DP-900 is one of the Fundamentals exams Microsoft kept when it cleared out much of the old catalogue in mid 2026. AI-900 is gone, AZ-204 is gone, and DP-900 is still here with a skills outline refreshed on 21 July 2026. The exam is 45 minutes, proctored, needs 700 to pass, costs $99 in the US and does not expire. Retakes follow the standard Microsoft policy of a 24 hour wait after the first fail. The skills outline has four domains.
Core data concepts are 25 to 30%: structured, semi-structured and unstructured data, file formats, transactional against analytical workloads, and the responsibilities of database administrators, data engineers and data analysts. Relational data on Azure is 20 to 25%, covering normalisation, common SQL statements and database objects, and the Azure SQL family plus the open-source database services. Non-relational data is 15 to 20%, mostly Blob storage, Azure Files, Table storage and Cosmos DB APIs. Analytics workloads are 25 to 30%, with ingestion, analytical stores, real-time analytics and Power BI, and the July update now names Azure Databricks and Microsoft Fabric as the large-scale analytics services. That last point matters more than it sounds.
Old DP-900 courses lean heavily on Synapse Analytics, and Microsoft's own direction is now Fabric. If you are using a prep course from 2023 or 2024, expect the analytics section to be out of date. What I do not like is how little of the exam asks you to think. A large share of the questions boil down to picking which Azure service fits a scenario, and the SQL section asks you to recognise a SELECT or a JOIN rather than write one. You can pass this exam having never queried a database, and that makes it weak evidence for a hiring manager. It is also worth being clear that there is nothing about AI here.
No vector search, no Azure AI services, no mention of Copilot in Fabric or Power BI. People often take DP-900 as the first step towards an AI career, and on its own it does not move you towards that at all. It teaches you what a data warehouse is, which is useful, but the AI-relevant material lives in AI-901 and AI-103 now. Where it does make sense: a career changer who wants a cheap, structured map of data concepts before tackling something harder, or someone at a Microsoft partner who needs Fundamentals headcount. For anyone who already works with data, the time is better spent on DP-600 or PL-300, which test things you actually do.