A certification mentioned on this page has been retired. Microsoft Certified: Azure AI Engineer Associate (AI-102) is no longer available to take. Microsoft reports the retirement date as 2026-06-30. The replacement is Microsoft Certified: Azure AI Apps and Agents Developer Associate (AI-103). Read any reference below as historical, not as advice to take this retired exam. Check the successor's current requirements before planning your preparation.
A certification exam is not a course. This page lists the vendor exams — the proctored, paid credentials issued by the cloud and platform vendors themselves — and compares them on what we can verify from each vendor's own page.
Why these are on a separate page from our rankings. Our main rankings score Coursera learning programmes on six factors: employer recognition, skill value, cost and value, time to complete, difficulty, and salary impact. A vendor exam teaches you nothing — it tests you, charges a fee, and in most cases expires. Scoring the two against each other produces a number that is meaningless in both directions, so we do not. Nothing on this page is ranked. We have verified what each credential is, what it costs and how it is sat; we have not assessed them on our six factors, and we are not going to invent scores to fill a column.
Course or exam: which do you actually need?
Read the job advert. If it names a specific credential — "AWS Certified…", "Microsoft Certified…" — you need the exam, and a Coursera certificate will not substitute. If it asks for skills, a course that leaves you with projects is usually the better use of the same money, and our course rankings cover those.
The two are not rivals. The common sequence is a course to learn the material and build something, then an exam only where an employer or a partner agreement actually requires the badge.
The vendor exams, compared
The table below compares 7 vendor certification exams on vendor, level, fee and exam format. Every figure comes from the vendor's own certification page and was verified on 22 August 2026; where a vendor publishes no figure, the cell says so rather than guessing.
| Credential | Vendor | Level | Fee | Exam format |
|---|---|---|---|---|
| AWS Certified Generative AI Developer - Professional | AWS | Advanced | $300 USD | 180 minutes · 75 questions |
| Databricks Certified Generative AI Engineer Associate | Databricks | Intermediate | $200 registration fee | 90 minutes · 45 scored questions · valid 2 years |
| Generative AI Leader | Google Cloud | Foundational | $99 (plus tax where applicable) | 90 minutes · 50–60 questions · valid 3 years |
| Professional Machine Learning Engineer | Google Cloud | Advanced | $200 (plus tax where applicable) | 2 hours · 50–60 questions |
| Microsoft Certified: Azure AI Apps and Agents Developer Associate | Microsoft | Intermediate | $165 USD (varies by country/region) | Not published by the vendor |
| NVIDIA-Certified Associate Generative AI LLMs | NVIDIA | Beginner | $125 | 60 minutes · 50–60 questions · valid 2 years |
| NVIDIA-Certified Professional Generative AI LLMs | NVIDIA | Advanced | $200 | 120 minutes · 60–70 questions · valid 2 years |
Fees are the vendor's list price before local tax, and several vendors price by region. Renewal is the cost people forget: most of these expire in two or three years, so the real figure is the fee multiplied by however long you intend to keep the credential current.
Exams that no longer exist
These still appear in older guides and job adverts. They cannot be sat.
- Microsoft Certified: Azure AI Engineer Associate — retired. Replaced by Microsoft Certified: Azure AI Apps and Agents Developer Associate (AI-103).
How to choose between them
Match the exam to the platform your employer actually runs. A Google Cloud credential is worth little in an AWS shop, and the reverse is equally true — these are vendor credentials and their value is concentrated inside that vendor's ecosystem. If you have no platform constraint, the question becomes which ecosystem you want to work in, which is a career decision rather than a certification one.
Level matters more than brand. A foundational exam aimed at business roles and a professional exam aimed at practitioners are not interchangeable regardless of whose logo is on them.
Frequently asked questions
What is the difference between an AI course and an AI certification exam?
A course teaches you material and issues a completion certificate; anyone who finishes it gets one. A certification exam teaches you nothing — it is a proctored test you pay to sit, and you can fail it. Employers reading a job application treat them differently, which is why a job advert naming a specific vendor credential will not accept a course certificate instead.
Are these exams ranked on this page?
No. Our six scoring factors were built for learning programmes and we have not assessed these exams against them. What this page gives you is verified fact — vendor, level, fee and format, taken from each vendor's own certification page — so you can compare like with like. Inventing scores to make it look like a ranking would tell you nothing true.
Do AI certification exams expire?
Most of them do. The NVIDIA and Databricks credentials here are valid two years and Google Cloud's Generative AI Leader three, per each vendor's own page. That recurring cost belongs in the decision: a credential you intend to hold for six years costs several times its sticker price.
Which vendor exam should I take?
The one for the platform your employer runs. These are vendor credentials and their value sits almost entirely inside that vendor's ecosystem, so a Google Cloud certification is worth little in an AWS shop. If nothing constrains you, choose the ecosystem you want to work in and take the entry-level credential there first.