AI-200 Exam Guide: The Four Skill Areas, the Format, and What the Official Learning Path Covers

The Azure AI Cloud Developer Associate certification rests on one exam, AI-200. It asks you to build and run AI back ends on Azure: containers, vector data stores, messaging and telemetry. Expect 120 minutes, four weighted skill areas, a pass mark of 700 on a scaled score, and English as the only language on offer.

What AI-200 covers

The study guide publishes four skill areas with a weighting for each. Here they are with the services named underneath them:

Area	Weight	Services named in the outline
1 · containers	20–25%	Azure Container Registry with ACR Tasks · Azure App Service · Azure Container Apps with KEDA · Azure Kubernetes Service (AKS)
2 · data services	25–30%	Azure Cosmos DB for NoSQL, including the change feed · Azure Database for PostgreSQL, with pgvector · Azure Managed Redis
3 · messaging and serverless	20–25%	Azure Functions · Azure Service Bus · Azure Event Grid
4 · security and observability	20–25%	Azure Key Vault · Azure App Configuration · OpenTelemetry · KQL queries over logs and metrics

Microsoft's own titles are longer. Area 1 is Develop containerized solutions on Azure. Area 2 is Develop AI solutions, a title the outline finishes with the words by using Azure data management services. Area 3 is Connect to and consume Azure services. Area 4 is Secure, monitor, and troubleshoot Azure solutions.

The shape is flat. One area carries 25–30% of the marks and the other three carry 20–25% each. So there is no small area you can safely skip. That is a planning observation of ours, not Microsoft's advice. It matters most for area 4. Secrets, configuration and telemetry are the parts developers tend to leave until last, and here they are worth as much as containers.

The outline arrives with two warnings attached, and both matter. Microsoft presents the sub-bullets as examples of the way a skill gets tested, and reserves the right to ask about neighbouring subjects. It also says most questions are about features that have reached general availability, with preview features possible where they are already widely used. So read the outline as a map of what gets tested, not as a full contents page. The page stamps its own last update as 5 May 2026.

The format: 120 minutes, and a 700 that is not 70%

The certification page gives you 120 minutes. The exam is proctored. Microsoft also warns that you may get interactive components to complete, which is worth knowing before the clock starts: those parts take longer than a multiple-choice question.

The pass mark is 700, and it is the most misread number on the page. The scoring page scores technical exams from 1 to 1,000. Because that result is scaled, it does not track the share of points you collected. Microsoft sets the bar from the skills under test and from the difficulty of the question set you were given. Easier sets therefore demand more points; harder sets demand fewer.

Three rules on that page are worth acting on. First, nothing is deducted for a wrong answer, so leave no item unanswered. Second, multi-part items are usually marked per component, which means partial credit counts. Third, a few items on any exam are unscored trials, invisible to you, so every question deserves a real answer.

Microsoft publishes no question count for AI-200. Other sites print exact numbers anyway. Those are estimates, and we would rather leave the field empty than fill it with a guess. The number you can actually plan around is 120 minutes.

Booking is already open: the certification page carries a Schedule through Pearson Vue link for this exam. Two details sit beside it. Microsoft asks you to register under a personal account rather than one issued by an employer or a school. Exam records held on an organizational account are lost once you leave it. On price, no worldwide number exists. The amount follows the country or region where you are proctored, and it appears during booking. A failed attempt can be retaken after 24 hours.

AI-200 is offered in English only

The certification page lists one language for this exam: English. That is a real constraint rather than a footnote, and it is easy to miss, because the training that surrounds the exam is not limited the same way. The matching course, AI-200T00-A, is published in twelve languages, with Japanese, German, Korean, French and Portuguese (Brazil) among them. The exam behind it is English.

One official policy softens this. The study guide says that when an exam is not available in the language you would prefer, you may ask for an extra 30 minutes to complete it. Request it when you book, not on the day.

Who AI-200 is for

AI-200 carries no prerequisite exam or certification. Microsoft publishes an audience profile instead, and this one is unusually concrete for a Microsoft page. Most audience profiles describe a role. This one names technologies you are expected to know already.

The candidate it describes works on every phase of putting AI solutions on Azure, with the weight on back-end services and components. The same person supports the whole development lifecycle, from gathering requirements through design, build, deployment, security and monitoring.

Then comes the list, and it is worth reading slowly. Microsoft expects you to be proficient in Python programming. It expects you to be implementing containerized applications on Azure. It names vector databases outright. It adds Azure data management services, Azure messaging and eventing, Azure monitoring and troubleshooting, and the Azure SDKs along with the third-party SDKs people use with them.

Those items are assumed knowledge rather than exam preparation. AI-200 is rated Intermediate and tied to the developer role, and the profile assumes you already write code against Azure services. If you have never built a container image or queried a vector index, that groundwork is bigger than the exam preparation itself. We say that as our own reading of the profile, not as Microsoft's wording.

