For years, AI-900 was the gentle way into AI on Azure: no code, mostly concepts, open to anyone. That exam has now gone. Its replacement, AI-901, keeps the same certification title but asks you to build things. If you studied with older AI-900 material, or you are wondering whether your existing badge still counts, this guide sets out exactly what changed.
| Fact | AI-900 (retired) | AI-901 (current) |
|---|---|---|
| Status | Retired 30 June 2026, 11:59 PM CST | Live; English version updated 15 April 2026 |
| Certification | Microsoft Certified: Azure AI Fundamentals | Microsoft Certified: Azure AI Fundamentals |
| Audience | Technical and non-technical beginners | Technical beginners planning to build AI solutions |
| Coding | None required | Basic Python syntax and programming techniques |
| Exam time | 45 minutes | 45 minutes |
| Pass mark | 700 on a scale of 1 to 1,000 | 700 on a scale of 1 to 1,000 |
| Fee | Priced by country | Priced by country |
We confirmed these dates on Microsoft Learn and the Microsoft Tech Community Skills Hub blog on 19 November 2026. The AI-900 study guide now carries a retirement notice, and the Azure AI Fundamentals certification page lists AI-901 as its only required exam.
The timeline
Microsoft announced the change on its Skills Hub blog in April 2026. AI-901 launched in beta, with the English version of the skills outline dated 15 April 2026. The blog said the certification would reach general availability in June 2026 and that, after 30 June 2026, candidates would need to pass AI-901 to earn it. The retired AI-900 study guide confirms the exam closed on 30 June 2026 at 11:59 PM Central Standard Time.
The AI-901 exam page lists 13 languages, including English, Arabic, Japanese, Spanish and Portuguese (Brazil). Microsoft updates the English version first and says localised versions usually follow about eight weeks later.
Did the certification name change?
No. This is the most common point of confusion. The credential is still called Microsoft Certified: Azure AI Fundamentals, and it still carries the Fundamentals badge. Microsoft describes this as updating the certification rather than retiring it. What changed is the exam you sit to earn it, and with it the level of skill being tested.
Existing AI-900 holders
If you passed AI-900, your certification remains valid. Microsoft states that existing holders need take no action, and that taking AI-901 will not rename the credential you already hold. Fundamentals certifications do not expire and need no annual renewal.
You might still choose to sit AI-901 if your role now involves building with generative AI, because it proves hands-on skills that AI-900 never tested. That is optional.
If an employer or job advert asks for "AI-900", your existing pass meets it. If you are certifying for the first time, AI-901 is now the only route, and it leads to the same credential.
How the skills changed
AI-900: five conceptual areas
The final AI-900 outline, skills measured as of 2 May 2025, had five roughly equal areas: AI workloads and considerations, machine learning principles on Azure, computer vision, natural language processing, and generative AI. Every objective started with "describe" or "identify".
AI-901: two areas, mostly practical
| AI-901 skill area | Weight | Sub-areas |
|---|---|---|
| Identify AI concepts and capabilities | 40 to 45% | Responsible AI principles; AI model components and configurations; AI workloads |
| Implement AI solutions by using Microsoft Foundry | 55 to 60% | Generative AI apps and agents; text and speech; computer vision and image generation; information extraction |
Over half the exam is now about implementation. Objectives use verbs such as create, deploy, build and extract.
Side by side
| Topic | AI-900 | AI-901 |
|---|---|---|
| Responsible AI principles | Yes, six principles | Yes, same six principles |
| Classic machine learning (regression, classification, clustering) | Yes, its own area | Not listed in the outline |
| Azure Machine Learning and automated ML | Yes | Not listed in the outline |
| Features, labels, training and validation data | Yes | Not listed in the outline |
| How generative models work, choosing a model | Partly, as features of generative AI | Yes, plus deployment options and parameters |
| Prompts | Not a separate objective | Write system and user prompts |
| Agents | Not covered | Create and test a single agent in the Foundry portal, build a client app for it |
| Foundry SDK | Not covered | Build a lightweight chat client |
| Text analysis and speech | Describe service capabilities | Build apps with text analysis and Azure Speech in Foundry Tools |
| Computer vision | Image classification, object detection, OCR, face | Multimodal visual input, image generation, a vision app |
| Information extraction | Document processing workloads | Azure Content Understanding for documents, forms, images, audio and video |
What is new
Generative AI as the centre. Instead of one area out of five, generative AI now runs through the whole exam. You need to understand how generative models work, how to choose one by capability, and which deployment options and parameters matter.
