New Microsoft Beta Exams 2026: Full Timeline and Current GA Status

✓ Update: No longer in beta — AI-300 is now generally available as Microsoft Certified: Machine Learning Operations Engineer Associate.

MLOps is the discipline of operationalizing machine learning — taking models from experiment to production reliably and at scale. AI-300 validates that you can design, implement, and manage end-to-end ML pipelines on Azure.

Expect the exam to cover Azure Machine Learning service deeply — model training pipelines, automated ML, responsible AI principles, model monitoring and drift detection, and CI/CD integration for ML workloads. This is not an introductory exam. Candidates should already be comfortable with Python, Azure ML SDK, and basic DevOps concepts.

✓ Update: No longer in beta — DP-750 is now generally available.

Databricks has become the dominant platform for large-scale data engineering, and Microsoft's partnership with Databricks has brought it deep into the Azure ecosystem. DP-750 fills a gap that practitioners have been pointing at for years — a dedicated Databricks certification from Microsoft itself.

The exam will likely cover Delta Lake, Unity Catalog, Databricks Workflows, Spark on Azure, integration with Azure Data Factory and Synapse, and data lakehouse architecture. If you're building pipelines on Databricks as your day job, this certification directly validates the work you're already doing.

✓ Update: No longer in beta — DP-800 is now generally available.

SQL and AI are colliding faster than most expected. Azure SQL is now shipping built-in vector search, AI model calling from T-SQL, and integration with Azure OpenAI. DP-800 validates that you can build AI-powered applications using SQL Server and Azure SQL Database as the foundation.

Expect topics like vector embeddings in SQL, calling OpenAI endpoints from database procedures, Azure SQL intelligent query processing, and building retrieval-augmented generation (RAG) pipelines with SQL as the data layer. This is a niche but fast-growing area, and early certifiers will stand out.

✓ Update: No longer in beta — AI-103 is now generally available as Microsoft Certified: Azure AI Apps and Agents Developer Associate.

This is arguably the most strategically important exam in the entire 2026 wave. AI agents — autonomous systems that can plan, use tools, and take actions — are where Microsoft is investing heavily through Copilot Studio, Azure AI Foundry, and the Semantic Kernel framework. AI-103 validates that you can build them.

Expect heavy coverage of Azure AI Foundry, prompt engineering, RAG (retrieval-augmented generation) implementation, function calling, agent orchestration, responsible AI safeguards, and deploying AI apps at production scale. Developers who master this content will be positioned for the most in-demand roles of 2026-2027.

The five exam domains are weighted heavily toward generative AI and agentic solutions (35–40%), with the remainder split across planning and responsible AI, computer vision, text analysis, and information extraction. Most candidates pass with 4–8 weeks of focused preparation when they already use Azure AI Foundry hands-on. Beta scoring releases 1–2 weeks after the beta period closes, but the credential is awarded as of your exam date — so taking the beta does not delay your earning the certification.

✓ Update: No longer in beta — AI-901 is now generally available (English version last updated April 15, 2026).

This is the entry-level credential for the entire AI-xxx exam family — the equivalent of AZ-900 for Azure or AB-900 for Microsoft 365 Copilot & Administration, but for AI. No coding experience required. AI-901 is designed to prove foundational AI literacy across concepts, responsible AI principles, and Microsoft's core AI services.

For professionals in non-technical roles who need to understand and communicate about AI initiatives — project managers, business analysts, compliance officers, executives — AI-901 provides a credible, vendor-validated baseline. It's also a natural starting point before pursuing AI-103 or AI-200.

SC-730 is the most unconventional exam in this wave. It's not targeting security engineers or administrators — it's targeting business leaders who need to understand cybersecurity well enough to make informed decisions, communicate with security teams, manage risk, and comply with regulations.

Think of it as the security equivalent of a business-focused certification. Topics will likely include threat landscape awareness, security governance frameworks, regulatory compliance (GDPR, NIS2, etc.), security investment decisions, incident response communication, and how to evaluate security posture without deep technical knowledge.

With cybersecurity breaches increasingly making board-level news, this credential fills a genuine gap between IT security teams and the executives who fund and oversee them.

✓ Update: No longer in beta — AI-200 is now generally available as Microsoft Certified: Azure AI Cloud Developer Associate.

Where AI-103 focuses on apps and agents, AI-200 zooms out to the backend engineering behind AI workloads. This exam validates that you can build and operate the Azure infrastructure an AI application depends on — containerized hosting, AI-ready data services, and the messaging and security layers around them.

Expect topics like Azure Container Apps and AKS hosting with KEDA autoscaling, vector similarity search in Cosmos DB and Azure Database for PostgreSQL with pgvector, Azure Managed Redis caching, Service Bus and Event Grid integration, Azure Functions, and securing/monitoring solutions with Key Vault, App Configuration, OpenTelemetry, and KQL.

Status: Still in beta. Microsoft has published the full skills-measured objectives, so our practice bank is already complete and accurate — you don't need to wait for GA to start preparing.

This is the exam the security community has been waiting for. AI systems introduce entirely new attack surfaces — prompt injection, model poisoning, data leakage through embeddings, adversarial inputs, and shadow AI deployed without governance. SC-500 validates that you can secure both cloud infrastructure and the AI workloads running on it.

Expect significant overlap with existing SC-series content (identity, endpoint security, SIEM/SOAR) but with a clear AI security lens added on top — covering AI model threat modelling, securing Azure OpenAI deployments, content filtering, data classification for AI pipelines, and AI governance frameworks like Microsoft's Responsible AI standard.

If you currently hold SC-200 or SC-300, SC-500 is a natural extension. The demand for professionals who understand both cloud security and AI security is growing faster than supply right now.