AI-900 and DP-900 were introduced as complementary Microsoft Azure fundamentals examinations, one centered on artificial intelligence and the other on data services. Their status is no longer symmetrical. Microsoft retired AI-900 on June 30, 2026, while DP-900 remains an active Azure Data Fundamentals examination as of October 2026. Microsoft now offers AI-901 in the AI fundamentals path with updated content. An older guide describing both AI-900 and DP-900 as newly launched, bookable credentials is historically interesting but unsuitable for a current candidate’s scheduling decision.
The subject matter still overlaps in useful ways. Data fundamentals help engineers understand relational stores, documents, analytical workloads and data movement; AI fundamentals introduces model use, responsible AI and the services needed to build AI-enabled solutions. The exam codes matter when choosing an appointment, but the transferable learning objective is recognizing what each workload does and which platform capabilities fit it. This guide preserves the original comparison while distinguishing historical AI-900 learning from active DP-900 and current AI-901 study.
What AI-900 Measured Historically
The retired AI-900 exam introduced AI workloads and responsible AI concepts, machine learning foundations, computer vision, natural language processing and conversational capabilities as represented in Azure at the time. The service catalog and exam weights evolved across versions. This knowledge can still help someone understand a classification problem or an image-processing scenario, but the original examination itself is no longer available. Microsoft’s retired-exams list is authoritative on the June 30 retirement date. An existing earned credential can remain on the holder’s transcript under Microsoft’s program rules without implying a new candidate can sit AI-900.
Using a historical AI-900 workbook to learn terminology is acceptable if its limitations are made clear. Some older service branding, interfaces and functionality have changed, and new material incorporates generative AI and Microsoft Foundry concepts. Do not label obsolete skill percentages as the current AI fundamentals blueprint. A current AI-901 exam outline is the more appropriate reference when someone plans a new certification attempt, accompanied by hands-on exercises that use maintained service documentation rather than a discontinued UI tutorial.
What AI-901 Changes for New Candidates
AI-901 is Microsoft’s successor for the contemporary Azure AI Fundamentals credential path, with updated emphasis aligned to current Azure AI capabilities and development workflows. The issuer’s examination overview expects foundational understanding of AI solution design and the ability to recognize appropriate services and responsible deployment choices. Candidates should use the latest Microsoft Learn guide to confirm exact objective weighting and available appointment details. AI-901 is not automatically the same test as AI-900 with only a different number; its new scope needs to be reviewed directly.
A practical AI fundamentals exercise can compare a deterministic rule-based system with a model-based classification approach, then examine limitations such as data quality, bias, confidence and sensitive-input handling. Another exercise might evaluate an AI service’s ability to summarize or categorize synthetic text while recording errors and human review requirements. A model producing plausible answers is not proof that it is appropriate for a regulated use case. The learner should be able to explain privacy, evaluation and monitoring before suggesting automatic deployment.
DP-900 Remains the Data Fundamentals Examination
DP-900 assesses foundational cloud data concepts and related Azure data services. Microsoft’s active certification page describes relational and non-relational data and the distinction between transactional and analytical workloads. It is intended for candidates beginning to work with data in the cloud. The credential is not required as a prerequisite for every later data engineering or database administration certification. A learner should focus on the practical concepts and use service names as examples, not as a substitute for understanding the data model or workload.
A relational database organizes records under schemas and supports query patterns such as filtering, joins and transactional consistency. A document database stores flexible records under a different modeling approach. A data warehouse or lake-based analytics system focuses on analyzing collected data at scale rather than serving each customer checkout in milliseconds. Understanding these distinctions helps determine why a service may be suitable for business intelligence but not for a high-volume order-processing transaction. DP-900 data fundamentals preparation should develop these decisions using the issuer’s current examples.
Compare Data Workloads With AI Workloads
A data pipeline might ingest inventory records, validate them and prepare a reliable dataset for reporting. An AI application could then use that dataset to produce predictions or support a search experience. The foundational skills connect, but the systems have different correctness criteria. A data process can be verified through schema, completeness and expected aggregates; an AI model may also require evaluation of accuracy, bias, drift and unsafe output. A high-quality architecture documents where the authoritative data resides and how downstream AI services are permitted to access it.
