Microsoft AI Certifications in 2026: The New Path, Explained

Microsoft reshaped its AI certification line in mid-2026: AI-900 retired on June 30 (replaced by AI-901), DP-100 retired on June 1 (replaced by AI-300), and a brand-new AI-103 credential arrived for generative-AI and agent developers. One thing that did not happen, despite what many roundups claim: AI-102 has not been retired — Microsoft frames AI-103 as a new credential, not an AI-102 replacement. If you’re googling any of these codes and getting confused, this is the map.

The 2026 reshuffle at a glance

If you hold AI-900 or DP-100: your certification stays valid until the expiration printed on it (and Fundamentals certs never expire at all), but retired exams can’t be renewed — at renewal time you take the successor, with no automatic conversion. If you hold AI-102: nothing changes — the exam and its renewals remain live. We’ll update this page if Microsoft posts a sunset date.

Tier 1: AI-901 — Azure AI Fundamentals (start here)

The on-ramp: $99, roughly 40–60 questions in 45–60 minutes, 700/1000 to pass, no coding, no prerequisites — and it never expires. It covers AI workloads, responsible AI, machine learning basics, computer vision, NLP, and (heaviest, at 20–25%) generative AI. For most people — including non-technical professionals — this is the right first and possibly only Microsoft AI cert. Start with the free study guide, follow the 2-week study plan, and gauge yourself with free practice questions.

Tier 2 (new): AI-103 — Azure AI App & Agent Developer Associate

Microsoft’s newest AI credential — and a successor to AI-102 in spirit only: Microsoft hasn’t retired AI-102 or positioned AI-103 as its replacement. Where AI-102 centers on wiring up pre-built Azure AI services, AI-103 centers on building generative AI applications and multi-agent solutions with Azure AI Foundry ($165, in beta since April 2026). It’s for developers — expect code, SDKs, and architecture judgment. Study materials are still catching up; if you’ve prepped for AI-102, your service knowledge transfers but the generative-AI and agent content is new ground. (We cover it now: free AI-103 practice questions and a free study guide.)

Tier 2 (data track): AI-300 — MLOps Engineer Associate

The successor to DP-100, with the emphasis moved from designing and training models to operationalizing them — deployment, monitoring, and MLOps practice. If your work is closer to running ML systems than building genAI apps, this is your track. (Interestingly, this mirrors the industry-wide shift: the vendor-neutral CertNexus CAIP also weights operations heaviest.)

Which path, by who you are

• Non-technical professional in a Microsoft shop → AI-901, done. (Worth-it analysis.)
• Developer building AI features → AI-901 as a warm-up if you want it, then AI-103 — or AI-102, which remains live and established.
• Data scientist / ML engineer → AI-300 (and compare the AWS equivalent, MLA-C01, if your stack is mixed).
• Hold DP-100 → nothing to do until renewal; then take AI-300. Hold AI-102 → it stays fully valid and renewable; watch for any future sunset announcement.
• Not actually a Microsoft shop? → step back and pick by ecosystem: all eleven AI certs compared.

Keep reading: AI-900 vs AI-901: what changed · The 2-week AI-901 study plan · AWS’s version of this same reshuffle

HOW TO // AI is not affiliated with or endorsed by Microsoft. AI-900, AI-901, AI-102, AI-103, DP-100, AI-300, and Azure are certifications and trademarks of Microsoft Corporation; we reference them descriptively. Retirement dates per Microsoft’s published certification updates; AI-102’s status and AI-103’s positioning verified against Microsoft’s own pages (AI-103 is presented as a new credential, not an AI-102 replacement). Verify current details with Microsoft Learn.