Microsoft’s Azure AI Fundamentals certification now requires exam AI-901. The old exam, AI-900, retired on 30 June 2026. The big change is that AI-901 puts 55-60% of its weight on actually building simple AI apps and agents with Microsoft Foundry, so it is more hands-on than AI-900 ever was.
If you already passed AI-900, you keep your certification and do not need to retake anything. If you are starting now, avoid AI-900 study material and follow the AI-901 study guide, which lists skills current as of 15 April 2026.
Key facts
| Item | Detail |
|---|---|
| Certification | Microsoft Certified: Azure AI Fundamentals |
| Exam | AI-901 (replaced AI-900, which retired on 30 June 2026) |
| Passing score | 700 or above |
| Prerequisites | None |
| Price | Set by country; check the exam page when you schedule |
| Languages | English plus Arabic, Chinese (Simplified and Traditional), French, German, Indonesian, Italian, Japanese, Korean, Portuguese (Brazil), Russian and Spanish |
| Expiry | Fundamentals certifications do not expire and need no renewal |
| Practice | Free official practice assessment on Microsoft’s AI Skills Navigator |
Microsoft notes that localised versions are usually updated about eight weeks after the English version, so if you take the exam in another language soon after a skills update, the questions may lag slightly. If the exam is not available in your preferred language you can request 30 extra minutes.
AI-900 vs AI-901: what actually changed
| AI-900 (retired) | AI-901 (current) | |
|---|---|---|
| Style | Mostly concepts: recognise workloads and match them to Azure services | Concepts plus implementation in Microsoft Foundry |
| Machine learning | Dedicated section on core ML principles | No separate ML section; focus moves to generative and agentic AI |
| Generative AI | One area among several | Central: prompts, model choice, deployment, agents |
| Technical depth | No coding expected | Familiarity with REST APIs, SDKs and CLIs expected |
That last row matters. Earlier versions of this certification were popular with project managers, pre-sales staff and business analysts because no technical background was needed. AI-901 still has no prerequisites, but you will be asked about building a chat client with the Foundry SDK and creating a single-agent solution. Non-developers can still pass, but they should budget extra time for the labs.
Skills measured
1. Identify AI concepts and capabilities (40-45%)
- Responsible AI principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability
- How generative AI models work, choosing a model by capability, and deployment options and configuration parameters
- AI workloads: generative and agentic AI, text analysis (keyword extraction, entity detection, sentiment, summarisation), speech recognition and synthesis, computer vision and image generation, and information extraction from text, images, audio and video
2. Implement AI solutions by using Microsoft Foundry (55-60%)
- Generative AI apps and agents: writing system and user prompts, deploying a model in the Foundry portal, building a lightweight chat client with the Foundry SDK, and creating and testing a single-agent solution
- Text and speech: a text analysis app, responding to spoken prompts with a multimodal model, and an app using Azure Speech in Foundry Tools
- Computer vision and image generation: interpreting images in prompts and creating visual outputs
- Information extraction: pulling data from documents, forms, images, audio and video with Azure Content Understanding in Foundry Tools
Most questions cover generally available features, though commonly used preview features can appear.
How to think about scenario questions
Many questions describe a business need and ask which capability or setting fits. For example, a bank wants to pull account numbers and dates from scanned loan forms: that is information extraction, not a chat model. A support team wants a bot that answers in a consistent, polite tone: that is a system prompt decision. A retailer wants to flag negative product reviews: that is sentiment analysis. Practise classifying everyday situations at work this way until the mapping feels automatic.
Who should take it, and who should not
Good fit: students and freshers wanting a recognised first AI credential; developers, testers and analysts starting to add AI features; and consultants or pre-sales staff on Azure projects who are willing to do the labs.
Weaker fit: experienced Azure developers, who will get more from an associate-level exam (see our guide to AI-200, the new Azure developer certification), and people who work mainly on AWS or Google Cloud. Our comparison of AI and ML certifications covers the alternatives.
If you already passed AI-900
Your Azure AI Fundamentals certification stays on your Microsoft Learn transcript and does not expire. There is no upgrade exam. If you want to show current skills, an associate-level AI or developer certification will add more than resitting a fundamentals exam.
A four-week study plan
This assumes about five to six hours a week alongside a full-time job.
| Week | Focus | Hands-on |
|---|---|---|
| 1 | Responsible AI, AI workloads, how generative models work | Work through the free AI-901 learning path on Microsoft Learn; write your own one-line definition of each workload |
| 2 | Foundry basics: projects, model catalogue, deployment, prompts | Deploy a model in the Foundry portal and test different system prompts and parameters |
| 3 | SDK chat client, single agent, speech and vision | Build a small chat client with the Foundry SDK, then a single-agent solution; try a multimodal prompt with an image |
| 4 | Information extraction and revision | Extract fields from a sample invoice with Content Understanding; take the official practice assessment and revise gaps |
Keeping costs and risks down
- Use a free or low-cost Azure subscription for labs and delete resource groups after every session. Deployed models and AI services can keep charging while idle.
- Set a budget alert in Azure Cost Management before you start.
- Register for the exam with a personal Microsoft account, not your work account. Microsoft warns that records tied to an organisational account can be lost when you leave the organisation.
- Ignore AI-900 practice tests and “updated” question dumps. They test a different skills outline, and memorised questions breach exam rules.
FAQ
Can I still take AI-900?
No. It retired on 30 June 2026. AI-901 is the only route to Azure AI Fundamentals now.
How long does preparation take?
Beginners studying part-time typically need three to five weeks. Developers already using Azure OpenAI or Foundry may need one to two.
Last reviewed: 8 October 2026.
Sources
- Microsoft Learn: Microsoft Certified: Azure AI Fundamentals
- Microsoft Learn: Study guide for Exam AI-901
- Microsoft Q&A: AI-900 retirement and AI-901

Leave a Reply