AI-901 Replaces AI-900: What Changed in 2026

AI-901 replaces AI-900 as the exam for Microsoft Certified: Azure AI Fundamentals, and the change is bigger than a new code. According to Microsoft Learn, AI-900 retired on June 30, 2026, and the AI-901 skills outline took effect on April 15, 2026.

The old exam asked you to describe AI. The new one asks you to build small things with it. This guide walks through what moved, what was dropped, and how to plan your preparation.

Does AI-901 replace AI-900 completely?

Yes. The Azure AI Fundamentals certification page now lists AI-901 as its only required exam, and Microsoft’s retired-exams list shows AI-900 retired on June 30, 2026.

A Fundamentals certification is Microsoft’s entry-level credential, meant for people at the start of a role rather than practitioners with years of experience. The certification keeps its name. Only the exam behind it changed.

If you earned AI-900 before it retired, Microsoft states that the certification stays on your Microsoft Learn transcript. There is nothing to re-sit.

What changed between AI-900 and AI-901?

The outline shrank from five domains to two, and the weight moved from describing AI to implementing it. Microsoft’s study guide puts 55-60% of AI-901 on building solutions with Microsoft Foundry.

Microsoft Foundry is Microsoft’s Azure platform for deploying models and building AI apps and agents. Foundry Tools is the family name Microsoft now uses for services such as Azure Speech and Azure Content Understanding.

AI-900 (retired June 30, 2026)	Weight	AI-901 (skills as of April 15, 2026)	Weight
Describe AI workloads and considerations	15-20%	Identify AI concepts and capabilities	40-45%
Describe fundamental principles of machine learning on Azure	15-20%	Implement AI solutions by using Microsoft Foundry	55-60%
Describe features of computer vision workloads	15-20%
Describe features of natural language processing workloads	15-20%
Describe features of generative AI workloads	20-25%

The verbs tell the story. AI-900 objectives start with “describe” and “identify”. Half of AI-901’s objectives start with “create”, “deploy”, “build” or “extract”.

What was removed from the exam?

The machine learning domain is gone. AI-900 tested regression, classification and clustering scenarios, features and labels, training and validation data, and Azure Machine Learning capabilities such as automated machine learning.

None of those objectives appear in the AI-901 outline. The service-by-service descriptions of Azure AI Vision and the Azure AI Face detection service are also absent.

What is new on the exam?

Hands-on Foundry work is new. According to the AI-901 study guide, candidates should be able to:

• create effective system and user prompts for generative AI models;
• 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 build a lightweight client for it;
• build a lightweight application by using Azure Speech in Foundry Tools;
• extract information from documents, forms, images, audio and video by using Azure Content Understanding in Foundry Tools.

An agent here means an AI solution that combines a model with instructions and tools so it can carry out a task, not just answer one prompt. AI-901 stops at a single agent; multi-agent orchestration belongs to the associate exam.

What stayed the same from AI-900?

Responsible AI stayed, almost word for word. Both outlines ask you to describe considerations for fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability.

The workload families also survive inside the first AI-901 domain. You still need to recognise text analysis, speech, computer vision and information extraction scenarios, now alongside generative and agentic AI.

What changes is depth. In AI-901 those same workloads reappear in the second domain as things you build, so knowing the names is no longer enough.

How many questions are on the AI-901 exam?

Microsoft does not publish a question count for AI-901. Its exam-experience page says most Microsoft Certification exams contain between 40 and 60 questions and that the number can vary.

The same page sets Fundamentals exams at 45 minutes of exam time and 65 minutes of seat time. Seat time is the total you should allow, including instructions, the candidate agreement and optional comments after the exam.

A score of 700 or greater is required to pass. Five minutes for unscheduled breaks are built into the exam time, but the clock keeps running while you are away.

Do you need to code for AI-901?

You need to read code comfortably, not write large programs. The AI-901 audience profile asks for knowledge of Python coding syntax and programming techniques and familiarity with Azure resources.

That is a real shift. The retired AI-900 profile said data science and software engineering experience were not required.

In practice, expect to follow short Python snippets that call a deployed model, a speech service or Content Understanding, and to judge what each line does. The study guide also notes you should be familiar with REST APIs, SDKs and CLIs.

Who is AI-901 for, and who is it not for?

AI-901 suits people at the beginning of a career in AI solution development, which is how Microsoft describes the audience.

It is a good fit for:

• developers and IT staff who want a first Azure AI credential before an associate exam;
• analysts or students comfortable reading Python who want proof they can work in Foundry;
• managers of AI projects who are willing to do the hands-on labs, not only the reading.

It is a weaker fit for:

• people who want a purely conceptual, no-code AI overview, which is what AI-900 used to offer;
• experienced Azure AI engineers, who are better served by AI-103, the associate exam covered in our guide on how AI-103 replaces AI-102.

How should you study for AI-901? A 4-week plan

Plan around the weights: a little under half for concepts, a little over half for building in Foundry. This plan assumes about five hours a week.

1. Week 1: concepts and responsible AI. Learn the six responsible AI principles until you can match each to a scenario. Map every workload type to a business example.
2. Week 2: models and prompts. Deploy a model in the Foundry portal, compare deployment options, and write system and user prompts that change the output in predictable ways.
3. Week 3: build the small apps. Create the lightweight chat client with the Foundry SDK, a single agent with its client, a speech app, and a vision prompt. Read every line of each script aloud.
4. Week 4: extraction and timed practice. Run Content Understanding on a form, an image and an audio file. Then sit full timed practice tests, review every miss by domain, and repeat.

Spaced, mixed review beats rereading; our post on three ways to study smarter explains why.

What happens if you already hold AI-900?

Nothing is taken away. Microsoft’s retired-exams page says that if you already earned a certification, it stays on the transcript in your Microsoft Learn profile.

Your credential still reads Microsoft Certified: Azure AI Fundamentals. Sitting AI-901 is optional for you, and worth it mainly if you want evidence of the newer Foundry skills.

If your next step is the associate level, compare the AI-103 changes or the back-end focused AI-200 changes before you decide.

Is AI-901 harder than AI-900?

It asks for more practical skill, so most candidates should plan for more preparation time. The difficulty comes from the 55-60% implementation domain and the Python reading, not from deeper theory.

AI-900 could be prepared mostly from reading. AI-901 rewards time spent in the Foundry portal, because questions about deploying, prompting and extracting are easier when you have done each task once.

Microsoft also removes Microsoft Learn access during Fundamentals exams. You cannot look anything up, so recall matters.

How do you know you are ready to book?

You are ready when you can explain each responsible AI principle with an example and walk through each build task without notes. Use Microsoft’s free practice assessment on AI Skills Navigator for the question style.

Then sit at least two full timed AI-901 practice tests and look at your score by domain, not only the total. If the Foundry domain trails, spend another week building before you book.

Before you schedule, recheck the AI-901 exam page for the current price in your country, the available languages and any outline update.

Related reading

• AI-103 replaces AI-102: what changed for Azure AI engineers
• AI-200 replaces AZ-204: what changed for Azure developers
• Are AI and data certifications worth taking in 2026?
• All CertCrafter articles