The direct answer: If you are already deep into DP-100 preparation and can sit the exam before June 1, 2026, finish it. If you are starting fresh today, go directly to AI-300.
DP-100 retires on June 1, 2026. That is approximately 6 weeks from now. AI-300 went live in beta in March 2026 and is expected to go generally available in May 2026. The decision you make in the next few weeks determines which certification you earn — and they are meaningfully different credentials for meaningfully different roles.
What Is the Difference Between DP-100 and AI-300?
| Factor | DP-100 (Retiring June 1, 2026) | AI-300 (New — Live May 2026) |
| Official name | Designing and Implementing a Data Science Solution on Azure | Operationalizing Machine Learning and Generative AI Solutions |
| Certification earned | Microsoft Certified: Azure Data Scientist Associate | Microsoft Certified: MLOps Engineer Associate |
| Exam cost | $165 USD | $165 USD |
| Passing score | 700/1000 | 700/1000 |
| Duration | ~120 minutes, 40–60 questions | ~120 minutes, 40–60 questions |
| Primary focus | Building and training ML models using Azure Machine Learning | Operationalizing, deploying, monitoring, and governing ML and GenAI systems |
| Generative AI coverage | Limited — LLM optimization at 10–15% | Deep — GenAIOps is a dedicated 20–25% domain |
| Infrastructure as Code | Not a primary focus | Core requirement — Bicep, Azure CLI, GitHub Actions |
| MLflow | Covered | Covered and extended |
| Microsoft Foundry | Not covered | Core platform — GenAIOps runs entirely through it |
| CI/CD pipelines | Mentioned | Dedicated domain — GitHub Actions workflows |
| Retirement date | June 1, 2026 | No retirement planned |
| Validity | 1 year renewable | 1 year renewable |
What Does DP-100 Actually Test?
DP-100 is built for data scientists who design, build, and train machine learning models on Azure. The exam validates your ability to take a business problem, acquire and explore data, train and evaluate models, and deploy them into production using Azure Machine Learning.
DP-100 Exam Domains
| Domain | Weight | What You Do |
| Design and prepare a machine learning solution | 20–25% | Set up Azure ML workspaces, compute resources, datastores, environments |
| Explore data and run experiments | 35–40% | Data preparation, feature engineering, experiment tracking with MLflow |
| Train and deploy models | 20–25% | Model training, evaluation, managed online and batch endpoints |
| Optimize language models for AI applications | 10–15% | Prompt engineering, RAG basics, LLM fine-tuning on Azure AI |
DP-100 is a data scientist exam. It starts with data, works through training, and ends at deployment. The deployment piece is real but not deep. The operations piece — what happens after deployment — is largely out of scope.
Core tools tested: Azure Machine Learning Studio, Azure Machine Learning SDK (Python), MLflow, Azure Databricks (basic), Automated ML, Azure AI Services.
What Does AI-300 Actually Test?
AI-300 is built for MLOps engineers — professionals whose job is to take models that data scientists have built and make them work reliably, automatically, and at scale in production. It also adds an entirely new dimension: operationalizing generative AI systems through Microsoft Foundry.
AI-300 Exam Domains
| Domain | Weight | What You Do |
| Design and implement MLOps infrastructure | 15–20% | Set up secure Azure ML workspaces, managed identities, RBAC, private networking, IaC with Bicep and Azure CLI |
| Implement ML model lifecycle and operations | 25–30% | Automate training pipelines with GitHub Actions, manage model registration and versioning, deploy to managed endpoints, monitor for data drift, implement rollback strategies |
| Design and implement GenAIOps infrastructure | 20–25% | Deploy generative AI solutions on Microsoft Foundry, configure foundation model deployments, manage prompt versioning, set up scalable inference |
| Implement generative AI quality assurance and observability | 10–15% | Run evaluation workflows using groundedness, relevance, and coherence metrics; monitor latency, throughput, token usage; implement safety evaluations |
| Optimize generative AI systems and model performance | 10–15% | Optimize RAG pipelines, tune retrieval strategies, implement fine-tuning with LoRA, manage synthetic data generation, promote fine-tuned models to production |
AI-300 is an operations engineering exam. It starts where DP-100 ends — at deployment — and goes much deeper into everything that makes production AI systems reliable, observable, and governed.
Core tools tested: Azure Machine Learning, Microsoft Foundry, GitHub Actions, Bicep, Azure CLI, MLflow, Azure AI Search (for RAG), managed online endpoints, prompt flow, Azure Monitor.
The Core Philosophical Difference
This is what matters most and what most articles miss.
DP-100 asks: Can you build a machine learning model?
AI-300 asks: Can you make machine learning and generative AI work reliably in production?
These are fundamentally different jobs. A data scientist who holds DP-100 validates that they can go from data to trained model. An MLOps engineer who holds AI-300 validates that they can take that model and build the automated systems, monitoring pipelines, governance controls, and operational frameworks that make it trustworthy and scalable at enterprise level.
