AWS vs Azure vs Google AI certifications compared
· 7 min read
AWS, Microsoft and Google each offer AI certifications at an entry level and at a technical level, but they divide the ground differently. This comparison covers three exams from AWS (AIF-C01, AIP-C01 and MLA-C02), three from Microsoft (AI-901, AI-103 and AI-300) and two from Google (Generative AI Leader and Professional Machine Learning Engineer). The right one depends far more on the platform you work on and your role than on any ranking between providers.
The eight exams at a glance
All figures are as of September 2026 and come from each provider's official exam and policy pages. Confirm them on the linked page before booking, because fees and formats change.
| Exam | Level | Duration | Questions | Fee (USD) | Focus |
|---|---|---|---|---|---|
| AWS Certified AI Practitioner (AIF-C01) | Foundational | 90 min | 65 | 100 | AI, ML and generative AI concepts and AWS AI services |
| AWS Certified Generative AI Developer – Professional (AIP-C01) | Professional | 180 min | 75 | 300 | Building, securing and operating generative AI applications on AWS |
| AWS Certified Machine Learning Engineer – Associate (MLA-C02, beta) | Associate | 170 min | 85 | 75 (beta price) | Data preparation, model development, deployment and operations for ML and foundation models |
| Microsoft Azure AI Fundamentals (AI-901) | Fundamentals | 45 min | Not published | Priced by country | AI concepts and building simple AI solutions with Microsoft Foundry |
| Azure AI Apps and Agents Developer Associate (AI-103) | Associate (Intermediate) | 120 min | Not published | Typically 165 | Generative AI, agents, vision, text analysis and information extraction on Azure |
| Machine Learning Operations Engineer Associate (AI-300) | Associate (Intermediate) | 120 min | Not published | Typically 165 | MLOps with Azure Machine Learning and GenAIOps with Microsoft Foundry |
| Google Generative AI Leader (GAIL) | Foundational | 90 min | 50–60 | 99 | Generative AI concepts, Google Cloud's offerings and business strategy |
| Google Professional Machine Learning Engineer (PMLE) | Professional | 120 min | 50–60 | 200 | Designing, building, serving, automating and monitoring ML systems on Google Cloud |
Notes on the table:
- Microsoft question counts. Microsoft does not publish a count per exam. Its exam-experience page says most certification exams contain between 40 and 60 questions and that the number varies.
- Microsoft fees. Microsoft prices exams by the country where they are proctored. Its FAQ says associate and expert exams typically cost US$165; the AI-901 page shows the price for your region rather than a fixed figure.
- MLA-C02 is in beta. The AWS page lists beta registration from 1 September 2026 and beta delivery from 29 September 2026, in English only. The outgoing MLA-C01 (130 minutes, 65 questions, US$150) has its last English sitting on 28 September 2026.
Our study guides carry the full detail for each: [AIF-C01](/blog/aif-c01-study-guide), [AIP-C01](/blog/aip-c01-study-guide), [MLA-C02](/blog/mla-c02-study-guide), [AI-901](/blog/ai-901-study-guide), [AI-103](/blog/ai-103-study-guide), [AI-300](/blog/ai-300-study-guide), [GAIL](/blog/gail-study-guide) and [PMLE](/blog/pmle-study-guide).
Entry level: AIF-C01 vs AI-901 vs GAIL
These three look like equivalents, but their intended candidates differ.
AIF-C01 is aimed at anyone exploring AI on AWS, and AWS names business roles alongside technical ones. Its exam guide spreads the weight across AI and ML fundamentals, generative AI fundamentals, applications of foundation models, responsible AI, and security and governance. It is the broadest of the three and has the most languages: twelve, according to the AWS page.
GAIL is the least technical. Google says it is for anyone in any job role, with or without hands-on technical experience. A large share of its weight sits on Google Cloud's generative AI offerings and on business strategy, so it suits managers and decision-makers who need to speak credibly about adoption.
AI-901 is the most hands-on. Microsoft describes the candidate as someone at the start of a career in AI solution development, and expects knowledge of Python syntax and familiarity with Azure resources. More than half of its weight is on implementing AI solutions with Microsoft Foundry. It is also the shortest exam here at 45 minutes, the standard duration for Microsoft fundamentals exams.
If you are choosing between them without a platform preference, the deciding question is how technical you are. Non-technical: GAIL or AIF-C01. Early-career developer: AI-901.
