After AI-900: The New Azure AI Fundamentals Path
Microsoft retired Exam AI-900 on June 30, 2026, but it did not retire the Azure AI Fundamentals certification itself. The credential now uses Exam AI-901, which reflects a broader shift in Microsoft’s AI platform toward Microsoft Foundry, generative AI, agents, and simple implementation tasks.
For learners looking at the Microsoft Certified: Azure AI Fundamentals path today, the important distinction is between the old exam and the continuing certification. Existing AI-900 holders do not lose the certification because the exam retired. New candidates should prepare for AI-901 rather than using an old AI-900 study plan as though nothing changed.
The transition also changes who the path best serves. AI-900 was friendly to both technical and nontechnical beginners and emphasized conceptual recognition. AI-901 still begins with fundamentals, but Microsoft now expects foundational technical skills, familiarity with Azure resources, and basic Python concepts because the exam includes implementing AI solutions with Foundry.
AI-900 is historical content now, not a current exam choice
AI-900 remains useful as a record of the previous Azure AI Fundamentals scope. Its content covered AI workloads and considerations, machine learning principles, computer vision, natural language processing, and generative AI in a largely conceptual way. Those ideas did not become irrelevant when the exam retired.
However, candidates should not treat older AI-900 fundamentals as a complete current plan. It can reinforce terminology and foundational concepts, but the active exam measures a different balance of knowledge and implementation. Any study resource should be checked against the AI-901 objectives before it is used as the primary guide.
Microsoft’s own transition messaging identifies AI-901 as the replacement exam, which makes the status distinction straightforward for candidates studying after June 30, 2026. Training providers, employers, and internal learning portals may take longer to update their labels, so the Microsoft exam page should be the final authority when an older AI-900 reference conflicts with current scheduling.
The certification continues under AI-901
Microsoft kept the Azure AI Fundamentals certification and changed the exam used to earn it. That matters because it avoids the common misunderstanding that the entire credential disappeared on June 30. The learning path is still a fundamentals credential; the validation method was refreshed to match Microsoft’s current AI platform.
The broader Azure AI Fundamentals value proposition also remains recognizable: it is an entry point for people who need a structured understanding of AI in Azure. What changed is the expectation that beginners should be able to connect concepts to lightweight implementation rather than only identify service categories.
The certification continuity also matters for employers and learning programs. A job description or internal training plan may still say AI-900 even though the active exam is AI-901. Candidates should verify whether the requirement is truly the retired exam code or the Azure AI Fundamentals certification. In most cases, the latter is the durable credential name that should guide an updated plan.
AI-901 is aimed at technical beginners
Microsoft describes AI-901 candidates as people at the beginning of a career in AI solution development. The exam still expects conceptual knowledge, but it also calls for basic Python syntax and programming techniques, familiarity with Azure resources, and awareness of REST APIs, SDKs, and command-line interfaces.
That does not make AI-901 an advanced developer exam. The implementation tasks are foundational. The shift simply means that learners should be comfortable moving from “what does this capability do?” to “how would I create, configure, or use a simple AI solution with the current platform?”
Nontechnical learners can still benefit from AI-901, but they may need a short preparation bridge before starting exam-specific labs. Basic Python variables, functions, API concepts, JSON, authentication, and Azure resource navigation can make the platform exercises much easier. Building that foundation first is more efficient than repeatedly getting blocked by unfamiliar development mechanics.
Microsoft Foundry is now central to the learning path
The most visible change is the emphasis on Microsoft Foundry. Instead of organizing the fundamentals experience mainly around separate Azure AI services, AI-901 brings models, generative AI applications, agents, and other workloads into a more unified development environment.
Learners who need background can still benefit from a broad Microsoft Azure AI, but they should then practice the current Foundry workflow. The exam expects familiarity with provisioning resources, selecting models or capabilities, configuring solutions, and understanding how basic application components fit together.
