Half of Microsoft's AI certifications are retiring by June 2026. Three exams gone, six replacements launching in overlapping waves, a brand new executive tier nobody saw coming, and zero clear guidance from Microsoft on what to do if you are mid-study right now. I have been through 20 years of Microsoft certification cycles. This one is the messiest transition I have seen.
TL;DR
AI-900, AI-102, and DP-100 retire by June 30, 2026. The replacements (AI-901, AI-103, AI-200, AI-300) launch in beta between April and May. For developers, take AI-102 before June or wait for AI-103 GA. For ML engineers, grab the AI-300 beta at 80% off. For architects, AB-100 is the hardest and most valuable cert in the new lineup. For everyone else, AB-730 is the $99 baseline that proves you can actually use AI at work.
The Great Reset: What Is Retiring and What Replaces It
Microsoft announced the retirements in a Tech Community blog post that buried the timeline details. Here is the actual transition map.
| Retiring Exam | Retirement Date | Replacement | Replacement Status |
|---|---|---|---|
| AI-900 (Azure AI Fundamentals) | June 30, 2026 | AI-901 (Azure AI Fundamentals v2) | Beta April 2026, GA expected June 2026 |
| AI-102 (Azure AI Engineer Associate) | June 30, 2026 | AI-103 (Azure AI App and Agent Developer) | Beta April 2026, GA expected June 2026 |
| DP-100 (Azure Data Scientist Associate) | June 1, 2026 | AI-300 (MLOps Engineer Associate) | Beta now, GA expected May 2026 |
Notice the problem. DP-100 retires June 1 but AI-300 might not be GA until May. AI-900 retires June 30 but AI-901 enters beta in April. There is a window where the old exam is dead and the new one is not fully live. If you are planning to certify, timing matters.
There is also a second wave of entirely new certifications that did not exist before:
| New Exam | Level | Cost | Status |
|---|---|---|---|
| AB-730 (AI Business Professional) | Fundamentals | $99 | Live |
| AB-731 (AI Transformation Leader) | Fundamentals | $99 | Live |
| AB-900 (M365 Copilot and Agent Administration) | Fundamentals | $99 | Live |
| AB-100 (Agentic AI Business Solutions Architect) | Expert | $165 | Live |
| AI-200 (Azure AI Cloud Developer) | Associate | $165 | Beta April 2026, GA expected July 2026 |
| SC-500 (Cloud and AI Security Engineer) | Associate | $165 | Beta May 2026, GA expected July 2026 |
That is 9 new or replacement exams. Microsoft now has more AI-specific certifications than AWS and Google combined. Whether that is a good thing depends on which ones you pick.
The Certifications That Actually Matter
I am ranking every new Microsoft AI certification by practical career value. Not marketing value. Not resume-padding value. What actually changes your job prospects or daily work.
Tier 1: Take These
AI-300 (MLOps Engineer Associate) - The real engineer cert.
This replaces DP-100 and it is a genuine upgrade. The old Data Scientist Associate exam tested you on Azure ML Studio notebooks and scikit-learn pipelines. AI-300 tests MLOps infrastructure: GitHub Actions for model CI/CD, Bicep templates for Azure AI Foundry deployment, model monitoring, RAG pipeline optimization, and generative AI operations. This is what ML engineering actually looks like in 2026.
Study time: 50-70 hours. The beta is open right now with Microsoft's standard 80% beta discount for the first 300 testers (check learn.microsoft.com for the current beta promo code, since codes change per cohort). At roughly $33 instead of $165, this is the best deal in the current certification cycle.
If I were starting today as a data scientist or ML engineer, this is my first cert. No contest.
AB-100 (Agentic AI Business Solutions Architect) - The hardest one.
This is expert-level and requires an active associate certification as a prerequisite. The exam covers multi-agent AI system design across Dynamics 365, Power Platform, Copilot Studio, and Azure AI. Not many people have deep knowledge across all four platforms. That small candidate pool is exactly what makes this cert valuable.
Study time: 60-80 hours. Cost: $165. If you work at the intersection of D365 and AI and you are already building agentic governance into Power Platform projects, this is the cert that separates you from everyone else. We took this one apart domain by domain in the AB-100 decode: what the Deploy-heavy weighting signals, the prerequisite gate that catches PL-600 holders, and a three-week prep plan.
AI-103 (Azure AI App and Agent Developer Associate) - The developer default.
This replaces AI-102 with a stronger focus on Azure AI Foundry, agentic patterns, and Copilot Studio integration. If you build AI-powered applications on Azure, this will be your primary credential starting mid-2026.
The problem: it is not GA yet. Beta is expected April 2026. If you need a cert now, take AI-102 before June 30. Your AI-102 credential stays valid for one year after the retirement date, and the knowledge overlaps significantly with AI-103.
Tier 2: Situationally Valuable
AB-730 (AI Business Professional) - The baseline everyone should have.
