AI-103 vs AI-200

AI-103 vs AI-200: How the Two Azure AI Associate Exams Differ

Updated July 2026•12 min read

AI-103 (Azure AI App and Agent Developer) and AI-200 (Azure AI Cloud Developer) are two separate Microsoft Associate exams, both now generally available. Here's how they differ, who each one is for, and which to take first.

Quick Answer

AI-103 and AI-200 are not the same exam. AI-103 is the application-focused Azure AI Engineer credential — Foundry agents, RAG, prompt engineering, AI services. AI-200 is the backend-platform-focused Azure AI Cloud Developer credential — containerized hosting, AI-ready data services (Cosmos DB, PostgreSQL with pgvector, Azure Managed Redis), event-driven integration, and securing/monitoring AI solutions on Azure.

Most application developers should take AI-103 first. Most backend and platform engineers should take AI-200 first. Many senior AI engineers eventually earn both.

The Common Confusion

A lot of searches for "exam AI-200" come from candidates who assumed AI-200 was the GA replacement for AI-103. That assumption is wrong. Microsoft treats AI-103 and AI-200 as two distinct Associate exams targeting different roles:

  • • AI-103 — Azure AI App and Agent Developer Associate (opened beta April 2026, now GA)
  • • AI-200 — Azure AI Cloud Developer Associate (opened beta May 2026, now GA)

Both stay separate now that they've reached GA. Each delivers its own credential, has its own skills outline, and follows its own annual-renewal cycle. They are complementary, not alternatives.

Side-by-Side Comparison

AttributeAI-103AI-200
Full credentialAzure AI App and Agent Developer AssociateAzure AI Cloud Developer Associate
Launched betaApril 2026May 2026
StatusGenerally availableGenerally available
LevelAssociateAssociate
Primary focusBuilding AI apps and agents on Azure AI FoundryBuilding the backend infrastructure behind AI solutions on Azure
Core platformAzure AI Foundry, Azure OpenAI, AI servicesContainer Apps/AKS, Cosmos DB, PostgreSQL (pgvector), Azure Managed Redis, Service Bus/Event Grid
Heaviest domainGenerative AI and agents (30–35%)AI data management services — Cosmos DB, PostgreSQL, Redis (25–30%)
Code-heavy?Yes — Python/C# Foundry SDK, OpenAI SDKYes — Azure SDKs, containerized backend services
Best forAI/ML developers, Copilot Studio devs, Azure OpenAI usersBackend developers, platform engineers, cloud engineers building AI infrastructure
Beta price (2026 launch)~$99 USD~$99 USD
Current price$165 USD$165 USD
Length120 min, 40–60 questions100 min, ~40 questions
Passing score700/1000700/1000
PrerequisitesNone formal — Python/C# recommendedNone formal — AI-103 or strong Azure backend/container experience recommended

What AI-103 Actually Covers

AI-103 validates that you can build AI applications and agents on Azure AI Foundry. The five domains:

  • • Plan and manage an Azure AI solution (25–30%) — responsible AI, content safety, RBAC for AI services, private networking, monitoring
  • • Implement generative AI and agentic solutions (30–35%) — Foundry agents, prompt engineering, RAG with AI Search, function calling, prompt flow, evaluation, fine-tuning
  • • Implement computer vision solutions (10–15%) — Image Analysis, Custom Vision, Face, Content Understanding, multimodal vision
  • • Implement text analysis solutions (10–15%) — Language service, Translator, Speech, Realtime API
  • • Implement information extraction (10–15%) — AI Search, vector and hybrid search, Document Intelligence, knowledge mining

If you write code against Azure OpenAI, the Foundry SDK, or any Azure AI service today, this is the exam that maps to your day-to-day work.

What AI-200 Actually Covers

AI-200 zooms out from "I built an AI app" to "I build and operate the Azure backend an AI application depends on." The confirmed exam blueprint has four domains:

  • • Develop containerized solutions on Azure (20–25%) — Azure Container Registry and Container Registry Tasks, deploying containers to App Service, Azure Container Apps with KEDA event-driven autoscaling, and AKS manifest deployments
  • • Develop AI solutions by using Azure data management services (25–30%, the largest domain) — Cosmos DB for NoSQL vector similarity search and change feed processing, Azure Database for PostgreSQL with pgvector and RAG patterns, and Azure Managed Redis caching and vector indexing
  • • Connect to and consume Azure services (20–25%) — Azure Service Bus (dead-letter queues, topics, subscriptions), Azure Event Grid, and Azure Functions
  • • Secure, monitor, troubleshoot Azure solutions (20–25%) — Azure Key Vault, Azure App Configuration, OpenTelemetry SDKs, and KQL for log/metric analysis

If your job title is Backend Developer, Cloud Engineer, or Platform Engineer connecting containerized services to vector databases, message queues, and event-driven pipelines in an organization running AI in production, AI-200 is the exam that maps to your role. See the full AI-200 domain breakdown and practice exam for exact percentages and sample questions.