What the official learning path actually contains

Microsoft publishes a free learning path for this exam through course AI-200T00-A, which is also what the "get trained" link points at. In building our AI-200 materials we walked all of it and counted what is there: nine learning paths, 24 modules, and 194 units. Of those units, 24 are hands-on exercises — exactly one per module. Those numbers are ours, not Microsoft's.

The nine paths, and the modules under each:

Learning path	Modules	Exam area
Implement container application hosting on Azure	2	1
Deploy and manage apps on Azure Container Apps	3	1
Deploy and monitor applications on Azure Kubernetes Service	3	1
Develop AI solutions with Azure Cosmos DB for NoSQL	3	2
Develop AI solutions with Azure Database for PostgreSQL	3	2
Enhance AI solutions with Azure Managed Redis	3	2
Integrate backend services for AI solutions	3	3
Manage application secrets and configuration for AI solutions	2	4
Observe and troubleshoot apps on Azure	2	4

Comparing the paths with the weightings shows two things. Area 2 gets nine modules for 25–30% of the marks, which is proportionate. Area 4 gets four modules for 20–25% — Key Vault, App Configuration, OpenTelemetry and KQL — so the training material is thinner there than the marks are. Plan your own reading accordingly.

We also found one module that maps to nothing in the published outline. The Redis path teaches event messaging with Azure Managed Redis, and the skills measured list mentions only two Redis subjects: data operations, and vector indexing for similarity search. That module is useful engineering. It is not, on the published evidence, examined material.

Two things Microsoft's own pages get wrong or leave out

There is no official practice assessment yet. Microsoft normally publishes a free practice assessment per exam, and many study plans start there. For AI-200 the certification page says the practice assessment is "not currently available". It also gives the usual timing: roughly 8 weeks after an exam leaves beta and reaches general availability.

Read that carefully, because it describes Microsoft's usual schedule and nothing else. It does not state where AI-200 sits today, and no date appears on the page. We are not going to infer one for you. Check the certification page yourself before you build a plan around a free practice assessment, because that line changes without an announcement.

What does exist for free is the exam sandbox. Spend twenty minutes in it before exam day. It uses the same software and the same item formats you will meet on the day. Be clear about what it is not, though. No AI-200 subject matter appears in it.

The study guide's "Study resources" table points at the wrong exam. This one is worth checking for yourself. Below the skills measured list, the AI-200 study guide carries a block of documentation links. On the copy we captured on 15 August 2026, those links included Blob Storage, Queue Storage, Event Hubs, API Management, Azure Container Instances and Microsoft Entra ID. Not one of those appears anywhere in AI-200's own outline.

The gaps run the other way too. That same table has no link for Azure Kubernetes Service, Azure Container Apps, Azure Database for PostgreSQL, Azure App Configuration, OpenTelemetry or KQL. All six are named in the skills measured list. It also links Azure Cache for Redis, while the outline asks about Azure Managed Redis.

Our reading is that this is a copy-paste from AZ-204, Microsoft's Azure Developer Associate exam, whose documentation set that list matches closely. We report it because it costs a candidate real time: anyone working through that table is reading for a different exam. Use the skills measured list as your scope, and treat the resources block below it as unreliable until Microsoft fixes it.

Does the certification expire?

Yes. Fundamentals certifications never expire, but this is an associate one, and the renewal page places associate, expert and specialty certifications on a yearly cycle. Azure AI Cloud Developer Associate therefore needs renewing every twelve months.

The renewal itself is a far smaller job than the exam, and it carries no fee. You sit a short online assessment through Microsoft Learn, open book and without a proctor, at any point during the six months that run up to your expiry date. Attempts inside that window are unlimited. A pass pushes your expiry date out by another year. Given the pace of change across Azure's AI services, we read that yearly check as a benefit rather than a burden.

How to prepare

Begin with the two official pages Microsoft keeps current: the study guide for the outline, and the certification page for the format, the language and the policies. Then work the nine learning paths, and actually do the 24 exercises. On an exam this practical, reading about a change feed processor or a KEDA scaling rule is not the same as having configured one.

Reading teaches you what each service does. The exam asks which one you would pick, and it asks under a clock. A managed identity or a stored secret. A Service Bus topic or an Event Grid subscription. A vector index tuned for recall or one tuned for latency. That is the gap practice questions close. Our AI-200 practice exams track this outline, and each question carries a link to the official Microsoft page it came from. No dumps, ever.

If you want the schedule rather than the scope, we wrote an AI-200 study plan. If you are choosing between this exam and the other Azure AI developer credentials, that comparison is here.

Independent study material. ciply.io is not affiliated with, endorsed by, or sponsored by Microsoft.