AWS vs Azure vs GCP AI certifications: a level-by-level comparison

If you want one AI cert and don't know which cloud to pick, the short version: AWS for breadth, Azure for enterprise + Microsoft-stack shops, GCP for serious ML engineering. The longer version is below — by level, by services covered, and by who each cert is actually for.

The cert ladders are not symmetrical across clouds. AWS has the cleanest three-level path right now (Foundational → Associate → Pro). Azure has a strong Foundational + Associate but no expert-level pure-AI cert. GCP has Foundational + Pro but skipped the associate tier. That asymmetry is itself part of the story.

Here's the side-by-side, then the deep dives.

At-a-glance map

Level	AWS	Azure / Microsoft	GCP
Foundational	AIF-C01 (AI Practitioner)	AI-900 (Azure AI Fundamentals)	Generative AI Leader
Associate	MLA-C01 (ML Engineer Associate)	AI-102 (Azure AI Engineer Associate); DP-100 (Data Scientist)	— (gap)
Professional / Expert	AIP-C01 (Generative AI Developer Pro)	— (gap)	PMLE (Professional ML Engineer)

A few things to notice:

• The associate tier is where Azure goes broader — two distinct certs (AI-102 for AI engineering, DP-100 for data science / ML).
• GCP has no associate-level AI cert. If you want a credential between the foundational Generative AI Leader and the Pro ML Engineer, there isn't one.
• AWS has the only "GenAI Developer Pro" cert as of 2026. AIP-C01 is GenAI-specific in a way none of the others are.

Foundational tier — AIF-C01 vs AI-900 vs Generative AI Leader

These three are the entry-level, conceptual, "I understand cloud AI without writing code" credentials. All three are roughly equal in difficulty (fairly approachable), all three cost about $99–$100, and all three target the same audience: PMs, BAs, sales engineers, technical decision-makers, and engineers stepping into cloud AI for the first time.

AWS AI Practitioner (AIF-C01)

Launched October 2024. $100 USD, 65 questions, 90 minutes.

Services covered:

• Amazon Bedrock (foundation models, agents, knowledge bases, guardrails)
• Amazon SageMaker (basics — Studio, JumpStart, model registry)
• Amazon Q (developer + business)
• Amazon Comprehend (NLP / sentiment / entity extraction)
• Amazon Transcribe (speech-to-text)
• Amazon Translate
• Amazon Polly (text-to-speech)
• Amazon Rekognition (vision)
• Amazon Textract (document extraction)
• Amazon Kendra (enterprise search)
• Amazon Lex (chatbots)

The exam is heavy on matching services to use cases. "A retail company wants to automate customer support email routing — which AWS service?" That kind of question. About 30% of the exam is responsible-AI / governance / explainability / bias mitigation, which surprises candidates who expected pure technology questions.

Azure AI Fundamentals (AI-900)

$99 USD, ~40 questions, 60 minutes. Never expires.

Services covered:

• Azure AI Services (the umbrella formerly known as Cognitive Services)
• Azure OpenAI Service (GPT-4, GPT-4o, DALL·E, Whisper)
• Azure Machine Learning Studio (low-code ML)
• Form Recognizer / Document Intelligence
• Azure AI Speech (recognition, synthesis, translation)
• Azure AI Vision (image analysis, OCR, custom vision)
• Azure AI Language (sentiment, key phrase, NER, conversational language understanding)
• Azure AI Search (formerly Cognitive Search)
• Azure Bot Service / Bot Framework

AI-900 leans more toward Azure's ML platform than AWS's AIF-C01 does. There's more hands-on flavor — questions about training a model in Azure ML designer, evaluating accuracy/precision/recall metrics. Less time on responsible AI than AIF-C01, more time on classical ML concepts.

The "never expires" status is real and meaningful. Microsoft's other fundamentals (AZ-900, DP-900, SC-900) are also lifetime — for fundamentals, this is the standard.

GCP Generative AI Leader

$99 USD. Newest of the three (introduced 2024). Explicitly non-technical — billed as a leadership / strategy cert.

Services covered:

• Vertex AI Generative AI (Gemini family, Imagen, Codey, MedLM)
• Gemini in Workspace
• Vertex AI Search and Conversation
• Vertex AI Model Garden (third-party models — Anthropic Claude, Meta Llama, etc.)
• Vertex AI Agent Builder
• Document AI (similar role to AWS Textract)
• Translation API
• Speech-to-Text / Text-to-Speech
• Vision AI

GAIL is the most strategy-flavored of the three. Expect questions on AI program governance, RAG patterns conceptually, prompt engineering basics, model selection criteria, and Google's responsible-AI principles. Less platform mechanics than AIF-C01 or AI-900.

If you're a leader / PM evaluating cloud AI providers — this is arguably the best cert for that audience because it's pitched at exactly that altitude.

Which Foundational to pick?

If you've already chosen a cloud: take that cloud's foundational cert. The transferable knowledge is roughly the same, but the named services aren't, and you'll save yourself a lot of "what's the AWS equivalent of Cognitive Services?" mental gymnastics.

