AWS published the detailed guide for AWS Certified Developer – Associate (DVA-C03) on October 27, 2026. The new version does not turn the exam into an AI-engineering certification. It updates the cloud developer role instead: assisted coding, security for AI services and agents, automated testing, containers, CI/CD, and troubleshooting are now explicit assessed skills.

The transition window is short. According to the AWS Training and Certification announcement dated September 1, 2026 (opens in a new tab) and the official certification page (opens in a new tab), registration for DVA-C03 opens on October 27, 2026, DVA-C02 remains available through November 30, 2026, and DVA-C03 delivery starts on December 1, 2026. If your preparation is already advanced, you now need to choose C02 or C03 based on the objectives you can demonstrate, not merely on the version number.

DVA-C03 at a glance

Item DVA-C03
Level Associate
Duration 130 minutes
Questions 65, including 50 scored and 15 unscored
Formats Multiple choice and multiple response
Minimum score 720 out of 1,000
Announced price USD 150, excluding tax and exchange-rate variations
Delivery begins December 1, 2026
Previous version DVA-C02, available through November 30, 2026

AWS therefore retains the familiar duration, question count, and passing threshold for this level. Unscored questions are not identified during the exam, and unanswered questions count as incorrect. Scoring is compensatory: you must meet the overall standard rather than pass every domain separately. These rules appear in the DVA-C03 exam guide (opens in a new tab).

One detail deserves a check when you book. The guide published today contains a sentence saying registration opens on December 1, while the AWS announcement, the certification page, and the comparison page place delivery on December 1 and registration on October 27. This appears to be an editorial inconsistency, but do not silently “correct” it in a corporate training plan: confirm the version offered in the booking portal before paying.

Similar domains, different weights

The official DVA-C02 versus DVA-C03 comparison (opens in a new tab) retains the four main domains but changes their weights:

Domain DVA-C02 DVA-C03 Practical impact
Development with AWS Services 32% 30% Still the largest domain, now including AI assistance and managed AI integration.
Security 26% 26% Stable weight, with a new task covering risks from AI services and agents.
Deployment / Testing and Deployment 24% 22% Testing becomes visible in the name; containers and AI assistance enter the skills list.
Troubleshooting and Optimization 18% 22% A four-point increase gives more importance to logs, traces, root cause, and optimization.

It would be a mistake to infer that development matters less simply because it loses two points. Domain 1 remains the largest and its scope expands. Likewise, the lower percentage for Domain 3 does not make testing less important. AWS makes testing more explicit while consolidating several previously dispersed skills.

AI assistance now spans all four domains

The most visible update is the presence of AI throughout the developer workflow. In Domain 1 (opens in a new tab), candidates must use assisted-development tools to generate, review, refactor, and optimize code, including identifying security issues. They must also integrate managed AWS AI services into applications.

In Domain 3 (opens in a new tab), these tools support test generation, regression automation, result analysis, and coverage. They can also support deployment approvals, environment provisioning, and post-deployment validation. Finally, Domain 4 (opens in a new tab) adds error analysis, troubleshooting suggestions, bottleneck detection, and optimization recommendations.

Knowing the name Amazon Q or Kiro is not enough. Strong preparation should show where assistance enters the workflow, what data it receives, how its recommendation is reviewed, and what evidence validates the result. For a study exercise, start with a small Lambda function. Generate a test, inspect the permissions and exposed data, require human review, run the test, and compare the tool’s diagnosis with metrics and logs. The target is a developer accountable for the outcome, not a user who accepts every suggestion.

Security for AI services and agents becomes a dedicated task

Domain 2 (opens in a new tab) remains 26% of scored content and introduces Task 2.3. It groups five concrete abilities:

This list shifts preparation toward scenarios. An agent that can call a refund tool, for example, needs more than a content-filtering rule. Its role must be restricted, tool access must be explicitly authorized, sessions must be isolated, high-impact actions must require approval, and the full request or personal data must not be copied into CloudWatch Logs.