Microsoft is not making a subtle naming change. They are formally recognizing that the discipline of operationalizing AI — MLOps and GenAIOps — is a distinct engineering role that deserves its own certification, not a footnote in a data science exam.
What Carries Over from DP-100 to AI-300?
If you hold DP-100 or have been studying for it, this is what transfers:
| DP-100 Knowledge | Relevance to AI-300 |
| Azure Machine Learning workspace setup | Direct — AI-300 builds on this foundation |
| MLflow tracking and model registration | Direct — AI-300 extends MLflow into full lifecycle governance |
| Managed online endpoints | Direct — AI-300 goes deeper into endpoint management and monitoring |
| Python for ML | Direct — same language, more DevOps context |
| Data exploration and feature engineering | Indirect — useful context but not directly tested |
| Automated ML | Indirect — useful background |
| Azure AI Services basics | Partial — AI-300 uses Microsoft Foundry which supersedes the old AI Services structure |
What DP-100 knowledge does NOT prepare you for in AI-300:
- Infrastructure as Code using Bicep
- GitHub Actions CI/CD pipeline construction
- Microsoft Foundry for GenAIOps
- Prompt versioning and foundation model deployment
- RAG pipeline optimization and retrieval tuning
- GenAI quality evaluation frameworks (groundedness, coherence, relevance)
- Data drift monitoring and automated rollback strategies
If you have studied DP-100 seriously, you are not starting from zero for AI-300. You are starting from the halfway point with a solid foundation that needs operational and GenAI depth added on top.
Who Should Take DP-100 Right Now?
Take DP-100 before June 1, 2026 if:
You have already completed most of your preparation. If you have studied DP-100 content for weeks or months, have practice exam experience, and feel close to ready — do not switch. Complete your preparation and sit the exam before the deadline. A passed DP-100 on your transcript is a valid credential until it expires regardless of retirement.
Your role is data science, not DevOps or operations. If your daily work involves building models, running experiments, and exploring data rather than building CI/CD pipelines and managing production infrastructure, DP-100 still accurately maps to your actual job.
You need a credential quickly. AI-300 is in general availability from May 2026, but the prep ecosystem — official instructor-led training, practice exams, community study guides — is still maturing. DP-100 has years of study materials, community answers, and established preparation paths. If you need to certify within the next few weeks, DP-100 is significantly lower-friction to prepare for.
Who Should Go Directly to AI-300?
Go to AI-300 if:
You are starting from scratch today. Spending 6 weeks preparing for a certification that retires in 6 weeks to avoid starting on AI-300 is almost never the right move for a new candidate.
Your role involves deploying, monitoring, or managing AI systems in production. If you work in MLOps, DevOps for AI, platform engineering for AI teams, or you manage generative AI deployments, AI-300 directly validates what you actually do.
You work with Microsoft Foundry or build generative AI applications. DP-100 has no Microsoft Foundry content. If this platform is part of your stack, AI-300 is the only Microsoft certification that validates your work.
You want a certification that will still be relevant in 3 years. The direction of enterprise AI is unambiguously toward operationalization, automation, and governance. AI-300 is built for that future. DP-100 was built for the past.
Salary and Career Impact: DP-100 vs AI-300
| Role | Average US Salary |
| Azure Data Scientist (DP-100 level) | $95,000 to $130,000 |
| MLOps Engineer (AI-300 level) | $110,000 to $155,000 |
| Senior MLOps Engineer / AI Platform Engineer | $140,000 to $185,000 |
| GenAI Operations Specialist | $120,000 to $160,000 |
MLOps engineering is currently one of the fastest-growing specializations in enterprise AI. Organizations are hiring aggressively for professionals who can bridge the gap between data science and production engineering. AI-300 is the first Microsoft certification purpose-built for exactly this profile.
DP-100 vs AI-300: The Decision Framework
Use this to make your decision in under 60 seconds:
| Your Situation | Take This Exam |
| Already deep into DP-100 prep — can test before June 1 | DP-100 — finish what you started |
| Starting from zero today | AI-300 — do not study a retiring exam |
| Your job is data science (building and training models) | DP-100 if time allows, then plan AI-300 as your next cert |
| Your job is MLOps or DevOps for AI | AI-300 directly |
| You work with Microsoft Foundry or generative AI systems | AI-300 — DP-100 does not cover this |
| You need a credential in the next 4 weeks | DP-100 has more mature study materials right now |
| You want the most future-relevant credential | AI-300 |
| You hold DP-100 and it is expiring soon | Renew DP-100 before June 1, then start AI-300 |
How to Prepare for AI-300
Step 1: Download the official AI-300 study guide from Microsoft Learn. The study guide lists every skill measured with exact domain weightings. This is your definitive preparation blueprint.
Step 2: Build hands-on experience with GitHub Actions and Bicep. This is the biggest skills gap for candidates coming from pure data science backgrounds. Build at least one end-to-end workflow — a GitHub Actions pipeline that deploys an Azure ML workspace using Bicep, runs a training job, registers a model, and deploys to a managed online endpoint.