This is the $99, 15-20 hour certification that proves you can use generative AI tools productively. Prompt engineering, business content drafting, AI-assisted analysis. It is not technical. That is the point. Every knowledge worker who uses M365 Copilot daily should have this. If you manage a team, make this the standard.
AB-900 (Copilot and Agent Administration Fundamentals) - For IT admins.
Heavy on Microsoft Purview, DLP policies, DSPM for AI, and SharePoint oversharing prevention. If you are the person responsible for rolling out Copilot licenses and making sure nobody accidentally shares confidential data through AI, this is your cert. Niche but critical.
SC-500 (Cloud and AI Security Engineer Associate) - The security crossover.
Beta expected May 2026. If it covers LLM application security, Purview DSPM for AI, and Copilot data protection as expected, this fills a gap that nobody else in the market has addressed. Security engineers who understand AI-specific attack surfaces are in short supply.
Tier 3: Strategic But Not Urgent
AB-731 (AI Transformation Leader) - The one nobody is talking about.
This is the first Microsoft certification explicitly designed for executives. Not technical people. Not developers. Directors, VPs, and department heads who need to understand AI business value, responsible AI governance, and adoption strategy.
The exam covers three domains: AI business value assessment (35-40%), Microsoft AI capabilities (35-40%), and implementation strategy (20-25%). Study time is 20-30 hours. Cost is $99.
Here is why this matters even if you are technical. Your CTO or VP of Engineering probably cannot articulate what Microsoft's AI stack actually does. They know "Copilot" and maybe "Azure OpenAI" and that is it. AB-731 gives them a structured framework for AI investment decisions. If you are trying to get AI initiatives funded, sending your leadership team through AB-731 prep materials removes the biggest blocker: executives who do not understand what they are approving. The Power Platform governance baseline is a good primer on why governance matters before you hand leadership the AB-731 study guide.
AI-901 (Azure AI Fundamentals) - Wait for it.
Direct replacement for AI-900. If you have zero AI background and want a fundamentals credential, wait for this instead of rushing through AI-900 before retirement. The content will be updated for 2026 tooling. Beta expected April, GA expected June.
AI-200 (Azure AI Cloud Developer Associate) - TBD.
Beta expected April 2026. The differentiation from AI-103 is not clear yet. This appears to be a broader cloud developer cert with AI integration focus, but until the study guide drops, I cannot recommend it over AI-103.
Do Certifications Actually Affect Your Career?
Here is the real question. Do these credentials translate to money and job offers?
| Role | Salary Range (USD, 2026) | Cert Premium vs Uncertified |
|---|---|---|
| AI/ML Engineer (certified) | $120,000 - $200,000 | +15-25% |
| Azure AI Engineer (AI-102/AI-103) | $110,000 - $165,000 | +12-20% |
| MLOps Engineer (AI-300 target) | $130,000 - $180,000 | +15-25% |
| Data Scientist (DP-100/AI-300) | $115,000 - $170,000 | +12-20% |
Certified professionals earn 15-25% more than uncertified peers in equivalent roles. That is not my opinion. That is what the salary data shows across multiple surveys.
But here is the nuance. Entry-level AI hiring dropped 73.4% in 2025. The market is not looking for people who can pass a fundamentals exam. It is looking for mid-to-senior practitioners who can ship production AI systems. A fundamentals cert alone does not move the needle. An associate or expert cert stacked on top of real project experience does.
The ROI math on AI-300: $165 exam fee (or $33 at beta pricing), 50-70 hours of study, and an MLOps Engineer role pays $130,000-$180,000. If certification moves you even one salary band, the return on 60 hours of study time is measured in tens of thousands of dollars per year.
Building AI solutions on Azure?
Start with the architecture patterns that actually work in production. What are Azure AI Services in 2026 maps the full service catalog, and Copilots vs Custom AI covers build-vs-buy decisions for real projects. Follow Alex on LinkedIn for architecture deep dives.
Microsoft vs AWS vs Google: The Honest Comparison
| Microsoft | AWS | Google Cloud | |
|---|---|---|---|
| AI-specific certs | 10+ | 2 | 2 |
| Fundamentals tier | AI-901, AB-730, AB-731, AB-900 | AI Practitioner ($150) | None |
| Developer/Engineer tier | AI-103, AI-200, AI-300 | ML Specialty ($300) | Professional ML Engineer ($200) |
| Architect/Expert tier | AB-100 | None | None |
| Security + AI cert | SC-500 | None | None |
| Business/Executive cert | AB-730, AB-731 | None | None |
| Cloud market share | ~24% | ~31-33% | ~10-12% |
| Salary range | $110K - $200K | $130K - $165K | $140K - $170K |
Microsoft has the broadest certification portfolio by far. AWS has two AI certs. Google has two. Microsoft has ten and counting.
Does breadth mean quality? Not automatically. But it does mean Microsoft is the only vendor where a business analyst, a developer, an ML engineer, an architect, a security engineer, and an executive can all get role-appropriate AI credentials from the same platform. AWS and Google do not have a cert for your CTO. Microsoft does.
