AWS opened registration for AWS Certified Solutions Architect – Professional SAP-C03 on October 27, 2026 and published the detailed guide that was still missing from its September announcement. Delivery of the new version begins on November 17, 2026. SAP-C02 remains available through its final testing date, November 16, 2026.

This is more than a service-list refresh. SAP-C03 replaces four domains organized around solution lifecycle stages with five blocks that better reflect current architecture work: cloud-native architecture, security and governance, cost, resilience and migration, and operational excellence. Generative AI, agents, post-quantum cryptography, modern data platforms, containers and automation are now explicit.

SAP-C03 at a glance

Item SAP-C03
Level Professional
Registration opens October 27, 2026
First delivery date November 17, 2026
Duration 180 minutes
Questions 75: 65 scored and 10 unscored
Formats Multiple choice and multiple response
Passing score 720 out of 1,000
Announced price USD 300, excluding taxes and regional variation
Languages English, Japanese, Korean, Brazilian Portuguese, Simplified Chinese and Latin American Spanish
Delivery Pearson VUE test center or online proctoring

These details appear in the SAP-C03 exam guide (opens in a new tab) and on the official certification page (opens in a new tab). The ten unscored questions are not identified. An unanswered question is incorrect, there is no additional penalty for guessing, and scoring remains compensatory: AWS assesses the overall score rather than requiring a pass in every domain.

A new five-domain structure

The official SAP-C02 and SAP-C03 comparison (opens in a new tab) shows a real reorganization.

SAP-C03 domain Weight Preparation impact
Architecture Design and Implementation Built for the Cloud 25% Cloud native, serverless, containers, networking, data and AI/ML integration are brought together.
Security, Compliance, and Governance 25% Identity, encryption, multi-account governance, compliance and responsible AI security.
Cost-Optimized Architecture Design 15% Pricing models, transfers, allocation, storage, databases and AI/ML cost.
Resilience, Migration, and Business Continuity 15% Migration, disaster recovery, resilience testing and continuity.
Operational Excellence and Automation 20% IaC, CI/CD, DevSecOps, observability and controlled remediation.

SAP-C02's four domains were weighted at 26%, 29%, 25% and 20%. Do not mechanically convert each old chapter to one new percentage. AWS consolidated scattered skills, moved objectives and created a full domain for operations and automation.

Rebuild your study matrix from the C03 guide. For every topic you already covered, mark it as retained, moved, expanded or new. This prevents two common mistakes: dropping a concept that still appears elsewhere, and overspending study time on a service that is no longer explicitly in scope.

AI architecture becomes a Solutions Architect skill

Domain 1 (opens in a new tab) adds a complete task for AI and machine learning integration. Candidates must design generative AI integrations, including multi-account patterns, agent architectures connected to tools, ML inference and RAG systems with ingestion, retrieval and vector storage.

The guide names Amazon Bedrock, Bedrock AgentCore, Bedrock Knowledge Bases and SageMaker AI endpoints. Model training and fine-tuning workflows are explicitly outside the target role. The architect instead chooses boundaries, data flows, identities, resilience, cost and controls around services that provide those capabilities.

A useful exercise is to design an internal assistant that searches a document base and invokes a business tool. Separate the user's identity, the agent's identity and the tool's identity; locate sensitive data; require human approval for a high-impact action; then explain what happens when the model, vector store or tool is unavailable. For practical AgentCore and multi-agent work beyond exam questions, OnLevelUp's review of the new AWS Jam experiences for Kiro and AgentCore is a useful but clearly separate resource.

Security: post-quantum, workload identity and responsible AI

Domain 2 (opens in a new tab) accounts for 25%. Familiar foundations remain: federation, least privilege, cross-account access, data perimeters, KMS, CloudHSM, Organizations, Control Tower, SCPs, RCPs, centralized logging and compliance evidence.

Three additions deserve focused preparation. First, the guide includes workload identity for hybrid and multi-cloud environments with IAM Roles Anywhere. Second, it adds post-quantum cryptography strategy, including ML-DSA and ML-KEM with AWS KMS. Third, its “security and responsible AI controls” task covers Bedrock Guardrails, access controls for agentic applications and human approval steps.