AWS explicitly excludes foundation-model selection, prompt engineering, RAG design, and organization-wide AI governance from the DVA-C03 target role. Amazon SageMaker AI also appears in the official out-of-scope services list (opens in a new tab). Do not turn the plan into a data-science curriculum. The exam treats AI as an application component and development tool.

Containers, testing, and diagnostics gain visibility

DVA-C03 adds building and managing images in Amazon ECR, followed by deploying them through Amazon ECS, Amazon EKS, or AWS Fargate. Candidates should connect image, tag, lifecycle policy, task definition or manifest, update strategy, and rollback. The objective is not full cluster administration: network-infrastructure design and server administration remain outside the target role.

The guide also relocates several skills. Unit testing with AWS SAM moves from development to Testing and Deployment; Lambda performance tuning moves from Domain 1 to Troubleshooting and Optimization; API Gateway stages and custom domains move into artifact preparation. These moves do not announce that the topics are disappearing. They show the context in which AWS expects them to be assessed.

The official in-scope services list (opens in a new tab) provides another view of the updated profile: ECR, ECS, and EKS for containers; Amazon Bedrock, Bedrock AgentCore, Amazon Quick, and Kiro for AI; Amazon Data Firehose for analytics; and AWS PrivateLink for private connectivity. The list is non-exhaustive and subject to change. Use it to identify gaps, not as a list of definitions to memorize.

Should you take DVA-C02 or DVA-C03?

Choose DVA-C02 by November 30 if you have finished a C02-aligned course, consistently pass realistic mock exams, and can book with a sensible margin. Passing C02 awards the same AWS Certified Developer – Associate certification under its normal validity cycle. AWS says an existing credential does not lose validity merely because the exam version changes.

Choose DVA-C03 from December 1 if you are starting now, your work already includes development assistants, containers, or managed AI services, or you want to validate the most current blueprint. Do not switch only because “C03 is newer.” Confirm that your materials cover the objectives released on October 27 and that any mock exam is explicitly aligned to C03.

If your goal is more specifically AI-oriented, OnLevelUp’s overview of the AWS AI certification path helps distinguish DVA-C03 from AI Practitioner, Machine Learning Engineer, and Generative AI Developer. DVA-C03 remains a broad cloud-development certification enhanced with practical AI use cases.

A C03-aligned preparation plan

Start with a four-domain diagnostic, then allocate effort according to both weight and weakness. A six-week plan can use this sequence:

  1. AWS development: SDKs, API Gateway, Lambda, events, SQS/SNS/EventBridge, data-store choice, and caching. Build a small asynchronous application with controlled retries.
  2. Core security: roles, STS, federation, presigned URLs, KMS, Secrets Manager, encryption, and certificates. Justify every permission.
  3. AI security: private connectivity, data handling, filtering, tool authorization, session isolation, and observability without sensitive leakage.
  4. Testing and delivery: SAM or CDK, test events, mocks, CodeBuild, CodePipeline, CodeDeploy, canary or blue/green strategies, and rollback.
  5. Containers and observability: ECR, ECS/EKS/Fargate at developer depth, structured logs, custom metrics, traces, alarms, and health probes.
  6. Timed review: domain sets followed by complete practice exams. For each mistake, record the decisive signal, the appropriate service, and why each distractor fails.

Use AI as a controlled tool inside those exercises rather than as a detached chapter. Ask it for a review or test, validate the output, then document limitations and human approval. That approach matches the guide better than memorizing a product list.

What to remember

DVA-C03 modernizes an existing certification without changing its level or primary job role. The four domains remain recognizable, but troubleshooting gains weight, testing becomes more visible, containers enter the blueprint clearly, and AI-assisted work spans the development lifecycle. The most specific new block is security for AI services and agents.

The immediate decision is straightforward: finish C02 by November 30 if you are ready, or rebuild your plan around the C03 guide for an exam from December 1. In either case, verify the code shown in the booking portal and use dated AWS pages and the official guide as your references, especially while the guide’s calendar inconsistency remains unresolved.

Official sources