A staged data engineering certification roadmap across AWS, Azure, Google Cloud, Databricks and Snowflake, tied to the skills each milestone should prove.
A surprising number of people planning a data engineering career in 2026 are building their roadmap around an exam that no longer exists. Microsoft's DP-203 — for years the default "Azure Data Engineer" answer in every forum thread — retired on 31 March 2025, along with the Azure Data Engineer Associate certification it awarded. AWS retired its Data Analytics and Database specialties back in 2024. If your plan came from an article written in 2023, it is a map of roads that have since been demolished.
Here is the current, verified picture — and more importantly, a way of sequencing certifications that survives vendor churn: tie each exam to a skills milestone, not to a brand. Short answer: start with a fundamentals-level data exam only if you are new to data work, make a cloud provider's data engineer certification (AWS Data Engineer – Associate, Google Professional Data Engineer, or Microsoft's Fabric-based DP-700) your central credential, and add a platform specialisation — Databricks or Snowflake — once you know which stack your target employers run. This article builds that sequence stage by stage.
Data engineering is the discipline of building and operating the systems that move, transform, store and serve data: ingestion pipelines, batch and streaming transformation, orchestration, warehousing and lakehouse storage, plus the governance, monitoring and cost control around all of it. The people who transition into it most successfully — analysts, backend developers, database administrators — arrive with one part of that picture and use structured study to fill in the rest.
That is the honest role of certifications here. No exam makes you a data engineer; a well-chosen sequence forces you through the full syllabus of one in a sensible order and gives employers legible evidence at each stage. So before choosing badges, name the milestones:
Now map exams onto those milestones.
If you already write SQL daily as an analyst or developer, you can skip this stage entirely; nothing here is a prerequisite for anything later.
If you are earlier in the transition, two low-cost exams do the milestone-one job well:
A realistic scenario: a financial analyst who lives in spreadsheets and dabbles in SQL takes DP-900 in a month of evenings. The certificate itself opens no doors — but the syllabus forces her through concepts (streaming, non-relational storage, warehouse vs lake) she would otherwise meet for the first time in a stage-3 exam, where the stakes and study load are much higher. That is what fundamentals exams are for.
Milestone two — SQL fluency and working Python — is deliberately certification-free in this roadmap. Vendor exams at this level exist, but employer weight sits in what you can build. Spend this stage writing pipelines end to end, however small: pull an API, transform with Python or SQL, load into a warehouse's free tier, schedule it. Two or three of these in a repository will do more in an interview than any stage-1 certificate, and they make stage 3 dramatically easier because cloud data-engineer exams assume you already think in pipelines.
This is the load-bearing certification of the whole path, so choose it by the same rule that governs every provider decision: what do your target employers run? All three options below have no formal prerequisites.
Launched in 2024, effectively replacing the retired Data Analytics – Specialty, and current as of August 2026 with no update announced. It costs $150 (varies by country), runs 65 questions in 130 minutes, needs a 720 scaled score (out of 1,000) to pass, and is valid three years. AWS suggests 2–3 years of general data engineering experience and 1–2 years hands-on with AWS — recommendations, not requirements. As an Associate-tier exam it is also the cheapest of the three cloud data-engineering credentials, and holding it earns AWS's 50% voucher towards a future exam. When your preparation is well advanced, timed runs through AWS Data Engineer Associate practice questions will show you which domains still leak marks.
The longest-established of the three and a Professional-tier credential, which employers read accordingly. It costs $200 plus tax, runs 40–50 questions in two hours (fewer questions than most Google Cloud professional exams), publishes no passing score — results are pass/fail — and is valid for two years, with a shorter renewal exam (20 questions, one hour, $100) now available. Google recommends three-plus years of industry experience including a year on Google Cloud. Note the page currently flags a coming update to reflect Google's product renames, so always study from the current official exam guide. ExamPractice's Google Professional Data Engineer practice questions are useful for testing yourself against the question style once the domains are covered.
Microsoft's current data engineering certification — and the fact that it is Fabric-based, not Synapse-based, is the single most important update in this field. DP-203 retired on 31 March 2025 and cannot be taken or renewed; DP-700 is its successor, with a 100-minute exam covering analytics solution management, data ingestion and transformation, and monitoring/optimisation, using SQL, PySpark and KQL. Passing is 700/1,000 scaled; pricing varies by country (associate tier, commonly cited around US$165). Like all Microsoft role-based certifications it expires after one year but renews free via an open-book online assessment. Choose it if your market is Microsoft-centric — and be aware you are also betting on Fabric adoption, which is newer than the AWS and Google Cloud data stacks it competes with.
| Factor | AWS DEA-C01 | Google Professional Data Engineer | Microsoft DP-700 |
|---|---|---|---|
| Tier | Associate | Professional | Associate |
| Cost | $150 (varies by region) | $200 + tax | Varies by country (~US$165 commonly cited) |
| Format | 65 questions, 130 min | 40–50 questions, 2 hrs | 100-minute exam (question count unpublished) |
| Passing | 720/1,000 scaled | Pass/fail, score unpublished | 700/1,000 scaled |
| Prerequisites | None | None | None |
| Validity/renewal | 3 years; paid renewal, 50% voucher | 2 years; $100 short renewal exam | 1 year; free online renewal |
| Core skills | AWS pipeline services, ingestion, transformation | End-to-end GCP data stack, ML-adjacent | Fabric, SQL, PySpark, KQL |
| Best for | The broadest cloud job market | Data-heavy GCP shops; strongest seniority signal | Microsoft/Fabric enterprises |
| Career path | Data engineer on AWS stacks | Senior/lead data engineer on GCP | Data engineer in the Microsoft ecosystem |
No universal winner: the AWS exam is the most affordable and most widely applicable; the Google credential carries Professional-tier weight and suits data-mature organisations; DP-700 is the only rational Microsoft-stack choice now that DP-203 is gone. If you are still genuinely undecided between provider ecosystems as a whole, that decision framework lives in our cloud certifications comparison — pick the ecosystem first, then return here.