Foundry also changes how learners should organize their notes. Instead of creating separate silos for every Azure AI service, it is useful to map workloads to common solution steps: choose a capability or model, provision resources, supply data or prompts, configure safety and evaluation, call the solution from an application, and monitor the result. That structure better reflects modern AI development.
Machine learning fundamentals still matter
Classification, regression, clustering, features, labels, training, validation, and model evaluation remain useful foundations because modern generative AI does not eliminate conventional machine learning. AI-901 reorganizes the exam, but candidates still need to understand what a model is doing and how different AI workloads differ.
A concise machine learning fundamentals can therefore remain relevant during the transition. The key is to avoid spending the entire study plan on classic algorithms. AI-901 gives much more weight to implementing current AI solutions, so machine learning theory should support—not crowd out—hands-on familiarity with Foundry.
The machine learning portion should also be studied as a decision framework. Candidates should know when a prediction problem looks like classification or regression, when grouping resembles clustering, and what training data contributes to a model. AI-901 does not require advanced mathematics, but practical solution work is easier when the learner can recognize the shape of the problem before choosing a tool.
Generative AI and agents are no longer optional side topics
AI-900 increased its generative AI coverage late in its lifecycle, but AI-901 makes modern generative applications and agents central to the path. Candidates should understand prompts, model behavior, grounding or context, multimodal capabilities, and the idea that an agent can combine a model with instructions, tools, and actions.
Foundational resources on generative AI and large language models and intelligent agents can help with the vocabulary, but exam preparation should connect those concepts to the Microsoft platform. The goal is not to become an agent-framework specialist; it is to understand what the components do and how a simple solution is assembled.
Agents also introduce a new security and governance dimension for beginners. A model that only generates text has limited direct authority, while an agent can be connected to tools that read data or perform actions. Even at a fundamentals level, candidates should understand that tool permissions, identity, human approval, and auditability affect whether an agent is safe to use.
Responsible AI remains a core foundation
Fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability continue to matter because practical AI development introduces real consequences. AI-901 keeps responsible AI in the fundamentals curriculum instead of treating it as an advanced governance topic.
Candidates should be able to recognize how design choices affect users and data. A technically functional model can still be unacceptable if it exposes sensitive information, systematically disadvantages a group, cannot explain important limitations, or operates without clear ownership. Responsible AI is part of solution quality, not a separate policy document.
Existing AI-900 knowledge can shorten the transition—but not replace it
Someone who previously prepared for AI-900 already understands many recurring concepts: AI workloads, common machine learning patterns, vision, language, generative AI, and responsible AI. That foundation reduces the amount of new theory required.
The study gap is mainly the platform and implementation emphasis. Candidates should compare their old notes with the AI-901 guide, then focus on Foundry, model use, prompts, generative applications, agents, and the lightweight coding or resource tasks the new exam expects. Re-reading AI-900 material from the beginning is less efficient than identifying what changed.
A useful transition exercise is to take the AI-901 skills outline and mark every objective as familiar, changed, or new. Candidates with AI-900 experience usually find responsible AI and workload concepts familiar, while Foundry implementation, agents, and technical tooling require the most new practice. That gap analysis prevents overstudying old material simply because it feels comfortable.
The new fundamentals path is closer to how AI work now begins
Modern AI adoption often starts with selecting an existing model, connecting data or context, designing prompts, adding safety controls, and integrating the result into an application. AI-901 reflects that reality more directly than an exam centered on recognizing a list of separate AI services.
For learners entering Microsoft certifications, that makes Azure AI Fundamentals a more practical bridge into later AI roles. The credential still establishes vocabulary and responsible-AI foundations, but the active exam now expects candidates to touch the tools used to build simple solutions. The path after AI-900 is therefore not disappearance; it is a more implementation-oriented version of the same fundamentals credential.
After earning the fundamentals credential, the next step should match the role rather than follow a fixed ladder. Developers may move toward Azure AI application and agent development, data professionals may deepen machine learning or data engineering, and security practitioners may focus on securing AI services and data. The fundamentals certification is most useful as a common vocabulary that supports those different directions.