Microsoft Azure AI Fundamentals (AI-900)
A conceptual certificate in machine learning and Azure AI services, useful before AI-102 or as literacy for non-specialists.
- Exam code
- AI-900
- Fee
- US$99 (check the official page)
- Registration
- Official registration page (opens in a new tab)
Fees and exam codes change. We link to the provider instead of copying prices that may be out of date.
AI-900 covers what machine learning is, which Azure services implement it, and what responsible AI means in practice. It is a literacy exam, not a modelling one.
Who it is for
Product managers, analysts and developers who need to discuss AI features without building models, and anyone preparing for the AI Engineer Associate (AI-102). No coding or maths prerequisite.
What it covers
- AI workloads — the categories: vision, language, speech, decision support and generative AI.
- Machine learning on Azure — regression, classification, clustering, and the difference between training and inference.
- Computer vision and NLP — image classification, object detection, OCR, sentiment, translation, entity extraction.
- Generative AI — what a language model does, prompts, grounding, and the Azure services that host models.
- Responsible AI — fairness, reliability, privacy, inclusiveness, transparency and accountability.
Format
Around 40 to 50 questions in 45 minutes, multiple choice, with a pass mark of 700 out of 1000. No expiry.
How to prepare
- Follow the Microsoft Learn AI-900 path, which includes hands-on sandboxes.
- Try Azure AI services in the portal: a vision call, a language call, a speech call. Seeing an API response makes the service descriptions concrete.
- Learn the vocabulary precisely; many questions turn on the difference between classification and clustering, or between training and inference.
- Take the practice assessment and review every explanation.
A week or two of study is realistic.
Is it worth it
As a signal of AI literacy for a non-specialist role, yes: it is inexpensive, non-expiring, and gives a shared vocabulary. As evidence that you can build AI systems, no — that is AI-102 or a portfolio. Be honest in an interview about which one you have.