Which Should You Take First?

Take AI-103 first if you are:

  • • A Python or C# developer shipping AI features in apps
  • • A Copilot Studio or Azure AI Foundry developer building agents
  • • An ML engineer moving from notebooks to production Foundry deployments
  • • A data scientist looking to formalize Azure AI development skills
  • • An existing AI-102 holder updating to the Foundry-centric blueprint

Take AI-200 first if you are:

  • • A backend developer building the containerized services that host AI workloads
  • • An engineer implementing vector search or RAG patterns with Cosmos DB or PostgreSQL/pgvector
  • • A platform engineer running Container Apps or AKS for AI-adjacent workloads
  • • An engineer wiring up Service Bus, Event Grid, or Functions around AI services
  • • A cloud engineer responsible for securing and monitoring AI solutions (Key Vault, App Configuration, OpenTelemetry, KQL)

Take both (most senior AI engineers eventually do):

  • • You build AI apps and you also own the backend infrastructure they run on
  • • You are positioning for a Lead Azure AI Engineer or Principal AI Architect role
  • • You want the broadest Azure AI credential coverage Microsoft offers in 2026

Sequencing: A Realistic 12-Week Plan for Both

If you decide to take both AI-103 and AI-200, a focused 12-week plan looks like this:

Weeks 1–6 — AI-103 Preparation

Follow the full 4-week AI-103 study plan (see our AI-103 Study Guide), then add 2 weeks of full-length mock exams and targeted review on Domain 2 (generative AI and agents, 30–35% of the exam). Book AI-103 in week 6.

Week 6 — Take AI-103

Scores for a scheduled exam typically release within a few days; if you take it during an active beta window, scoring can take 1–2 weeks — but the credential is awarded as of your exam date either way. Move on to AI-200 prep while waiting.

Weeks 7–11 — AI-200 Preparation

Pivot to backend platform topics. Spend week 7 on Container Apps, KEDA event-driven autoscaling, and AKS manifest deployments. Weeks 8–9 on the largest domain — Cosmos DB vector search, Azure Database for PostgreSQL with pgvector and RAG patterns, and Azure Managed Redis. Week 10 on Service Bus, Event Grid, and Azure Functions. Week 11 on Key Vault, App Configuration, OpenTelemetry/KQL, and mock exams.

Week 12 — Take AI-200

AI-200 is now generally available. The AI-103 foundation from weeks 1–6 transfers directly to AI-200's responsible-AI and Azure-fundamentals topics — focus your final prep on what AI-200 covers beyond AI-103: containers, vector-database data patterns, and messaging.

Frequently Asked Questions

Are AI-103 and AI-200 the same exam?

No. AI-103 (Azure AI App and Agent Developer Associate) and AI-200 (Azure AI Cloud Developer Associate) are two separate Microsoft exams, both now generally available. They cover complementary slices of the Azure AI developer role. AI-103 is application-focused (Foundry agents, RAG, prompt engineering, AI services). AI-200 is backend-platform-focused (containerized hosting, AI-ready data services, event-driven integration, and securing/monitoring AI solutions on Azure). Each exam delivers a distinct credential.

What is the difference between AI-103 and AI-200?

AI-103 validates that you can build AI apps and agents on Azure AI Foundry — heavy focus on generative AI, agents, RAG, prompt flow, Document Intelligence, Content Understanding, and responsible AI safeguards. AI-200 validates that you can build the backend infrastructure behind AI solutions on Azure — containerized hosting on Container Apps and AKS, AI-ready data services (Cosmos DB vector search, Azure Database for PostgreSQL with pgvector, Azure Managed Redis), event- and message-based integration via Service Bus and Event Grid, and securing/monitoring solutions with Key Vault, App Configuration, OpenTelemetry, and KQL. AI-103 is "I build AI apps." AI-200 is "I build and operate the cloud infrastructure those AI apps run on."

