CertifyDeckDP-100 retired on June 1, 2026 and can no longer be scheduled. AI-300 is the current related path for operationalizing machine learning, but it expands the scope with infrastructure as code, CI/CD, Microsoft Foundry, GenAIOps, generative evaluation, and RAG.
What DP-100 measured
DP-100 targeted the Azure Data Scientist role and emphasized data exploration, model training, jobs, Azure Machine Learning, MLflow, and machine learning deployment.
The exam and credential are retired. The knowledge remains useful, but any course presenting DP-100 as bookable is outdated.
What AI-300 adds
AI-300 tests secure workspaces, registries, Bicep, Azure CLI, GitHub Actions, progressive delivery, rollback, drift, alerts, and retraining. It sits between data science and DevOps.
It also adds GenAIOps: Foundry, foundation models, provisioned throughput, prompt source control, quality and safety evaluation, traces, token cost, RAG optimization, and fine-tuning.
What to reuse and what to learn
Reuse Python, experiments, metrics, MLflow, training, pipelines, registration, and endpoints. Ensure the material reflects the current Azure Machine Learning experience.
Add GitHub Actions, Bicep, managed identities, RBAC, private networking, progressive delivery, observability, Foundry, RAG, generative evaluation, and customized-model operations.
Recommended 2026 path
If you were nearly ready for DP-100, compare your knowledge against the AI-300 outline, retain the Azure ML labs, and focus new effort on IaC, CI/CD, and generative domains.
If starting fresh, learn Python, ML, and Azure fundamentals before following the AI-300 preparation plan. Do not purchase material promising access to retired DP-100.
Frequently asked questions
Does AI-300 replace DP-100?
It is the closest current operations path, but not a one-to-one replacement because it adds GenAIOps, Foundry, IaC, and production automation.
Can I still take DP-100?
No. Microsoft retired the exam on June 1, 2026.
Will a DP-100 course prepare me for AI-300?
Only partly. Azure ML, MLflow, training, and endpoints transfer, but you must add GitHub Actions, Bicep, Foundry, generative evaluation, RAG, and operational fine-tuning.