Microsoft Certified: Azure AI Fundamentals (AI‑900) Practice Exams
About the Azure AI-900 exam
Exam at a glance
Microsoft's entry-point AI credential at the fundamentals tier. Unlike Microsoft's role-based associate and expert certifications, Fundamentals certifications do not expire — once you pass, it's permanent.
AI-900 → AI-901 transition (2026)
Microsoft is replacing AI-900 with AI-901 on 30 June 2026. The certification (Microsoft Certified: Azure AI Fundamentals) is unchanged — only the exam version transitions. AI-901 is currently in beta and already accepts registrations. Both exams cover the same five skill areas (AI workloads + considerations, ML principles, computer vision, NLP, generative AI), with AI-901 adding deeper coverage of Microsoft Foundry implementation. Anyone holding an AI-900 pass keeps their certification; only new candidates need to switch to AI-901 after the retirement date.
Who this exam is for
AI-900 covers AI/ML concepts and the Azure AI services that implement them. It's conceptual rather than hands-on, which makes it a strong fit for business analysts, sales engineers, project managers, technical pre-sales, and IT generalists adding AI knowledge to their toolkit. Microsoft explicitly designs the exam for candidates from both technical and non-technical backgrounds — no prior data-science or software-engineering experience is required.
Domain weighting
- Describe AI workloads and considerations: 15–20%
- Describe fundamental principles of machine learning on Azure: 20–25%
- Describe features of computer vision workloads on Azure: 15–20%
- Describe features of Natural Language Processing (NLP) workloads on Azure: 15–20%
- Describe features of generative AI workloads on Azure: 20–25%
Prerequisites
None. Familiarity with basic cloud concepts and client-server applications is helpful but not mandatory. Microsoft positions AI-900 as a starting point — many candidates take it before AZ-900 (Azure Fundamentals), and you do not need AZ-900 first.
Why take this certification
- Permanent credential, no renewal. Microsoft's Fundamentals-tier certifications never expire — you earn it once and it stays on your transcript forever. That's a meaningfully better deal than role-based Microsoft, AWS, and GCP certifications that require renewal every 1–3 years.
- AI literacy signal for non-engineers. AI-900 is the most-cited "I understand AI concepts well enough to have an informed conversation" credential in 2026 hiring. It's specifically valued in pre-sales, product management, and consulting roles where you need to scope AI solutions without necessarily building them.
- Foundation for the AI track. AI-900 maps directly to the next step: AI-102 Azure AI Engineer Associate, which validates hands-on implementation of Azure AI services and Azure OpenAI.
- Generative AI coverage. The April 2026 refresh added a full domain on generative AI workloads (LLMs, prompt design, RAG, Azure OpenAI Service, Azure AI Foundry) — making AI-900 one of the few entry-level certifications that genuinely reflects the 2024–2026 AI landscape.
Important: AI-900 retiring June 30, 2026
Microsoft is retiring exam AI-900 on June 30, 2026 and replacing it with AI-901. Both exams lead to the same Microsoft Certified: Azure AI Fundamentals credential. Skills measured are nearly identical; AI-901 adds emphasis on implementing AI solutions with Microsoft Foundry. If you have time before June 30, AI-900 is fine; otherwise plan for AI-901. Candidates who pass AI-900 keep the certification permanently.
What you'll learn in the AI-900 exam
AI-900 is concept-heavy and service-aware. You won't write code, but you will need to know what each Azure AI service does, when to choose it over alternatives, and how Microsoft frames responsible AI.
AI workload categories
- Machine Learning — supervised vs unsupervised vs reinforcement; classification, regression, clustering.
- Computer vision — image classification, object detection, OCR, semantic segmentation, face detection.
- Natural Language Processing (NLP) — sentiment analysis, key phrase extraction, entity recognition, translation, speech-to-text and text-to-speech.
- Generative AI — large language models, prompts, completions, embeddings, retrieval-augmented generation (RAG).
- Knowledge mining — extracting structured insights from unstructured content via Azure AI Search.
- Document intelligence — extracting fields from forms, invoices, and receipts via Azure AI Document Intelligence.
Responsible AI principles
Microsoft's six Responsible AI principles are tested directly — expect questions that give you a scenario and ask which principle applies:
- Fairness — AI systems treat all people equitably.
- Reliability and safety — AI systems perform reliably and safely.
- Privacy and security — AI systems are secure and respect privacy.
- Inclusiveness — AI systems empower everyone and engage people.
- Transparency — AI systems are understandable.
- Accountability — people are accountable for AI systems.
Azure AI services you'll be tested on
- Azure Machine Learning — workspace, designer, automated ML, compute targets, model deployment basics.
- Computer vision — Azure AI Vision (image analysis, OCR), Custom Vision (image classification, object detection), Face service.
- NLP services — Azure AI Language (sentiment, entities, question answering, conversational language understanding), Translator, Speech (speech-to-text, text-to-speech, translation).
- Generative AI — Azure OpenAI Service, Azure AI Foundry (formerly Azure AI Studio), prompt engineering basics, when to use RAG vs fine-tuning.
- Document intelligence — prebuilt vs custom models, form recognizer basics.
Foundation-model concepts
The generative AI domain assumes you understand: what a large language model is, how tokens and context windows work conceptually, embeddings and vector search, RAG architecture, prompt engineering basics, and the trade-offs between prompt design, RAG, and fine-tuning.
How the practice exams help
Each free question and every premium exam mirrors Microsoft's scenario-style format — a short stem describing a business case, four to six plausible options, one or two correct. Detailed explanations cover not just why the right answer is right but why the distractors are wrong, so you learn the service trade-offs rather than memorizing service names.
How to prepare for the AI-900 exam
AI-900 is one of the shorter prep curves in the Microsoft catalog. Most IT generalists pass with 2–4 weeks of focused study. Recommended approach:
- Walk the official learning path (1–2 weeks). Microsoft Learn hosts the free AI-900 learning path covering all five domains. The modules are short (15–30 min each) and include knowledge checks. This is the single highest-value resource — the questions on the real exam are written to match the framing used in these modules.
- Hands-on with the free tier (1 week). Create a free Azure account and spend a few hours in Azure AI Foundry, Azure AI Vision, and Azure AI Language. Run an image through Vision, send a sentence through Translator, and try a prompt or two against Azure OpenAI. The point is recognition — you should see a service name on the exam and immediately picture the portal blade, not just the brand.
- Memorize the six Responsible AI principles. Three to five questions on every sitting test these directly. The principles are short, distinct, and reliably worth points — there's no excuse for missing them.
- Practice exams (3–5 days). Take timed practice tests to identify weak areas. Detailed explanations on every answer option help you learn the reasoning, not just memorize answers. Aim for consistent 85%+ scores before scheduling your exam — Fundamentals exams reward thoroughness over depth.
Recommended timeline
2–3 weeks for candidates with some IT background. Complete beginners with no cloud or AI exposure should allow 4–6 weeks. Most candidates underestimate the generative AI domain — budget extra time there if the LLM landscape is new to you.
Official resources
Start with the official AI-900 exam page and the linked study guide. Microsoft Learn also publishes a free practice assessment that closely mirrors the question format on the real exam. For deeper context, the Azure AI services documentation is the canonical reference for every service in scope.