Developer level: AIP-C01 vs AI-103
AIP-C01 is a professional-level exam, and AWS recommends two or more years building production-grade applications on AWS or with open-source technologies plus a year of hands-on generative AI work. It is the longest and most expensive exam in this comparison at 180 minutes and US$300. Its guide puts the most weight on foundation model integration, data management and compliance, followed by implementation and integration, then safety, security and governance.
AI-103 is an associate-level exam for Azure AI engineers who build agents and AI solutions with Microsoft Foundry. It covers more than generative AI: computer vision, text analysis and information extraction each carry their own domain. One practical difference is that Microsoft allows access to Microsoft Learn during associate and expert exams, though not fundamentals exams. The exam timer keeps running while you search, so it helps with a detail you half-remember, not with a topic you never studied.
Of the two Google exams in this comparison, PMLE is the technical option, and it is aimed at ML engineers rather than application developers.
ML engineering: MLA-C02 vs AI-300 vs PMLE
MLA-C02 is AWS's updated associate exam. Its guide balances four domains: data preparation, model and foundation model development, deployment and orchestration, and operating, monitoring and securing solutions. AWS recommends at least a year of experience with its ML services.
AI-300 is about operations more than model building. Microsoft expects a data science background, Python, and entry-level DevOps skills such as GitHub Actions and the command line, plus infrastructure as code with Bicep and the Azure CLI. It is currently offered in English only.
PMLE is Google's professional-level ML exam and the most experience-heavy: Google recommends three or more years of industry experience, including a year or more on Google Cloud. Google notes that the exam was recently updated and directs candidates to the new guide, so check that any preparation material follows it.
Scoring and exam experience
The providers report results differently, which matters when you read practice scores.
- Microsoft reports technical exam scores on a scale of 1 to 1,000 with 700 as the pass mark, and states that this is a scaled score that may not equal 70% of the points.
- AWS also uses scaled scores. The AIF-C01 guide gives 700 as the passing score and says the exam includes 15 unscored questions. AWS's FAQ adds that passing scores are set statistically and can change.
- Google's exam pages for GAIL and PMLE do not list a passing score.
For delivery, all three providers offer online proctoring and test centres for these exams. AWS delivers through Pearson VUE, Microsoft schedules through Pearson VUE, and Google offers online-proctored or onsite-proctored sittings.
How to decide
Use the provider you work with. An exam that tests the services you already operate is easier to prepare for, and it tells an employer on that platform exactly what you can do. If you have no platform yet, look at the job postings you want and count which cloud appears most often.
Then match level to experience. The professional exams (AIP-C01, PMLE) state multi-year experience recommendations; sitting them without that experience means studying services you have never operated.
Finally, consider format. If you prefer shorter exams, GAIL and AI-901 are the lightest. If you want a credential that clearly signals hands-on build skills, AIP-C01, AI-103 and PMLE carry more weight because of what they test, not because of their provider.
For a wider view across Anthropic, NVIDIA, Databricks and others, see [how to choose an AI certification](/blog/how-to-choose-an-ai-certification) or browse the [certifications page](/certifications).
Frequently asked questions
Which is easiest: AIF-C01, AI-901 or GAIL?
It depends on your background. GAIL requires no hands-on experience, AIF-C01 is broad but conceptual, and AI-901 expects Python syntax knowledge. A non-technical candidate will usually find GAIL or AIF-C01 more approachable.
Is AIP-C01 harder than AI-103?
AIP-C01 is a professional-level exam with longer recommended experience and a 180-minute sitting, while AI-103 is associate level. By the providers' own positioning, AIP-C01 targets more experienced candidates.
Should I sit MLA-C01 or MLA-C02?
From 29 September 2026, English candidates can only sit MLA-C02, because AWS lists 28 September 2026 as the last English sitting of MLA-C01. MLA-C01 stays available in Japanese, Korean and Simplified Chinese until MLA-C02 reaches general availability. Prepare against the MLA-C02 guide.
Do these certifications expire?
AWS states its certifications are valid for three years, and Google lists three years for Generative AI Leader. Google and Microsoft each publish renewal rules for their credentials; check the renewal page for yours.
Can I use Microsoft Learn during AI-901?
No. Microsoft allows Microsoft Learn access during associate and expert exams such as AI-103 and AI-300, but not during fundamentals exams.
CertifyAIx is independent and not affiliated with any certification provider. All CertifyAIx practice questions are original.