If you haven't chosen: AIF-C01 has the broadest service surface and the most weight on responsible AI, which is increasingly what enterprises want to talk about. AI-900 is the easiest of the three and never expires. GAIL is the only one specifically pitched at non-engineering leadership.

Associate tier — MLA-C01 vs AI-102 vs DP-100 (no GCP equivalent)

Now we're in genuinely different territory. The associate-level certs assume hands-on experience and test deeper service knowledge.

AWS ML Engineer Associate (MLA-C01)

Launched August 2024. $150 USD, 65 questions, 170 minutes. Replaced the old ML Specialty (MLS-C01).

Services covered:

• Amazon SageMaker (deep — Studio, Pipelines, Feature Store, Model Registry, Model Monitor, Clarify, Data Wrangler, Ground Truth, JumpStart, Canvas)
• Amazon Bedrock for fine-tuning + provisioned throughput
• AWS Glue (data prep)
• Amazon S3 + S3 Tables + Lake Formation (data lake patterns)
• Amazon Athena, Redshift (analytics for ML)
• Amazon Kinesis Data Streams / Firehose (streaming features)
• Step Functions (orchestration)
• CloudWatch Container Insights for ML monitoring

The cert is operational ML, not pure modeling. Expect questions on monitoring drift, retraining triggers, A/B testing model versions, cost optimization for inference, MLOps patterns. If you came in expecting "build a CNN from scratch", you'll be disappointed (and unprepared).

Azure AI Engineer Associate (AI-102)

$165 USD, ~50–60 questions, 100 minutes. Got a meaningful refresh in early 2025 to add agentic-solutions content.

Services covered:

• Azure OpenAI Service (deep — including fine-tuning, completions, embeddings, function calling, assistants API, Azure AI Foundry)
• Azure AI Services (formerly Cognitive Services — full suite)
• Azure AI Search (deep — vector search, hybrid retrieval, semantic ranking, RAG patterns)
• Azure AI Document Intelligence (formerly Form Recognizer)
• Azure AI Speech (custom speech, custom voice, real-time translation)
• Azure AI Language (custom NER, classification, conversational language understanding)
• Azure AI Vision (custom vision, face, video indexer)
• Azure AI Content Safety
• Container apps for deploying AI models
• Azure AI Agent Service (the new agentic content from the 2025 refresh)

AI-102 is the cert most directly comparable to MLA-C01 in scope — both expect you to ship AI workloads to production and operate them. The difference is service emphasis: AI-102 is about Azure OpenAI + RAG + AI Search, MLA-C01 is about SageMaker + Bedrock at scale.

Azure Data Scientist Associate (DP-100)

$165 USD. Distinct from AI-102 — DP-100 is data-science / classical ML focused, AI-102 is GenAI / cognitive services focused.

Services covered:

• Azure Machine Learning workspace (deep — compute clusters, environments, experiments, jobs, endpoints, MLflow integration)
• Azure ML SDK / CLI
• AutoML
• ML Pipelines
• Model registry and deployment
• Responsible AI dashboard (interpretability, fairness, error analysis)
• Azure Synapse Analytics for data prep
• Azure Databricks integration
• Compute optimizations (CPU vs GPU, spot, low-priority)

If you're a data scientist building custom models, DP-100 is the cert. If you're an AI engineer shipping Azure OpenAI applications, AI-102 is the cert. They overlap maybe 20%, mostly in the deployment / monitoring topics.

GCP — there's no associate-level AI cert

This is a real gap in GCP's catalog as of 2026. Google has the Cloud Digital Leader (foundational), the Generative AI Leader (foundational), and the Professional ML Engineer (which is genuinely pro-tier). The path from GAIL to PMLE is steep — there's no intermediate credential.

If you want a GCP-specific intermediate signal: the Associate Cloud Engineer (ACE) cert, while not AI-focused, covers Vertex AI deployment basics. Some engineers position it as "I can run AI workloads on GCP without being an AI specialist." It's a workaround, not a clean answer.

Which Associate to pick?

• Building custom models / classical ML: DP-100 (Azure) is the most focused.
• Shipping Azure OpenAI apps to production: AI-102 (Azure).
• Operating SageMaker + Bedrock at scale on AWS: MLA-C01 (AWS).
• GCP-only: skip to PMLE; there's no intermediate option.

The closest cross-cloud parallel is MLA-C01 ≈ AI-102 — both test "ship AI to production and operate it." Different service surfaces, similar engineering altitude.

Professional / Expert tier — AIP-C01 vs PMLE (no Azure equivalent)

AWS Generative AI Developer Professional (AIP-C01)

$300 USD, 75 questions, 180 minutes. Launched in 2025 as AWS's first GenAI-specific Pro cert.

