AI-900 vs AZ-900: Azure AI Fundamentals or Azure Fundamentals?

AI-900 vs AZ-900 in 2026: one of them has changed

Most AI-900 vs AZ-900 comparisons online describe two similar beginner exams: one about Azure, one about AI concepts. That picture is out of date. Microsoft retired Exam AI-900 on June 30, 2026. The Microsoft Certified: Azure AI Fundamentals certification still exists, but you now earn it by passing Exam AI-901, which is more technical than AI-900 was.

So the real comparison today is AI-901 vs AZ-900. Both are Fundamentals exams with the same format rules and a certification that does not expire. They differ in subject and, more than before, in how technical they are.

All facts below come from Microsoft Learn's certification pages and study guides and Microsoft's Tech Community announcements, checked on 7 November 2026. If you already hold Azure AI Fundamentals from AI-900, it stays valid: Microsoft says retirement does not revoke earned certifications.

Side-by-side comparison

Both list the same 13 languages: English, Arabic, Chinese (Simplified and Traditional), French, German, Indonesian, Italian, Japanese, Korean, Portuguese (Brazil), Russian and Spanish.

What each exam covers

AZ-900

AZ-900 is about the platform. You describe the shared responsibility model, cloud models, IaaS, PaaS and SaaS; Azure regions, availability zones, resource groups, subscriptions and management groups; compute, networking and storage options; identity and security services such as Microsoft Entra ID, Conditional Access, RBAC and Defender for Cloud; and governance and cost tools such as Azure Policy, resource locks, tags, the pricing calculator and Azure Monitor. Every objective starts with "describe" or "compare".

AI-901

AI-901 has two halves. The first covers concepts: responsible AI principles, how generative models work, choosing and configuring models, and AI workloads from agentic AI to speech, vision and information extraction. The second, and larger, half asks you to implement: deploy a model and use it in the Foundry portal, build a lightweight chat client with the Foundry SDK, create and test a single agent, and build small apps for text, speech, vision and information extraction with Foundry Tools such as Azure Speech and Content Understanding.

The overlap between the two exams is small. AZ-900 mentions no AI services; AI-901 assumes you are already familiar with Azure resources, which is exactly what AZ-900 teaches.

Which is harder?

For most people, AI-901 is now the harder of the two. AZ-900 can be passed with no technical background through study alone. AI-901 expects you to read Python and to have built things in Foundry, and more than half of the exam is about implementation.

That is a change from the old pairing. Microsoft's own comparison describes AI-900's audience as "nontechnical or technical beginners" with no coding required, and AI-901's as "technical beginners planning to build AI solutions" with basic Python. If someone tells you AI-900 is the easy one, they are describing the retired exam.

Microsoft does not publish pass rates for either exam.

Where the difficulty lies is different too. On AZ-900, the traps are similar-sounding services and governance tools: knowing when Azure Policy, a resource lock or an RBAC role assignment is the right answer. On AI-901, the traps are practical: which model suits a task, which deployment setting changes the output, and which Foundry tool extracts information from a form or a video. The concepts half of AI-901 also rewards careful reading of responsible AI scenarios, where several principles can seem to apply and you have to pick the one the question is really about. If you have never written code, budget for that learning curve before you book AI-901, rather than discovering it in the exam.

Which should you take first?

This is our editorial recommendation, based on Microsoft's audience descriptions.

Taking both is reasonable and inexpensive in time compared with role-based exams. If you do both, AZ-900 then AI-901 is the natural order, because the second builds on the first.

Time to prepare and where each leads

Preparation time differs more than the identical 45-minute format suggests. These are our estimates; Microsoft does not publish study hours.

The extra time for AI-901 goes into hands-on work: setting up a Foundry project, deploying a model, building small client apps and an agent. Budget a little Azure spending for that, and delete resources after each session.

Where they lead also differs. AZ-900 is the usual first step towards the Azure Administrator path and other role-based Azure exams. Azure AI Fundamentals is the first step on Microsoft's AI developer path, which continues at associate level with exams for building AI apps and agents. If you want both an infrastructure and an AI direction, AZ-900 first keeps the sequence logical.

Preparing for each

For AZ-900, use the free Microsoft Learn path, explore the Azure portal with a free or trial account, and focus on the distinctions questions test most: IaaS vs PaaS vs SaaS, regions vs availability zones, and Azure Policy vs resource locks vs RBAC. Our AZ-900 study guide has a short plan, and our free AZ-900 practice test shows where you stand.

For AI-901, plan time in a real Azure subscription with a Foundry project, and practise the build tasks, not only the concepts. Be careful with older AI-900 courses and question banks: they cover the concepts half but not the Foundry implementation half. Our AI-900 study guide has a 2-week AI-901 plan, and our Azure AI Fundamentals practice test helps you find weak areas.

Microsoft offers free Practice Assessments for both exams; for Azure AI Fundamentals it is on AI Skills Navigator and requires sign-in. Neither exam lets you open Microsoft Learn during the test, so practise without notes and within 45 minutes.