GCP PCA prep, professional cloud architect roadmap with ARIA
The Google Professional Cloud Architect (PCA) is a 120-minute, 50-question exam with a passing score around 72%, built for engineers who already design production GCP systems. Google does not formally require the Associate Cloud Engineer exam, but I strongly recommend you sit ACE first, and the PCA assumes roughly three years of cloud experience including one year on GCP. Four canonical case studies must be pre-read. 
ARIA runs the adaptive evaluation, builds your architect roadmap, and stands behind it with a pass guarantee tied to five measurable conditions.
Start your PCA roadmap. About five minutes to the first signal.
TL;DR
- PCA is the advanced GCP architect exam, current as of 2026: 50 questions, 120 minutes, ~72% to pass, no labs.
- Four domains, weighted unevenly. Cloud Solution Design alone accounts for 40% of the exam, the largest single block on any major architect cert.
- Four case studies (Mountkirk Games, EHR Healthcare, Helicopter Racing League, TerramEarth) drive roughly a quarter of the questions and must be read in advance.
- Most engineers with ACE-level background finish in 8 to 12 weeks at 45 to 60 minutes per day.
- Pass guarantee eligibility requires every milestone done, two mock exams passed, one gauntlet at 80%+, and a live readiness score of 80 or higher when you sit the exam.
What the PCA exam is
PCA is design-heavy, scenario-driven, and shorter than its AWS counterpart. You get 50 questions in 120 minutes, which is a comfortable 2.4 minutes per item, but the questions themselves are long. A typical item is a two-paragraph scenario with four answers that all technically work, and the correct one hinges on a constraint buried in the prompt or, more often, in the case-study brief that the question references.
Domain weights, current as of 2026
| Domain | Weight |
|---|---|
| Cloud Solution Design | 40% |
| Manage and Provision Infrastructure | 25% |
| Security and Compliance | 20% |
| Analyze and Optimize | 15% |
Solution Design at 40% is the largest single domain block on any current architect cert. That weight is the reason the prep approach for PCA looks different from PCA-shaped AWS or Azure preps. Almost half of your study time should land in design tradeoff reasoning, not service recall.
The four canonical case studies
Google publishes four case studies in advance, and the exam draws scenarios from all of them. Read each one twice before sitting the exam. Then read it again the night before.
- Mountkirk Games. A mobile gaming company moving multi-region. Latency-sensitive multiplayer, analytics on player behavior, autoscaling under unpredictable launch spikes.
- EHR Healthcare. A healthcare records vendor balancing HIPAA, hybrid connectivity, and regulated multi-tenant data isolation. Compliance constraints carry through almost every answer.
- Helicopter Racing League. A streaming and analytics company with global broadcast, real-time ML inference on race telemetry, and edge delivery.
- TerramEarth. A heavy-equipment manufacturer ingesting telemetry from 20 million vehicles, ETL pipelines, and connected-vehicle architectures with intermittent connectivity.
The same GCP service can be the right answer for one case and the wrong answer for another, because the case studies set different constraints. Cloud Spanner is correct for EHR's strongly consistent global writes, and overkill for TerramEarth's append-heavy telemetry.
Positioning vs ACE, SAP-C02, and AZ-305
ACE verifies you can deploy and operate GCP services. PCA verifies you can design the architecture in the first place. The two exams overlap in vocabulary and diverge entirely in question style: ACE asks "which command", PCA asks "which design".
Against SAP-C02, PCA is shorter (50 vs 75 questions, 120 vs 180 minutes) but more design-focused per question. AWS spreads its design weight across five domains; Google concentrates it into one 40% block. Against AZ-305, PCA shares the design-only philosophy, but PCA's case studies are uniquely public and uniquely heavy.
How ARIA preps you for it
I treat PCA differently from associate-tier certs and even from other professional architect certs, because the case-study load reshapes the prep itself.