Step 3: Get hands-on with Microsoft Foundry. Set up a Foundry hub and project. Deploy a foundation model. Build a prompt flow. Enable continuous evaluation. This is the GenAIOps loop that AI-300 tests, and reading about it is not sufficient.
Step 4: Understand the five AI-300 domains by weight. ML model lifecycle and operations (25–30%) is the heaviest domain. GenAIOps infrastructure (20–25%) is second. These two domains together account for roughly 50 percent of the exam. Give them disproportionate preparation time.
Step 5: Practice with current materials. Our Microsoft exam preparation materials cover current active Microsoft exams. Use practice questions aligned to the AI-300 blueprint to identify gaps before exam day.
How to Prepare for DP-100 Before June 1, 2026
You have approximately 6 weeks. This is achievable but requires a focused plan.
Week 1–2: Domain 2 (Explore data and run experiments — 35–40%). This is the heaviest domain. Prioritize it first. Use Azure Machine Learning Studio hands-on labs.
Week 3: Domain 1 (Design and prepare — 20–25%) and Domain 3 (Train and deploy — 20–25%). These share tools and concepts. Study them together.
Week 4: Domain 4 (Optimize language models — 10–15%). Focus on RAG basics, prompt engineering concepts, and LLM fine-tuning in Azure AI.
Week 5: Full practice exams. Identify gaps and revisit weak areas.
Week 6: Final review. Book your exam. Do not delay — June 1 is the hard cutoff.
Our DP-100 exam preparation materials include up-to-date practice questions aligned to the current DP-100 exam blueprint.
What About DP-750? How Does It Fit?
DP-750 (Azure Databricks Data Engineer Associate) is a separate new certification that was also launched in beta in March 2026. It is not a replacement for DP-100 — it targets a different role.
| Certification | Role | Focus |
| DP-100 (retiring) | Data Scientist | Build and train ML models using Azure Machine Learning |
| AI-300 (new) | MLOps Engineer | Operationalize ML and GenAI in production |
| DP-750 (new) | Databricks Data Engineer | Build scalable data pipelines using Azure Databricks |
If your work involves Databricks as your primary data engineering platform, DP-750 is relevant independently of the DP-100 to AI-300 transition. It is a separate path, not a competing one.
For the full picture of all Microsoft certification changes happening right now, our Microsoft certifications retiring in 2026 guide covers every retirement date, replacement, and action plan across all tracks.
Frequently Asked Questions: DP-100 vs AI-300
What is the difference between DP-100 and AI-300?
DP-100 validates data science skills — building, training, and deploying ML models. AI-300 validates MLOps engineering skills — operationalizing, automating, monitoring, and governing both ML and generative AI systems in production. DP-100 retires June 1, 2026. AI-300 is its replacement and is live now.
When does DP-100 retire?
DP-100 retires on June 1, 2026. After this date, the exam can no longer be taken. Certifications already earned remain valid on your transcript until they expire.
Is AI-300 harder than DP-100?
AI-300 requires a broader skill set — Python data science skills plus GitHub Actions, Bicep, IaC, Microsoft Foundry, and GenAIOps. For candidates with a pure data science background and no DevOps or operations experience, AI-300 will require more preparation. For candidates already working in MLOps or platform engineering, it directly validates daily work.
Does DP-100 knowledge transfer to AI-300?
Yes, significantly. Azure Machine Learning workspace skills, MLflow, managed endpoints, and Python for ML all carry over. The new areas to add are infrastructure as code, CI/CD pipeline automation, Microsoft Foundry, and GenAI operations.
Can I still take DP-100 in May 2026?
Yes. DP-100 is available until June 1, 2026. If you can prepare and test before that date, you can still earn the Azure Data Scientist Associate certification.
What certification does AI-300 earn you?
AI-300 earns you the Microsoft Certified: MLOps Engineer Associate (also officially titled Operationalizing Machine Learning and Generative AI Solutions). The credential is valid for one year and renewable.
Should I take both DP-100 and AI-300?
If you can take DP-100 before June 1 and your role spans both data science and operations, earning both before DP-100 retires gives you two distinct credentials on your transcript. After June 1, DP-100 is gone and only AI-300 is available.
What is Microsoft Foundry and why does it matter for AI-300?
Microsoft Foundry is Microsoft’s unified platform for building, deploying, and governing generative AI applications and agents. AI-300 tests GenAIOps through Foundry extensively — foundation model deployment, prompt management, evaluation, and observability all run through it. Candidates who have not used Foundry hands-on will find this the steepest part of AI-300 preparation.
What is MLOps and who is it for?
MLOps (Machine Learning Operations) is the engineering discipline of applying DevOps practices to machine learning systems — automating training pipelines, managing model versions, monitoring production models, and maintaining CI/CD workflows for AI. AI-300 is the first Microsoft certification purpose-built to validate these skills.
Where can I practice for AI-300?
The Microsoft Learn study guide for AI-300 is the starting point. Official instructor-led training launched in late March 2026. Our Microsoft exam preparation section includes current practice materials for active Microsoft certifications.