Services covered:

• Amazon Bedrock at depth (custom models via continued pre-training, model evaluation, agents with multi-step reasoning, knowledge bases with hybrid search, guardrails configuration)
• Amazon Bedrock Studio + Bedrock IDE
• SageMaker JumpStart for foundation model fine-tuning
• SageMaker for hosting custom models
• AWS App Runner / ECS Fargate for inference services
• Amazon OpenSearch as a vector store
• Amazon Q for code generation use cases
• IAM roles for cross-service GenAI access
• AWS Step Functions for orchestrating complex agent workflows

AIP-C01 is the only major-cloud cert specifically dedicated to GenAI development — not classical ML, not "AI services" broadly. Expect deep questions on retrieval-augmented generation architectures, model evaluation strategies (HHEM, ROUGE, custom evals), token cost optimization, hallucination mitigation, and multi-agent orchestration.

This is a brand-new cert. Salary data is too thin to quote with confidence — see the AIF-C01 salary post for adjacent role context.

Google Cloud Professional ML Engineer (PMLE)

$200 USD. One of the highest-paying single cloud certs per levels.fyi, partly because the candidate pool is small.

Services covered:

• Vertex AI Workbench (managed notebooks)
• Vertex AI Pipelines (Kubeflow Pipelines on managed infrastructure)
• Vertex AI Training (custom containers, hyperparameter tuning)
• Vertex AI Prediction (online + batch endpoints, custom serving containers)
• Vertex AI Model Registry + Model Monitoring
• Vertex AI Feature Store
• Vertex AI Generative AI (Gemini, Model Garden, agents)
• Vertex AI Search and Conversation
• BigQuery ML (in-database ML)
• TensorFlow Extended (TFX) integration
• Kubeflow on GKE for self-managed ML
• Dataflow for ML data pipelines
• Cloud Composer (Airflow) for orchestration
• AutoML Tables / Vision / NLP

PMLE is broader than AIP-C01. It covers classical ML, MLOps, AND GenAI — all on Vertex AI's relatively unified surface. The exam is scenario-heavy in the way GCP Pro exams are: long case studies that hinge on architectural trade-offs ("which solution best balances cost, latency, and accuracy under these constraints?").

Microsoft — there's no Expert-level pure AI cert

As of 2026, Microsoft has no expert-level AI cert. AI-102 is the top of the AI ladder. The closest expert-level credential touching AI is Azure Solutions Architect Expert (AZ-305), which has scattered AI questions in the context of broader architecture, or Microsoft Cybersecurity Architect (SC-100), which touches AI security obliquely.

If Microsoft adds an "AI Architect Expert" cert in 2026 or 2027, expect it to consolidate AI-102 + DP-100 expertise into a more strategic exam. As of right now: it doesn't exist.

Which Pro to pick?

• GenAI-only focus on AWS: AIP-C01 is the deepest credential available anywhere right now for that scope.
• End-to-end ML engineering on GCP, including GenAI: PMLE is broader but still GCP-specific.
• Microsoft-stack senior AI roles: there's no exam — pair AI-102 with AZ-305 or DP-100 instead.

The Pro tier is where the cross-cloud comparison breaks down most. Each cloud took a different bet about what "professional AI engineer" means.

A note on cert renewals

This matters more for AI certs than for other categories because AI changes fast.

• AWS AI certs: 3-year validity. Renew by re-passing the current version.
• Azure AI certs: 1-year validity for role-based (AI-102, DP-100), but free renewal via unproctored online assessment on Microsoft Learn starting 6 months before expiration. Fundamentals (AI-900) never expires.
• GCP AI certs: 3 years for Foundational/Associate, 2 years for Professional. Renew by re-passing.

Microsoft's renewal model is dramatically friendlier than the others. For AI specifically, where the underlying services (Azure OpenAI, Bedrock, Vertex AI) refresh every few months, the renewal cost compounds. Worth factoring in if you're choosing between two roughly equal credentials.

My recommendation, by role

• AI / ML Product Manager: GAIL (GCP) or AIF-C01 (AWS) — the strategy-tier certs. Both. Or one and the other later.
• Backend engineer adding AI to a product: AI-102 (Azure) if your stack is Microsoft-leaning, MLA-C01 + AIF-C01 (AWS) if cloud-leaning.
• Data scientist: DP-100 (Azure) for classical ML, PMLE (GCP) for the broader scope.
• Senior ML engineer / MLOps lead: PMLE (GCP) if your stack is anywhere near Vertex AI, MLA-C01 (AWS) otherwise. Add AIP-C01 (AWS) if your team is GenAI-heavy.
• AI safety / responsible AI work: AIF-C01 (AWS) covers this best at the foundational tier. None of the higher-tier certs go deep on responsible AI as an isolated topic.

What to do this week

If you're already studying for one of these: drill the questions. Browse the AIF-C01 bank, the MLA-C01 bank, AI-102, PMLE, or any of the others on CertLabPro.

If you're picking your first AI cert: identify which cloud your employer (or target employer) uses, then take the foundational cert at that cloud. The compounding from there — to associate, to pro — is much faster than starting on the wrong cloud and pivoting later.

If you're trying to figure out which cloud's AI ladder is "best": there's no winner. Each cloud's catalog reflects different bets about what AI engineering means. Pick the one whose bets match your work.