The CAT evaluation calibrates against design questions. I open every cert with a CAT adaptive test. For PCA, the candidate population is biased toward engineers who already hold ACE or have GCP production experience, so the evaluation converges higher in the 1 to 5 difficulty band. Most of your evaluation time lands in Analytical and Complex design scenarios, not Foundational service recall. That gives me a sharper read on which design domains are strong and which still need scaffolding.
The roadmap is longer than the question count suggests. PCA has only 50 questions, but the case-study reading load adds real weeks. Working engineers studying 30 to 45 minutes a day typically end up with a 10 to 12 week roadmap, and the first phase often includes dedicated case-study comprehension milestones before any GCP service drilling starts. See the roadmap overview for how phases, milestones, and tasks fit together.
The gauntlet emphasizes case-study clusters. The gauntlet is the long-form exam-conditions session I unlock at 80% readiness, and on PCA it is the closest analog to exam day. I structure gauntlet sessions around case-study clusters so you build the muscle of holding an entire case context in working memory across several questions. Short isolated practice does not surface that fatigue pattern.
The error backlog tags case-study traps separately. PCA candidates miss questions for three reasons: they do not know a service detail, they reasoned past a stated constraint, or they applied the right answer from the wrong case study. The third bucket is its own tag in the backlog, with revisit intervals weighted toward case-study mock-exam format rather than isolated drill. Same wrong answer, different remediation path.
Readiness gates the demo test and the gauntlet. The demo test is locked until 60% readiness. The gauntlet is locked at 80%. Neither is a marketing number. Both reflect the point at which the next session type produces a useful signal instead of noise. See readiness and decay for how the score moves day to day.
Common pitfalls on PCA
These are the topics that quietly cost candidates the most points. For each one, here is what I do during prep.
Case-study mapping. The most common failure mode on PCA is picking a technically correct answer that is wrong for the specified case. Cloud Spanner solves global strong consistency, which EHR Healthcare wants and TerramEarth does not. Cloud Run is right for stateless event-driven workloads in Mountkirk, and wrong for long-running ML inference at Helicopter Racing League. I drill case-study mapping as its own milestone, with practice items tagged by case so you build per-case intuition, not generic GCP intuition.
GKE vs Cloud Run vs GKE Autopilot. Three compute options for containers, with overlapping use cases and very different operational shapes. The exam will give you a workload profile (request-driven, batch, stateful, multi-tenant, regulated) and ask you to pick. I run a dedicated decision-framework session that walks you through the discriminating questions: do you need pod-level networking control, do you need to manage the control plane, what is your cold-start tolerance. By exam day the choice should be reflexive, not deliberative.
BigQuery vs Bigtable vs Spanner vs Firestore. Four data services, four access patterns. BigQuery for analytical scans over large columnar data. Bigtable for low-latency wide-column at scale. Spanner for globally consistent transactional. Firestore for document, mobile-friendly. The exam tests these by giving you a read pattern and a write pattern and asking you to pick. I cover the access-pattern matrix in a dedicated milestone with calculator-style scenarios so you stop pattern-matching on service names.
Anthos vs hybrid alternatives. Anthos is the headline answer for hybrid and multicloud, but the exam will sometimes give you a constraint where Anthos is overkill and a simpler hybrid pattern (VPN, Cloud Interconnect, Migrate to Containers) is the correct answer. I generate scenarios that contrast Anthos against lighter options so you do not default-pick the marketing answer.
IAM hierarchy at scale. Organization Policy, folder-level IAM, project-level IAM, resource-level IAM, VPC Service Controls, and Workload Identity Federation each solve a specific class of problem. The exam will give you a multi-team, multi-environment scenario and ask you to pick the least operationally expensive option that meets the security requirement. I drill the hierarchy as its own milestone, with explicit contrast between organization policy constraints and IAM grants, because conflating them is a routine cause of wrong answers.
Networking choices: Shared VPC vs VPC Peering vs Network Connectivity Center. Three composition models with different tradeoffs in centralized governance, transitivity, and operational overhead. Shared VPC for centralized network admin in a host-and-service-projects model. VPC Peering for non-transitive direct connections. Network Connectivity Center for hub-and-spoke at scale. I run a hybrid networking milestone with the full matrix, and revisit it inside Solution Design when case studies require it.