Should I take AI-103 or AI-200 first?

Take AI-103 first if you are a developer building AI applications today — Copilot Studio developers, Azure OpenAI users, anyone shipping RAG or agent solutions. AI-103 is more immediately practical and the content directly transfers to day-to-day work. Take AI-200 first if you are a backend or platform engineer responsible for the containers, data services, and messaging infrastructure that AI workloads run on. Most career developers benefit from AI-103 first, then layering AI-200 once they have hands-on production deployment experience.

When did AI-103 and AI-200 open?

AI-103 (Azure AI App and Agent Developer Associate) opened in beta in April 2026 and has since reached general availability. AI-200 (Azure AI Cloud Developer Associate) opened in beta in May 2026 and has also since reached general availability. Passing either during its beta period earned a permanent credential identical to a post-GA pass, at a discounted price.

How much does each exam cost?

Both AI-103 and AI-200 now cost the standard $165 USD Associate-exam fee following general availability. During their respective beta periods, pricing was discounted to roughly $99 USD. Microsoft also frequently issues additional voucher offers via Cloud Skills Challenges, virtual training days, and partner programs.

Do AI-103 and AI-200 overlap?

There is modest overlap, primarily in foundational topics: responsible AI principles, basic Azure resource and identity management, and shared knowledge of which Azure AI services exist. Beyond those foundations, AI-103 dives into application-level patterns (prompt engineering, agents, RAG) while AI-200 dives into backend-platform patterns (containers, vector databases, messaging, security). Studying for one builds a useful foundation for the other, but neither replaces the other.

Which is harder, AI-103 or AI-200?

Both are Associate-level and roughly equally hard but in different ways. AI-103 rewards developers who have built with Azure AI Foundry hands-on — its difficulty comes from architectural judgment between similar components (agent vs prompt flow, Document Intelligence vs Content Understanding). AI-200 rewards backend engineers — its difficulty comes from breadth across containers, vector-database data patterns, and messaging, particularly the newer retrieval-augmented-generation patterns in Cosmos DB and pgvector that are unfamiliar even to experienced Azure developers. A pure application developer will find AI-103 easier; a pure backend engineer will find AI-200 easier.

Can I take AI-200 without taking AI-103?

Yes. Microsoft has not announced a hard prerequisite chain. AI-200 expects familiarity with Azure AI services but does not require you to have first passed AI-103 — Microsoft typically expects equivalent knowledge from experience, not necessarily a prior certification. That said, candidates who pass AI-103 first carry a strong foundation in Azure AI Foundry that makes AI-200 easier.

Will AI-103 and AI-200 stay separate after GA?

Yes — now that both exams are generally available, they remain separate exams targeting different roles (Application Developer vs Cloud Developer). Microsoft's certification roadmap positions them as complementary credentials, not as different codes for the same content. Each retains its own exam code, skills outline, credential name, and annual renewal cycle.

Is AI-200 worth getting?

For backend and platform engineers building the infrastructure behind AI solutions, yes. AI-200 is the most explicit Microsoft credential for "I can build and run the Azure backend an AI application depends on" — covering containerized hosting, AI-ready data services like Cosmos DB and PostgreSQL with pgvector, event-driven integration, and securing AI workloads. As enterprises move AI from prototype to production, demand for engineers who understand the backend platform (not just the AI services) is growing fast. For pure application developers who build AI apps but do not own the backend infrastructure, AI-103 is the more immediately valuable credential.

How should I prepare if I want both?

Take AI-103 first to build the application and AI-services foundation, then add AI-200 to layer in the containerization, data-platform, and messaging skills. Plan 6–8 weeks for AI-103 (assuming you already use Azure AI services), then 4–6 weeks for AI-200 if you already know Container Apps/AKS (10-12 weeks if you don't). Together that's roughly 3 months for both Associate credentials, well-positioned for senior Azure AI developer and backend architect roles.

Start with AI-103 — Practice 40 Questions Free

AI-103 is the most immediately practical Azure AI Engineer exam for developers. Start with 40 free questions covering Foundry agents, RAG, vision, text, and information extraction.

Start Free AI-103 Practice →

Start with AI-200 — Practice 40 Questions Free

AI-200 is the platform-focused Azure AI Cloud Developer exam. Start with 40 free questions covering containerized AI hosting, vector data services, and Azure integration.

Start Free AI-200 Practice →

Related Azure AI Resources