Do not memorize these labels in isolation. Build an architecture decision: which data needs protection, against which threat, at what point in its path, with what evidence and migration plan? For an agent, add tool-level authorization, session isolation and action logging without unnecessarily copying sensitive content.

Cost: AI and transfer decisions join the blueprint

Domain 3 (opens in a new tab) is the smallest at 15%, but it is not simply a Savings Plans chapter. It covers purchasing options, rightsizing, transfer costs, multi-Region design, tagging, chargeback and showback, storage, databases and observability.

SAP-C03 adds AI/ML optimization: selecting a service tier, batching, caching, cross-Region inference routing and token usage monitoring. Practice scenarios where the lowest unit price becomes expensive because of transfer, replication, latency or low utilization. The correct answer must satisfy all constraints, not merely reduce one line item.

Resilience and migration: prove that recovery works

Domain 4 (opens in a new tab) retains migration, RTO, RPO, replication and disaster recovery, while making testing more concrete. It names Resilience Hub, Fault Injection Service, Application Recovery Controller, multi-cluster ECS/EKS strategies, backup locks and chaos engineering.

For every design, explain how failover starts, how data is validated after recovery and how failback is controlled. A “multi-Region” diagram without a test procedure, restoration evidence or cost estimate is incomplete.

Migration still tests decisions: dependency discovery, waves, rehost, replatform, refactor and modernization. AWS Transform appears in examples. Study why a model fits and how you control transitional risk instead of memorizing only tool names.

Operational excellence is a full domain

Domain 5 (opens in a new tab) represents 20% of scored content. It covers CloudFormation, CDK, policy as code, secrets in IaC, StackSets, multi-account CI/CD, scanning, signing, rollback-capable deployments, observability and remediation.

Important additions include evaluating AI applications, monitoring model or agent drift and performance, synthetic transactions, service-level objectives, and remediation guarded by circuit breakers, rate limits, bulkheads and exponential backoff. Automation is not correct merely because it acts quickly: it must limit blast radius, produce evidence and support rollback.

Added and removed services: read the list carefully

AWS adds Bedrock, Security Lake, Verified Permissions, VPC Lattice, CDK, Clean Rooms, Cloud WAN, OpenTelemetry, Fault Injection Service, Resilience Hub, Transform and Verified Access, among others, to the explicit in-scope list. IoT, End User Computing and Media Services categories disappear, along with several individual services.

This is not an invitation to memorize hundreds of product cards. Use the list as a final gap check after studying architecture decisions. AWS also notes that parent services now implicitly include some features that used to be listed separately: a textual deletion does not always mean the underlying concept can never appear.

Choose SAP-C02 or SAP-C03

Take SAP-C02 by November 16 if your preparation is complete, your results are stable and you can book a date without rushing. The credential keeps its normal validity cycle; AWS does not invalidate an earned title when the exam version changes.

Choose SAP-C03 from November 17 if you are starting now, want to validate current architecture practice, or already work with modern data platforms, containers, generative AI, DevSecOps and tested resilience. Do not mix C02 and C03 practice tests without clear labels.

If AI is your main direction, compare the generalist architect role with OnLevelUp's AWS AI certification path. SAP-C03 assesses AI as part of a complete architecture; it does not replace a specialized AI credential or data science curriculum. The AWS training on AI compute and EKS efficiency can also supply hands-on context for two newly visible areas without becoming an official SAP-C03 course.

A six-step transition plan

  1. Download the C03 guide and official comparison, and record the access date.
  2. Reclassify your C02 notes under the five C03 domains.
  3. Create a separate backlog for AI and agents, post-quantum security, data platforms, containers, DevSecOps and tested resilience.
  4. Build three cross-domain scenarios: a multi-account platform, an agentic application and multi-Region recovery.
  5. For each scenario, justify security, cost, resilience and operations, not only the primary service.
  6. Finish with timed sets explicitly aligned to SAP-C03 and record the decisive clue behind each answer.

The detailed guide published on October 27 finally makes this transition possible without guesswork. The priority is not to bolt an “AI chapter” onto an old course. It is to revisit how architecture, security, cost and operational decisions interact.

Official sources