Agents. AI-901 expects you to create and test a single agent solution in the Foundry portal and to build a lightweight client application that talks to it. Agents did not appear in AI-900 at all.
Microsoft Foundry. The portal and SDK are now the main tools. Expect to deploy a model and interact with it in the portal, and to read or reason about simple Python that calls a deployed model.
Multimodal models. Speech and vision are approached partly through multimodal models: answering spoken prompts with a deployed model, and interpreting images in prompts.
Content Understanding. Information extraction now centres on Azure Content Understanding in Foundry Tools, across documents, images, audio and video.
Microsoft notes that most AI-901 questions cover generally available features, but commonly used preview features may appear too, and that you should be familiar with REST APIs, SDKs and CLIs. Study the current Foundry experience, not screenshots from older courses.
What was dropped
Comparing the two outlines, several AI-900 topics are no longer listed in AI-901:
- Regression, classification and clustering as named machine learning techniques
- Features and labels, and training versus validation datasets
- Automated machine learning and Azure Machine Learning capabilities
- Face detection and facial analysis as a separate service objective
Microsoft says the bullets under each skill illustrate the assessment and that related topics may still appear, so a basic grasp of machine learning terms is still worth having. It simply no longer has its own section.
Many free AI-900 question banks are still online. Around half of their content tests topics AI-901 no longer lists, and none of it tests agents, prompts or the Foundry SDK. Using them as your main practice will leave large gaps.
A study plan for AI-901
This four week plan assumes about six hours a week and basic Python. If you have never written Python, add a week first to learn variables, functions, imports and how to call a library.
- Week 1: concepts. Responsible AI principles, how generative models work, model choice, deployment types and parameters, and the common AI workloads.
- Week 2: generative AI and agents. Set up access to Microsoft Foundry, deploy a model, practise system and user prompts, build a small chat client with the Foundry SDK, then create and test a single agent.
- Week 3: text, speech, vision and extraction. Build a small text analysis app, try spoken prompts and Azure Speech, send images to a multimodal model, generate an image, and extract fields from a document with Content Understanding.
- Week 4: practice and review. Take Microsoft's free practice assessment, which now runs on AI Skills Navigator and needs a sign in, then sit timed mocks and review every answer.
Microsoft's own course for the exam is AI-901T00, Introduction to AI in Azure. Use it alongside the official study guide. When you reach week four, MockDen's AI-901 practice and mocks follow the Microsoft format, including Yes/No statement sets and short code reading items.
Try the new format
The best way to feel the difference between AI-900 and AI-901 is to answer some questions. Try the free AI-901 practice on MockDen, see which sub-areas need work, and build your plan from there. MockDen is an independent study site and is not affiliated with Microsoft.
Last checked against official sources on 19 November 2026.
Frequently asked questions
Has AI-900 been retired?
Yes. Microsoft's AI-900 study guide states the exam retired on 30 June 2026 at 11:59 PM Central Standard Time. AI-901 is now the only exam for the certification.
Did the certification name change?
No. It is still Microsoft Certified: Azure AI Fundamentals. Only the exam behind it changed.
I already passed AI-900. Do I need to do anything?
No. Microsoft says existing holders need take no action, and Fundamentals certifications do not expire. You may take AI-901 if you want to show newer skills.
Do I need to code for AI-901?
You need basic Python. Microsoft's audience profile asks for knowledge of Python coding syntax and programming techniques, plus familiarity with Azure resources.
What is the passing score for AI-901?
700 on a scaled score from 1 to 1,000, the same as other Microsoft technical exams.
- Microsoft Learn, Study guide for Exam AI-901
- Microsoft Learn, Study guide for Exam AI-900 (retired)
- Microsoft Learn, Exam AI-901
- Microsoft Tech Community, Evolving the Microsoft Certified: Azure AI Fundamentals Certification
- Microsoft Learn, Exam duration and exam experience
MockDen is independent and not affiliated with the exam owners. Always confirm fees and dates on the official page before booking.