Data quality and privacy matter to both disciplines. A machine learning service cannot correct missing business context simply because its infrastructure scales. Likewise, a well-structured relational table does not make a model’s prediction trustworthy. Ensure that training or inference inputs respect classification and permissions, and consider the consequences of incorrect output. Fundamentals certification should lead to more careful engineering conversations rather than promises that all data becomes valuable once it is sent to the cloud.
Choose the Appropriate Learning Path
A person planning to support reporting, databases or analytics may begin with active DP-900 and progress toward hands-on work involving SQL queries, data modeling, validation and pipeline monitoring. Someone working toward AI application development may start with the updated AI-901 fundamentals objectives and then move to service-specific engineering. Do not assume either fundamentals credential automatically qualifies its holder for a specialist job. A portfolio containing a small data project or carefully evaluated AI prototype adds practical evidence to the examination result.
A candidate who previously earned AI-900 does not need to rewrite history and call that older credential AI-901. They may still learn modern content from AI-901, but the precise label and issue date belong to the original certification transcript. Conversely, someone new to the field must not book the retired AI-900 because a third-party course still describes it as available. Check the official scheduling portal for current offerings and use the actual exam code when buying preparation materials. Avoid promotional guarantees of salary increases or job placement.
Design a Short Hands-On Study Sequence
Begin with a sample business problem, such as analyzing synthetic store orders. Model it as relational data, write simple queries and compare the results with a document-oriented representation. Identify whether the task is transactional reporting or large-scale analytics. Then create a separate AI example, such as evaluating text classification on non-sensitive test data, and record where the model was correct or uncertain. These exercises demonstrate why the data foundation and the AI capability solve related but different problems.
For DP-900, practice explaining the difference between a normalized transactional database, an analytical model and a streaming ingestion system. For AI-901, focus on identifying appropriate AI workloads, responsible use and relevant current platform services. Use official Microsoft Learn study guides and work within a controlled subscription, reviewing possible costs and deleting resources after the lab. A strong candidate should be able to describe what the system does and how its errors would be detected, not merely recite a catalogue of Azure product names.
Check How the Fundamentals Fit Later Specializations
AI and data credentials should be evaluated against later work, not only against which examination is easier. A future database administrator benefits from understanding relational design, transactional consistency and access controls before deepening platform-specific operations. A data engineer needs pipeline reliability, data validation and analytical modeling. An AI application developer must add evaluation, privacy and operational monitoring to basic model selection. None of these professional responsibilities is established automatically by passing a fundamentals exam. Learners can use a small portfolio and the current role-based certification map to identify gaps rather than trying to complete every foundational badge.
A useful comparison exercise begins with synthetic support tickets and transactional order records. First design a relational or non-relational data store and define access, retention and quality rules. Then consider whether an AI service should classify ticket topics or summarize records. Verify sample outputs against known labels and document cases where the model is wrong. Restrict which records can reach the service and protect customer identifiers. The data system has a source-of-truth and integrity problem; the AI component has an additional probabilistic evaluation problem. Understanding that difference is more important than memorizing a feature list from an obsolete AI-900 syllabus.
For scheduling, distinguish old course material from the actual issuer credential. A retired AI-900 practice test can be used for historical background but cannot confirm readiness for AI-901, whose current objectives should be read directly. An active DP-900 study guide should also be checked for the latest revision rather than assumed unchanged from its launch. Record the exam code and intended date before choosing paid resources, and remember that practical cloud service costs can arise during labs regardless of whether training content is freely accessible.
The accurate October 2026 comparison is therefore straightforward: AI-900 is retired, DP-900 is active, and the current AI fundamentals examination is AI-901. Historical AI-900 material remains useful when labeled and updated for service changes, while DP-900 continues to teach enduring data concepts. Choose the exam that aligns with the intended technical work, validate the current issuer requirements and supplement the fundamentals with practice projects. The value lies in understanding data and AI workloads rather than treating an expired exam code as a permanent career route.