Cloud provider exams certify breadth across one vendor's stack. The platform vendors certify depth in the engines where much real data engineering happens — and job adverts increasingly name them directly.
Databricks runs its own certification track for data engineers built around Spark and the lakehouse model. Snowflake likewise offers role-based certifications for practitioners on its warehouse platform. Our research for this article did not extend to verifying the current exam names, formats or fees for either vendor — both programmes have changed over time, so take the details from the official certification pages (databricks.com and snowflake.com) rather than from third-party summaries, ours included. What the market signal supports saying is this: a Databricks credential is the natural stage-4 move if your milestone-4 platform is Spark and lakehouse architecture, while Snowflake certification fits warehouse-centric analytics engineering; and either one layered on top of a stage-3 cloud certification reads as a specialist profile, not a generalist one.
Sequence matters here. A platform certification before a cloud certification leaves gaps interviewers will find — Databricks jobs still run on AWS, Azure or GCP storage, networking and IAM. Cloud first, platform second.
Worth considering, case by case:
Generally skip for this career path: multiple fundamentals badges, database-administration specialty exams (AWS retired its Database – Specialty in 2024; Google's Professional Cloud Database Engineer targets a heavier, DBA-flavoured profile with a recommended five years' experience), and any exam whose current version you cannot confirm on an official page.
Because readers arrive at data engineering from different directions, here is the staged path applied to the three most common starting points.
The analyst (SQL-strong, infrastructure-weak). Stage 1 optional but useful — Google's Associate Data Practitioner fits well. Six months of Python and pipeline projects. Then the stage-3 exam for whichever cloud her employer runs; if that is Microsoft, DP-700, and her PL-300 background compounds nicely. Databricks or Snowflake specialisation in year two, matching the platform in-house.
The backend developer (Python-strong, data-modelling-weak). Skip stage 1. Go almost straight at AWS DEA-C01 or the Google Professional Data Engineer, using the exam objectives to force the warehousing, modelling and governance study he would otherwise avoid. His gap is rarely code; it is understanding why pipelines are built the way they are.
The career changer from outside tech. Stage 1 is genuinely for you — DP-900 or Associate Data Practitioner — but the certification path alone will not carry a zero-experience CV into a mid-level data engineering role; plan for an adjacent first job (analytics, support engineering, junior BI) as the bridge. The wider playbook for that kind of move is our roadmap on certifications for switching careers.
Certifications do not come with salaries attached, and no reliable survey publishes a "data engineer certification premium." The nearest well-sourced anchors: Robert Half's 2026 US Technology Salary Guide lists midpoint salaries of $153,750 for data scientists and $117,250 for data analysts — data engineers typically sit between the two in US market guides, but Robert Half's data engineering midpoint was not verified for this article, so treat the bracket as context rather than a figure. Skillsoft's 2025 IT Skills and Salary Report found IT professionals worldwide reporting about 5% year-on-year earnings growth. All such numbers vary heavily by country, sector and experience; check the named surveys directly for your own market rather than relying on aggregator sites.
If your target employers run AWS, the path is: Python-and-SQL projects → AWS Certified Data Engineer – Associate → Databricks or Snowflake depending on the in-house platform. If they run Google Cloud: Associate Data Practitioner (optional) → Professional Data Engineer → platform specialisation. If they are a Microsoft shop: DP-900 (optional) → DP-700 Fabric Data Engineer → the same stage-4 choice. Every route shares the same skeleton — fundamentals if needed, one cloud data credential at the centre, one platform specialisation on top — and that skeleton is the real answer, because it will outlive the next round of exam retirements. Browse the full certification exams directory to see what preparation resources exist for whichever branch you pick, and confirm every fee and exam version on the vendor's official page before you book: in data engineering, more than anywhere else in cloud, the exams move.
No — it cannot be taken. DP-203 and the Azure Data Engineer Associate certification retired on 31 March 2025. Existing holders keep the credential on their transcript, but new candidates should target DP-700, the Fabric Data Engineer Associate, which is Microsoft's current data engineering certification.
No employer universally requires one, and a portfolio of working pipelines plus SQL/Python skill is the core of any transition. Certifications accelerate rather than replace that: they structure your study across a full vendor stack and give recruiters a legible signal, which matters most when your job history does not yet say "data."
Cloud first in almost every case. Platform specialisations assume the cloud storage, networking and identity context that provider exams teach, and hiring managers read platform-only profiles as narrow. The exception: your current employer runs one of these platforms and will hand you real work on it now — real work beats sequence purity.
There is no official figure and study time depends heavily on your starting point, so treat any quoted hour-count sceptically. As an editorial estimate of shape rather than duration: analysts and developers typically spread the fundamentals-to-central-certification stretch across several months of part-time study, with platform specialisation following in a second year. Consistency matters more than intensity — the exams reward accumulated hands-on time.
Exam facts in this guide were checked against official certification-provider pages on . Fees, exam codes and policies change — confirm on the provider’s own site before you book.
Put it into practice
Reading about an exam only takes you so far. Work through practice questions for your certification and find the gaps before exam day does.
·10 min read
A five-step decision process for choosing a professional certification: goals, market demand, true cost, prerequisites and renewal — with a comparison worksheet.
Continue reading·12 min read
The strongest professional certifications for 2026, ranked on demand, recognition and cost of ownership — with the year's retirements and AI-driven changes factored in.
Continue reading·8 min read
How certifications actually influence recruiters when you change employers, which types move offers upward, and how to time one around a job search.
Continue reading
