Cross-vendor comparison

AI certifications compared (2026): AWS, Google, Microsoft, Anthropic

How to read this comparison

AI certifications split along a few axes, and confusing them is the most common mistake people make when choosing one:

• Level. Foundational exams test literacy — what GenAI is, what the platform offers, when to use it. Associate / professional exams test building and operating. A foundational badge and a professional badge from the same vendor signal very different things.
• Focus. Most certs are vendor-specific and assume their platform (Bedrock, Azure AI, Google Cloud). Anthropic's is model-and-pattern specific (Claude, agentic). Read the cert for the stack you actually ship on.
• Format and cost. Proctored multiple-choice exams with a scaled pass mark are the norm; a few "certificates" are really self-paced course completions, which is a different kind of credential.
• Recognition. Vendor exams from AWS, Microsoft, Google, Anthropic, and NVIDIA carry weight at partners and consultancies. Many "Agentic AI Certified" badges in the wild are third-party marketing — treat anything not backed by a named vendor with suspicion.

One disambiguation up front, because it trips people up: AWS AIF-C01 (Certified AI Practitioner) is the foundational exam, pass mark 700; AWS AIP-C01 (Certified Generative AI Developer – Professional) is a separate professional exam, pass mark 750. Same vendor, different tiers. The full per-domain blueprints are in the exam guide.

The major AI certifications, side by side

The table compares the credentials a senior engineer or AI leader is most likely to weigh in 2026 — across vendors and levels. Figures are confirmed against official pages where noted; cells marked * are region-priced or were not published on the official cert page at the time of writing (confirm before you register).

* Confirm before registering. AWS does not publish a single global price for AIP-C01 on the pages we used. Microsoft AI-102 / AI-900 prices (~$165 / ~$99) and the AI-102 pass mark are standard published figures not printed on the current cert page; AI-900's 700 pass mark is officially stated. CCA-F is partner-gated (free for early Claude Partner Network members, then about $99). Google AI Essentials is a Coursera course (subscription priced), not a proctored vendor exam.

By vendor: AWS

AWS runs the most structured AI ladder. The AI Practitioner (AIF-C01) is the foundational on-ramp — broad GenAI literacy, pass 700, $100 — useful if AI is new to you, and usually skippable for an experienced AWS engineer. The Generative AI Developer – Professional (AIP-C01) is the building-grade exam: foundation-model integration, RAG, AI safety and governance, operations, and testing on Bedrock, with a 750 pass mark. If you target AWS-partner GenAI work, AIP-C01 is the credential that signals it. Per-domain weights for both are in the exam guide; we keep a fuller AWS-specific walkthrough in the AWS AI Practitioner page.

By vendor: Anthropic

Anthropic's Claude Certified Architect – Foundations (CCA-F) is the odd one out — not tied to a cloud, but to a model family and a way of building. It is scenario-based (you reason through 4 of 6 scenarios rather than recalling trivia), pass mark 720, and weighted toward agentic architecture, tool / MCP design, Claude Code workflows, and prompting. It is the clearest single signal for the agentic path, with the caveat that the badge is obtained through the Claude Partner Network even though the prep is open to everyone. Detail on the exam and how to prepare is in the Anthropic certification guide.

By vendor: Google

Google has two genuinely different things that share the word "AI." The Generative AI Leader is a foundational Google Cloud exam aimed at decision-makers — GenAI fundamentals and Google Cloud's offerings, $99, no prerequisites. The Professional Machine Learning Engineer is a senior, hands-on Google Cloud exam (model building, pipelines, MLOps, $200) for experienced practitioners. Separately, Google AI Essentials is a short Coursera course with a completion certificate, not a proctored Google Cloud credential — fine for general productivity, but don't mistake it for the Cloud exams. If your stack is Google Cloud and you build models, the Professional ML Engineer is the credential that carries weight.

By vendor: Microsoft

The Azure AI Engineer Associate (AI-102) tests building AI solutions with Azure AI services and Azure OpenAI — including generative and agentic solutions — at the associate level. Azure AI Fundamentals (AI-900) is the foundational literacy exam (pass 700). Microsoft is also rolling out agent-specific credentials — an AI Agent Builder Associate (Copilot Studio, in beta) and an advanced Agentic AI Business Solutions Architect — which lean on open standards like MCP and Agent2Agent. Treat the newer ones as emerging until they are generally available.