The salary numbers are interesting. Google Cloud ML Engineer certification correlates with the highest single-cert salary bump at roughly 25%. AWS ML Specialty appears in 40% more job postings than competitors for ML-specific roles. Microsoft Azure AI sits in the middle on both metrics but has 3x more certification paths.
My honest take: if you work in a Microsoft shop (D365, Power Platform, M365, Azure), Microsoft certs are the obvious choice. If you are cloud-agnostic and optimizing purely for salary, Google's ML Engineer cert has the best ROI per exam. If you want the most job postings to apply to, AWS ML Specialty gives you volume.
Most competitive professionals hold one vendor cert plus one vendor-neutral credential (Stanford ML Specialization, fast.ai, or Deep Learning Specialization). The vendor cert proves you can deploy. The theory cert proves you understand why.
The Free Credentials Everyone Ignores
Microsoft offers Applied Skills assessments. These are free, hands-on lab credentials. Not traditional exams. You complete a real task in a sandbox environment and earn a credential if you pass.
Current AI-related Applied Skills:
- Create agents in Microsoft Copilot Studio - build and deploy a working agent
- Develop generative AI apps with Azure OpenAI and Semantic Kernel - code-level assessment
- Generate reports with AI research agents - practical AI application
- Prepare security and compliance to support Microsoft 365 Copilot - governance skills
These do not carry the weight of a full certification on a resume. But they are free, they prove hands-on ability (not just exam knowledge), and they take 2-4 hours each. If you are building a Copilot-powered solution and want to validate your skills without spending $165, start here.
What I Would Actually Do
If I were planning my certification path right now, here is exactly what I would do based on career stage.
Early career (0-3 years):
Wait for AI-901 (April beta). Pass it. Then immediately start studying for AI-103. Those two credentials plus a GitHub portfolio of actual projects is enough to get interviews. Skip AB-730 unless your employer pays for it. Your time is better spent building things.
Mid-career developer (3-7 years):
Take AI-102 right now, before June 30. The content overlaps with AI-103 and you get a valid credential immediately instead of waiting for a beta. Then stack the Semantic Kernel Applied Skills credential on top. If you are building AI-powered apps on Azure, the combination of AI-102 + hands-on Semantic Kernel experience puts you ahead of 90% of candidates.
Senior engineer or ML specialist (7+ years):
AI-300 beta. Today. The standard 80% beta discount is available for the first 300 testers (check learn.microsoft.com for the current code, since beta promo codes rotate per cohort). This is the most technical cert in the new lineup and it validates what you probably already do: MLOps pipelines, model deployment, monitoring, and generative AI operations. At roughly $33, the risk-reward ratio is absurd.
Architect:
AB-100, but only after you have a qualifying associate cert (that is a hard prerequisite). If you already hold AI-102 or any of the 14 qualifying associate certs, start studying now. If you are building agentic solutions across D365 and Azure AI, this is the credential that proves cross-platform architecture competence. Small candidate pool means high differentiation.
Manager or director:
AB-731 (AI Transformation Leader). The $99 you spend and 20-30 hours you invest will give you the vocabulary and framework to evaluate AI initiatives, set governance policy, and make funding decisions without depending on your technical team to translate everything. Then send your entire team through AB-730.
IT admin managing Copilot rollout:
AB-900 first, then the Applied Skills credential for Copilot security and compliance. These two together cover the governance side of AI deployment that most organizations are ignoring until something goes wrong.
The Timeline That Matters
Here is what to do month by month.
March 2026: Decide your path. If you are taking AI-102 or DP-100, register now. If you want the AI-300 beta discount, apply before slots fill.
April 2026: AI-901 and AI-103 betas open. Early adopters can take these at reduced cost. Beta scores take longer to process (8-10 weeks).
May 2026: SC-500 beta opens. AI-300 expected GA. Last full month to take DP-100 before June 1 retirement.
June 2026: AI-900, AI-102, DP-100 all retired by June 30. AI-901 expected GA. AI-103 expected GA. The old era is officially over.
July 2026: AI-200 and SC-500 expected GA. The full new certification portfolio is in place.
The window between now and June is the most important. After June, the path is clear. Right now, you have a choice between the known (old exams with defined study guides) and the unknown (new exams with beta-quality materials). I would take the known exam now and the new exam later. Having both on your profile shows continuity.
The certification market is shifting from "prove you know AI concepts" to "prove you can deploy AI systems." Every new exam in the 2026 lineup emphasizes hands-on skills, production deployment, and operational concerns over theory. That is the right direction. Study accordingly.
Microsoft AI Builder Series
- AI Certifications in 2026 - Which ones actually matter
- What Are Azure AI Services in 2026 - The full service map
- Copilots vs Custom AI - When to build and when to buy
AZ365.ai - Azure and AI insights for architects building on Microsoft. Follow Alex on LinkedIn for architecture deep dives.
This article was originally published at az365.ai. I'm Alex Pechenizkiy, an Azure and Power Platform solutions architect writing honest, vendor-neutral analysis of the Microsoft AI stack. More at az365.ai.
