this fits among Microsoft exams, AI-901 is the AI-track sibling to the cloud-track AZ-900 Azure Fundamentals. Many people take both as their entry pair before moving to an associate cert like AZ-104.
AI-901 skills measured (official domains)
AI-901 has two domains. The weighting is heavily tilted toward hands-on implementation, which is the single most important thing to internalize before you study.
| Domain | Weight |
|---|---|
| Identify AI concepts and capabilities | 40–45% |
| Implement AI solutions by using Microsoft Foundry | 55–60% |
Domain 1 — Identify AI concepts and capabilities (40–45%)
This is the descriptive half, closest to old AI-900 material. It breaks into three areas:
- Responsible AI principles — fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. Six principles; expect at least one scenario question mapping a situation to the right principle.
- AI model components and configurations — how generative AI models work, choosing an appropriate model by capability, and model deployment options and configuration parameters.
- AI workloads — recognizing scenarios for generative and agentic AI, text analysis (keyword extraction, entity detection, sentiment analysis, summarization), speech recognition and synthesis, computer vision and image generation, and information extraction across text, images, audio, and video.
Note that even the "concepts" domain now lists agentic AI as a workload you must recognize. That word appears nowhere in the retiring AI-900 outline.
Domain 2 — Implement AI solutions by using Microsoft Foundry (55–60%)
This is the new center of gravity, and it's where unprepared candidates lose. Every objective starts with a verb like create, deploy, build, or test — you're expected to have actually done these things in the Foundry portal at ai.azure.com. The sub-areas:
- Generative AI apps and agents — write system and user prompts, deploy a model and interact with it in the Foundry portal, build a lightweight chat client with the Foundry SDK, and create and test a single-agent solution.
- Text and speech — build a text-analysis app, respond to spoken prompts with a deployed multimodal model, and use Azure Speech in Foundry Tools.
- Computer vision and image generation — interpret visual input with a multimodal model, generate new visual outputs, and build a vision-enabled app.
- Information extraction — use Azure Content Understanding to pull structured data from documents, forms, images, audio, and video.
Microsoft's own note on the exam: most questions cover generally available features, you should be familiar with REST APIs, SDKs, and CLIs, and Preview features may appear if they're commonly used. Translation — don't try to read your way through Domain 2.
AI-901 exam details at a glance
| Detail | Value |
|---|---|
| Exam code | AI-901 |
| Certification | Microsoft Certified: Azure AI Fundamentals |
| Cost | $99 USD (varies by region) |
| Passing score | 700 / 1000 (scaled) |
| Questions | ~40–60 (Microsoft doesn't publish a fixed count for fundamentals) |
| Duration | ~45–60 minutes |
| Prerequisites | None (but basic Python + Azure familiarity expected) |
| Skills outline updated | April 15, 2026 |
A note on the question count and duration: Microsoft does not publish fixed figures for fundamentals exams, so the ~40–60 questions and ~45–60 minutes are reasonable expectations based on the format, not guarantees. The passing score of 700/1000 is official and confirmed in Microsoft's exam scoring documentation. Because the score is scaled, it doesn't translate to a clean "you need X correct."
How to pass AI-901
The fastest reliable path: split your prep along the 40/60 domain split and spend the larger share of your time inside the Foundry portal, not in slides.
- Get hands-on in Foundry first. Open ai.azure.com, deploy a model, send a chat completion, build and test a single agent, and run Content Understanding against a document. Doing each task once is worth more than any amount of reading for Domain 2. This is non-negotiable.
- Get your Python to "read and tweak" level. You won't write production code, but you need to follow a Foundry SDK snippet and predict what it does. If you can read a basic chat-completion script and change the prompt, you're at the right level.
- Drill responsible AI. The six principles (fairness, reliability/safety, privacy/security, inclusiveness, transparency, accountability) are easy points if you can map a scenario to the right one. Memorize them cold.
- Use Microsoft Learn's learning paths and the free practice assessment. They're written to the exact outline. Pair them with the official study guide so you're not studying stale AI-900 content — a real risk right now, since most third-party guides haven't caught up to the April 2026 refresh.
Two to four weeks of part-time study is typical for a beginner. If you already hold AZ-900 or have touched any LLM API, you can compress the concepts half and concentrate on Foundry hands-on.
My honest take on whether it's worth it: AI-901 is a vocabulary and signal cert. It gets your résumé past keyword filters and proves you understand Azure's AI stack — but on its own it won't land an AI role. Pair it with an associate-level cert and a small portfolio of things you've actually built. Fundamentals get the interview scheduled; the portfolio gets you through it. That's true of every fundamentals cert, and AI-901 is no exception — the value is being early on the credential while the SERP and hiring filters are still catching up.
Build the Foundry skills, don't just read about them
Because Domain 2 is 55–60% of the exam and entirely hands-on, the candidates who pass AI-901 comfortably are the ones who've deployed a model and built an agent at least once. Reading the objectives isn't the same as doing them.
Build those reps with CloudaQube's hands-on AI and cloud labs — practice deploying models, writing prompts, and standing up agents in a live environment so the Foundry questions feel like review, not first contact. If you're mapping a broader path, see which AI skills employers are hiring for in 2026 and where this cert fits on the AI engineer career path.
Sources: Microsoft Learn — AI-901 study guide, Microsoft Learn — Exam AI-901, Microsoft Certified: Azure AI Fundamentals.