The emerging agentic-AI certs

A new category is forming around agentic AI specifically — systems where a model runs in a loop and calls tools — and it is where the agentic AI engineer role is headed. The credentials are real but young:

• NVIDIA Agentic AI Professional (NCP-AAI) — a professional exam on architecting, deploying, and governing agentic and multi-agent systems (60–70 questions, 120 min, $200). NVIDIA also offers the established associate-level NCA-GENL for GenAI / LLM fundamentals ($125).
• Microsoft's agent certifications — the AI Agent Builder Associate (Copilot Studio, beta) and the advanced Agentic AI Business Solutions Architect — bring agentic patterns into the Microsoft stack.
• Anthropic's CCA-F is, in practice, the most agentic-focused of the mainstream exams, even though it is branded as foundational.

There is, as of mid-2026, no vendor-neutral agentic-AI certification with broad recognition — the cross-industry effort (the Linux Foundation's Agentic AI Foundation, backed by Anthropic, OpenAI, Google, Microsoft, AWS, Cloudflare and others) is steering open standards like MCP, not running an exam. So pick a vendor exam that matches your stack, and weigh the skill over the badge.

Which should you choose?

Map the cert to where you are heading, not to the longest list of badges:

• New to AI, any cloud: one foundational exam is plenty — AIF-C01, Google's Generative AI Leader, or AI-900 (mind the AI-900 retirement). They are interchangeable literacy signals.
• Engineer on AWS: skip the practitioner, go for AIP-C01, and add CCA-F for agentic depth.
• Engineer on Azure: AI-102 today, but watch the June 2026 transition and the newer agent certs.
• ML practitioner on Google Cloud: the Professional Machine Learning Engineer is the senior, defensible choice.
• Claude / agentic builder, any platform: CCA-F is the clearest signal, optionally with NVIDIA's NCP-AAI as the agentic specialism matures.
• Everyone: a certification opens a door; a portfolio walks through it. Spend at least as long building a real agent, an eval suite, or an MCP server as you do on the exam.

For the ROI verdict on each — worth it, maybe, or skip — see which AI certifications are worth it for a senior engineer. For the domain-by-domain exam blueprints, see the AI certification exam guide. And for how a curriculum can map to these domains while producing the portfolio, see the curriculum.

Frequently asked questions

What are the best AI certifications in 2026?

There is no universal best — it depends on your platform and level. For literacy, the foundational exams (AWS AIF-C01, Google Cloud Generative AI Leader, Microsoft AI-900) are interchangeable on-ramps. For engineers, the credentials that signal real building are Anthropic's CCA-F, AWS AIP-C01, Microsoft AI-102, and Google's Professional Machine Learning Engineer. Choose by the stack you ship on.

Which AI certification should I get?

Pick by destination. New to AI: one foundational exam. Building on AWS: AIP-C01, plus CCA-F for agentic depth. On Azure: AI-102, watching the June 2026 exam transition. On Google Cloud as an ML practitioner: the Professional Machine Learning Engineer. On the Claude or agentic path regardless of cloud: CCA-F. Then invest equally in a portfolio.

AWS vs Google vs Microsoft AI certification?

Each is vendor-specific and assumes its own platform, so the right one is the cloud you actually build on. AWS has the most structured ladder (AIF-C01 foundational, AIP-C01 professional). Google splits between business literacy (Generative AI Leader) and senior ML (Professional ML Engineer). Microsoft's AI-900 and AI-102 are in transition in 2026, with newer agent-focused certs arriving. If you are cloud-agnostic, weigh the model-specific CCA-F instead.

What is the Anthropic certification?

The Claude Certified Architect – Foundations (CCA-F) is Anthropic's official credential. It is scenario-based multiple choice (you sit 4 of 6 scenarios), with a 720/1000 pass mark, weighted toward agentic architecture, tool and MCP design, Claude Code workflows, and prompting. It is obtained through the Claude Partner Network, though the prep is open to everyone. It is the most agentic-focused of the mainstream exams.

Are AI certifications worth it?

They help you pass filters, especially at consultancies and cloud partners where vendor certs matter, but they do not qualify you on their own — most teams weigh a portfolio and interview performance more heavily. Treat a cert as screen-ready, not hired, and pair it with demonstrable builds. Our full verdict per cert is on the worth-it page.

What is the most advanced AI certification?

Among the mainstream exams, the professional-tier credentials are the most advanced: AWS AIP-C01 (Generative AI Developer – Professional) and Google Cloud's Professional Machine Learning Engineer. For agentic systems specifically, NVIDIA's Agentic AI Professional (NCP-AAI) and Microsoft's advanced Agentic AI Business Solutions Architect sit at the top, though both are newer and less established.