Is DP-203 Retired? Yes, and DP-700 Is the Fabric Exam That Replaced It

2026/07/13

Click to upload or drag and drop

PDF, DOCX, PPTX, TXT, JPG, JPEG, PNG, HEIC, ODP, ODT, BMP, or TIFF

up to 20MB

Please wait, your quiz is being created...

Uploading...

Yes. DP-203 is retired. Azure Data Engineer Associate (Exam DP-203) retired on March 31, 2025, and its certification page now carries a retirement warning. The exam that replaced it is DP-700: Implementing Data Engineering Solutions Using Microsoft Fabric, which earns the Microsoft Certified: Fabric Data Engineer Associate certification.

The name change tells you everything. Microsoft did not refresh its data engineering exam. It moved data engineering onto Fabric, and quietly deleted the platform the old exam was built around.

Azure Synapse has been erased from Microsoft's data certifications

This is the finding that matters, and you can verify it yourself in about two minutes. We counted the term "Azure Synapse" in the skills-measured section of the current official outlines:

ExamStatus"Azure Synapse" mentions
DP-203 (Azure Data Engineer Associate)Retired March 31, 2025The exam was built on it
DP-900 (Azure Data Fundamentals)Live0
DP-700 (Fabric Data Engineer Associate)Live0

Zero, on both live exams. DP-900's large-scale analytics bullet used to be a Synapse bullet; it now names only "Azure Databricks and Microsoft Fabric." A product that was the centerpiece of Microsoft's data platform certification track is no longer named once in the fundamentals exam or in the associate exam that replaced its own.

If you are studying from any data engineering course recorded before 2025, you are studying Synapse pipelines and dedicated SQL pools for an exam that does not mention them.

What DP-700 actually tests

Three domains, and here is the unusual part: they are all weighted 30 to 35 percent. There is no dominant domain and no domain you can safely under-prepare. It is the most evenly balanced Microsoft exam we have measured.

DomainWeight
Implement and manage an analytics solution30 to 35%
Ingest and transform data30 to 35%
Monitor and optimize an analytics solution30 to 35%

The vocabulary is entirely Fabric: pipelines (6 mentions), Spark (4), notebooks (4), eventstreams (3), KQL (3), lakehouse. The skills section is only about 370 words, which is short, and short outlines are deceptive. Fewer words does not mean less exam, it means each bullet covers more ground.

DP-700's skills are stamped as of July 21, 2026, so if you are reading an outline dated earlier than that, it is stale.

DP-203 vs DP-700, term by term

The cleanest way to see what happened is to count the same words in both official outlines. DP-203's skills section runs 615 words; DP-700's runs 370. Here is what each one names.

Term in the official outlineDP-203 (retired)DP-700 (live)
Azure Synapse80
Fabric04
Lakehouse01
Eventstream03
KQL03
Stream Analytics30
Dedicated SQL pool10
Spark54
Pipeline176

It is a clean inversion. Every Synapse-era term (Synapse itself, Stream Analytics, dedicated SQL pools) went to zero. Every Fabric-era term (Fabric, lakehouse, eventstream, KQL) came from zero. Only Spark survived the transition roughly intact, because Spark is Spark wherever Microsoft hosts it.

Spark is, in fact, the practical answer to "what do I keep?" If you spent the DP-203 era writing Spark transformations, that knowledge transfers directly. If you spent it tuning dedicated SQL pools and wiring Stream Analytics jobs, that is the part of your preparation the new exam has no interest in.

The weighting changed shape too

DP-203 had a dominant domain: "Develop data processing" was 40 to 45 percent, nearly half the paper, and you could pass by being very good at pipelines and merely adequate elsewhere. Its other two domains were storage design at 15 to 20 percent and security, monitoring and optimization at 30 to 35 percent.

DP-700 flattened that. Three domains, 30 to 35 percent each, no dominant one, no safe one to under-prepare. Notice what got promoted: "Monitor and optimize an analytics solution" is now a full third of the exam, up from being one part of a shared domain. Microsoft has decided that a data engineer who can build a pipeline but cannot tell you why it got slow is not a data engineer.

That is also why the DP-203 tactic of skimming the smallest domain does not work anymore. There is no smallest domain.

Do I have to retake anything if I passed DP-203?

No. A certification you earned stays earned. Microsoft keeps retired certifications on the Active Certifications section of your transcript, and a retired certification does not vanish from your record. What changes is the market: "Azure Data Engineer Associate" now describes a credential nobody can newly obtain, on a platform Microsoft has moved off. Recruiters will start reading it as a date stamp rather than a current skill.

If data engineering is your actual job, DP-700 is the one that says you can do the work in the tool Microsoft is now shipping.

Where does DP-900 fit?

DP-900 (Azure Data Fundamentals) is still live, still $99, confirmed in Microsoft's published pricing feed, and still the sensible on-ramp. It is not retiring, and its objectives refreshed on July 21, 2026 with, once again, every functional group marked "No change" in Microsoft's own change log. No domain renamed, no weight moved.

Its four domains: core data concepts (25 to 30%), analytics workload (25 to 30%), relational data (20 to 25%), non-relational data (15 to 20%). Its skills section is around 331 words, the shortest of any Microsoft exam we have measured, which makes it a fast credential but a thin one.

The honest path in 2026 is DP-900 to prove you know what a data lake is, then DP-700 to prove you can build one in Fabric. The old DP-900 to DP-203 ladder no longer has a top rung.

The rest of the data track moved too

DP-203 is not the only casualty. DP-100 (Azure Data Scientist Associate) retired on June 1, 2025, and its replacement is AI-300, which is not a data science exam at all: the certification is Machine Learning Operations Engineer Associate. The role itself was redefined. Building models gave way to shipping and observing them, and roughly half of AI-300 is generative AI operations, which did not exist on DP-100.

Put the two together and Microsoft's data track has been rebuilt end to end inside about fifteen months: the data engineer moved from Synapse to Fabric, and the data scientist became an MLOps engineer.

How to prepare for a track that keeps moving

The recurring problem with all of this is that the paid question banks lag the exams by a year or more, and on a track that has changed this fast, a year-old bank is not slightly stale, it is testing a retired product. The Synapse count is the proof: you can still buy DP-203 banks today, for an exam that has not existed since March 2025.

The way around it is to stop buying snapshots of somebody else's outline and generate practice questions from current material: the official study guide, Microsoft Learn modules, and your own notes from the tools you actually use. If your pipelines already pull from a dozen systems, the notes you keep while wiring those sources together into one warehouse are better exam material than any bank, because they are about the ingestion problems the exam is genuinely testing.

Upload what you have and build a DP-900 practice test from your own Azure data notes, weighted to the four domains Microsoft publishes today. If you are earlier in the ladder, AZ-900 is the cloud fundamentals starting point and SC-900 the security one, both $99 and both confirmed. For anything else, the certification exam generator turns any source document into practice questions.

Z tej samej rodziny narzędzi