The Microsoft AI Solution Architect Certification Path in 2026

If your role is to design AI solutions across the Microsoft stack, the certification path you may have bookmarked a year ago no longer holds. The exams that used to anchor the architect track are retiring in 2026, and the Expert architect credential itself has been re-cut around agentic AI. This is the current route: the live credentials, in order, with the honest caveats an architect should carry.

One scope note up front. This is the path to an AI solution architect on the Microsoft data and AI stack. If you are after the Azure infrastructure architect track (the AZ-900 to AZ-305 spine), that is a different path and is not covered here.

Cert facts trace to official Microsoft Learn credential pages, linked inline, and reflect the June 2026 check. New and beta exams are marked as such; verify status on the linked page before you book.

The fastest credible path, in one line

AI-901 → AI-103 → AB-100.

Three steps that end on the only live Expert architect credential in the current catalog. The stretch version adds breadth (below).

Who this path is for

Entry: AI-901 Azure AI Fundamentals

A breadth architect needs a shared vocabulary across AI workloads, responsible AI, and, critically for this role, generative AI, Copilot, RAG, and grounding. The Azure AI Fundamentals certification, earned through the AI-901 exam, is the live entry point and adds exactly that generative-AI content over the older AI-900.

Earn the certification through AI-901, not AI-900: the AI-900 exam retires June 30, 2026, while the certification continues through AI-901. It does not expire. One note for partner-minded architects: Fundamentals are generally excluded from Solutions Partner skilling scores, so earn this for your own grounding, not for partner points; verify the current partner rules.

Core: the architect’s working surface

AI-103 Azure AI Apps and Agents Developer

AI-103 is the load-bearing core credential. It covers planning and managing an Azure AI solution plus generative AI and agentic solutions, computer vision, text analysis, and information extraction, which is the design surface an architect must be able to speak to. It is the forward-safe successor to AI-102 (retiring June 30, 2026), so it is the only future-proof choice here. It expects Python app-development experience.

Hedge it honestly: AI-103 is recent enough that real prep-time and difficulty evidence does not exist yet, so treat any “how hard is it” estimate as provisional, and verify whether it counts toward partner skilling before you rely on that.

AI-300 Machine Learning Operations Engineer (breadth)

AI-300 gives the architect the ML-lifecycle and MLOps breadth that the retired DP-100 used to supply, now reframed around MLOps, GenAIOps, and gen-AI quality and observability. Those are exactly the operational concerns an AI-solution designer has to architect around. It is the explicit replacement for DP-100, and Microsoft added it as a skilling option for the AI Platform on Azure specialization, which makes it the clearest partner-designation hook in this path. It is also new, so its difficulty is unproven.

DP-600 Fabric Analytics Engineer (data breadth)

DP-600 adds the data and analytics dimension (semantic models, lakehouses, warehouses; SQL, KQL, DAX) so the architect can design end-to-end solutions that sit on a data platform, not just model APIs. A timely cost lever: Microsoft has promoted Fabric certifications through Data Days 2026 with free exam vouchers, announced for June 15 to August 7, 2026, so check the current offer before booking.

GH-300 GitHub Copilot (governance lens)

A breadth architect increasingly owns the Copilot governance story (content exclusion, intellectual property, licensing, responsible AI), which is GH-300’s hardest and most architect-relevant content. Lived experience splits on difficulty: straightforward if you use Copilot daily, but the enterprise-policy and heavily-weighted responsible-AI sections require real study.

Advanced: AB-100, the Expert architect credential

AB-100, Agentic AI Business Solutions Architect Expert, is the terminal credential and the only live Expert cert in the catalog that fits “designs AI solutions.” Microsoft positions it as the forward path for the retiring PL-600, though it was re-cut around agentic AI rather than being a like-for-like replacement. Its skills map directly to the role: plan, design, and deploy AI-powered business solutions.

Three things to plan for. It is current and earnable, but newly introduced, so treat its difficulty and employer recognition as still forming. It sits behind a gate: you must already hold one of the qualifying Associate certifications listed on its credential page, and earning AI-103 in the core step is the cleanest way to satisfy that gate. And because Expert certs are typically excluded from partner skilling points, treat AB-100 as role credibility, and earn your partner points from the Associate layer (AI-300 and AI-103).

For the habit of keeping a plan like this current as the portfolio keeps reshuffling, see how a Power Platform architect keeps up with Microsoft’s AI certification changes.

What to skip, and why

The old architect spine leaned on exams that are retiring or absent from the current AI catalog. Route around them:

• AI-900 exam retires June 30, 2026. Earn the certification through AI-901.
• AI-102 retires June 30, 2026. Sit AI-103.
• DP-100 is retired. Use AI-300 for the MLOps breadth.
• PL-600 retires June 30, 2026. AB-100 is the recommended forward path, not a one-to-one replacement.
• PL-200 retires August 31, 2026. If your work leans functional and low-code, that track now points at the Intelligent Applications Builder Associate (AB-410), which is in beta.

The rule for anything with a future retirement date: only start it if you can realistically pass before the deadline, because the certification cannot be earned once the exam retires. For the wider Power Platform family, see our Power Platform certification path guide.

Do not quote a salary number to yourself

Architect salaries you will see attached to these exams are role-level figures from job boards and training vendors, not cert-attributed outcomes, and AB-100 is far too new to have any credible salary signal. Earn this path for the design capability and the architect standing it builds, not for a percentage on a marketing page.

Train for the path

The credentials mark the route; the capability is what the role actually rewards. Citizen Development Academy teaches the Power Platform, Copilot, and AI capabilities behind these credentials, built and delivered by Microsoft Certified Trainers. Explore the courses to build that foundation. Citizen Development Academy is a Microsoft Learning Partner; partner status governs the standards under which we deliver Microsoft training, and Microsoft does not endorse this article or any individual certification plan.

FAQ

Is this the Azure infrastructure architect path?

No. This is the AI solution architect path on the Microsoft data and AI stack, ending at AB-100. The Azure infrastructure architect track (AZ-900 to AZ-305) is a separate path and is not part of this route.

What replaced AI-102, DP-100, and PL-600 for architects?

AI-103 replaces AI-102, AI-300 replaces DP-100, and AB-100 is the recommended forward path for PL-600 (re-cut around agentic AI, not a one-to-one swap). Build toward the successors, not the retiring codes.

Do I have to earn all four core credentials?

No. The minimum credible path is AI-901, then AI-103, then AB-100. AI-300, DP-600, and GH-300 add the breadth the architect role rewards; add them where they fit your work, and use the Data Days 2026 voucher window for the Fabric exam while it lasts.

Do these certifications expire?

AI-901 (Fundamentals) does not expire. Associate certifications typically renew every 12 months through a free online assessment on Microsoft Learn; confirm terms on each credential page, since the newest ones (AI-103, AI-300) can still change. The AB-100 Expert certification also renews every 12 months.