Data migration patterns. Datastream for change data capture from operational databases. Database Migration Service for managed homogeneous and heterogeneous migrations. Storage Transfer Service for object storage at scale. Transfer Appliance when bandwidth is the constraint. Each maps to a specific question shape, and the exam tests the discrimination explicitly. I cover all four in a migration milestone, then practice cross-discrimination in mock-exam format.
Cost design. Committed Use Discounts, Sustained Use Discounts, Spot VMs, custom machine types, and resource-level commitments compose into a real cost-architecture problem. The exam routinely asks you to optimize cost without sacrificing availability or performance, and the right answer is usually a combination, not a single lever. I run cost milestones with calculator-style scenarios where you choose a commitment shape and a fallback policy together.
Common questions
Do I need to pass ACE before attempting PCA?
Google does not require it. I strongly recommend it. The Associate Cloud Engineer exam covers the day-to-day GCP service vocabulary that the architect exam assumes you already own. If you skip ACE without equivalent production experience, the PCA scenarios will read like a foreign language for the first three weeks.
Do I really need to read all four case studies before sitting PCA?
Yes. Mountkirk Games, EHR Healthcare, Helicopter Racing League, and TerramEarth are not optional context. Roughly a quarter of the exam draws directly on these case studies, and the answers depend on constraints that are only stated in the case-study brief. Walking in cold is the single most common reason candidates fail by a few points.
How does ARIA simulate Google's case-study question style?
I generate scenario clusters that mirror the case-study format: a paragraph or two of business and technical context, followed by a question whose correct answer hinges on a constraint hidden in the brief. Practice items are tagged by case study so you build a per-case mental model, not a generic GCP one. The gauntlet then strings several case-study items together so you practice holding context across questions, the way the real exam does.
How long should I expect to study for PCA?
At 30 minutes a day, plan on 14 to 18 weeks. At 45 minutes a day, 10 to 12 weeks. At 60 minutes a day, 8 to 10 weeks. These bands assume ACE-level vocabulary and at least one year of GCP exposure. If your CAT evaluation puts you in the Novice or Developing band on more than two domains, add four to six weeks for case-study reading.
How does PCA compare to AWS SAP-C02 and Azure AZ-305?
All three sit at the professional architect tier and reward tradeoff reasoning over service recall. PCA is shorter (50 questions in 120 minutes) and uniquely leans on four pre-published case studies. SAP-C02 is longer and broader (75 questions in 180 minutes) and tests cross-account complexity heavily. AZ-305 is design-only and pairs with separate Azure implementation exams. If you architect across all three clouds in your day job, PCA's case-study format is the closest to a real consulting engagement.
Can I switch between ACE and PCA inside ClaudeLab?
Yes. Each certification gets its own evaluation, roadmap, readiness score, and error backlog. The active-cert switcher on the dashboard moves you between them in one tap. Streaks count any roadmap task on either cert.
Where do I see whether I am eligible for the pass guarantee?
On the dashboard, once all five conditions hold. The check runs after every milestone validation, and the eligibility flag flips automatically. Read the full breakdown of the conditions on the pass guarantee page, and the AI cert prep article for the structural reasoning behind the design.
Start your PCA roadmap
The cheapest possible signal is a 15 to 25 question CAT evaluation against the PCA blueprint. The output is a domain-by-domain skill estimate, a phase-by-phase roadmap weighted toward Cloud Solution Design, and your day-one task. If the evaluation tells you that you are still ACE-shaped, the system will say so honestly and point you at ACE first. If it tells you that you are 80% of the way there already, the roadmap will be short and dense, and the case-study reading will be the longest single block.
Either way, the measurement is more useful than another two weeks of unmeasured study. Open the PCA onboarding flow and start the evaluation. From there, practice sessions take over the daily cadence, and I pick the next task every time you reopen the app.