In one line: An index of the AI and machine learning certifications that are still available as of 2026-10 (Anthropic, AWS, Google Cloud, NVIDIA, Databricks, Oracle, GitHub), which ones fit which role, which were retired in 2026, and links to the free study guide for each on this site. Last reviewed: 2026-10-21
Certifications are most useful when you work with a specific vendor or cloud, sell or advise on it (solutions, consulting, pre-sales), or need structure for self-study. For hands-on AI engineering roles, projects matter more; see the AI engineer roadmap.
Start with one foundational exam that matches your stack, then one associate or professional exam aligned with your target role. Two well-chosen certifications beat six random ones.
Several popular AI exams were retired in 2026: AWS Machine Learning Engineer - Associate (MLA-C01) and Microsoft's AI-900, AI-102 and DP-100. Older roadmaps still recommend them. Check the retired list below before you study.
Every certification below links to a free study guide on this site (fact sheet, notes, practice plan, and for most flashcards and practice questions). All of them are on the Certifications hub.
Prices and formats change. Figures here come from the study guides on this site, which were last updated between May and October 2026; confirm on the vendor's page before booking.
Anthropic launched its certification program in March 2026 and expanded it to four exams in July 2026. Exams are delivered by Pearson VUE and registered through the Anthropic Partner Academy; the study guides note that free Claude Partner Network membership is required.
Microsoft retired its main AI certifications in 2026 (see below). This site keeps an Azure generative AI study track for learning the platform. Check Microsoft Learn credentials for any newer AI credentials before planning around Azure.
Read the official exam guide for domain weights. Spend time in proportion.
Do practice exams early to find weak domains, then study those. Cheat sheets and fact sheets help for last-week review.
Get hands-on with the vendor's console or API. Scenario questions reward people who have actually built something. The projects in build/ double as exam preparation for the developer-level exams.
Yes for consulting, pre-sales, solutions architecture and partner roles, where vendors and clients count them; for structured learning; and for getting past CV filters in some companies.
Less so for research roles and for hands-on AI engineering at product companies, where interviewers ask about what you built. See AI engineer interview.
Watch renewal costs: some AI certifications are valid for only 12 months (Anthropic's) or two years (NVIDIA's, Databricks'), because the field moves fast.