Introduction
This is Part 2 of my PCA certification post series — my experience renewing the PCA certification. Part 1 focuses on my experience—how I prepared and what the exam was like first time. Part 3 shares the resources I used and my notes on GCP products. Part 4 includes my notes on case studies.
To renew PCA certification, you can choose either:
- Renewal exam — 1 hour, 25 questions, 1 case study, $100, during your renewal eligibility period (60 days before and 30 days after expiration date)
- Classic exam — 2 hours, 50 questions, 2 case studies, $200, any time
To reduce the cost, I chose the Renewal exam, which "focuses on practical generative AI solutions in enterprise settings"
How I Prepared
I scheduled the exam three weeks in advance.
1. Focus on Vertex AI and AI-related services (10–12 hours over two weeks)
I don't use Vertex AI daily. I had a general idea of what it is and which products are available on the platform, but I needed a deeper understanding.
Google Cloud provides a suggested renewal learning path on Google Cloud Skills Boost . I checked it briefly, but I didn't find it very suited for me (similar to my experience the first time — see Part 1). In my opinion, many of the courses felt too marketing-focused. Often, a long video would contain only a few really important points, and the rest was general explanations.
I mainly completed labs to better understand the platform. In some labs, I simply explored the interface, navigated through services, and tested features. I find it helpful to attach a visual understanding of products to what I read in the documentation. I liked "Multimodal Retrieval Augmented Generation (RAG) using the Gemini API in Vertex AI." lab.
Generally I didn't rely heavily on that path.
Mostly, I read Vertex products documentation and took notes. Whenever I felt lost, I checked ChatGPT, Claude, Gemini to clarify things. Quite often, some products are abstractions or platforms built on top of lower-level AI services (for example, Vector Search, Vertex AI Search, Retail Search). Others are very similar products but different versions. Sometimes the naming was confusing, especially when the same name was used for different products in different contexts (for example, Video API, Vision API, and Vertex AI Vision). Sometimes it wasn’t obvious for me how to clearly sort them out within the platform.
If helpful, you can check my notes. A useful tip: if you’re unsure what a product actually does, read its pricing page — it’s usually much more concrete and practical than documentation or marketing materials.
2. Review of classical questions (around 10 hours in the final week)
I reviewed my notes from the first time I took the exam to refresh knowledge of products I don't use daily. I also prepared notes for potential case studies.
This time, I found some additional resources and used them to double-check my understanding and coverage. It helped confirm that I had covered most topics.
On the last day before the exam, I reviewed practice questions I could find (see Part 1).
What the Exam Was Like
1. In my opinion, a large portion of the exam still focuses on classic cloud architecture topics, even when questions are framed in an AI context.
There were many traditional questions about:
- Storage classes and cost optimization (e.g. lifecycles)
- Kubernetes and Compute Engine (e.g. cost optimization using Spot VMs)
- Security and Networks (e.g. hybrid connections between on-premises environments, Google Cloud, and other clouds)
2. The AI-related questions were mostly practical and product-focused.
You need to clearly understand what each service does (e.g., Vertex AI components, Workbench, Builders, etc.). However, I don’t remember many highly specific, configuration-level questions. Some questions required understanding trade-offs between using pre-trained models and training custom models, as well as knowing when conversational AI platforms like Dialogflow are appropriate.
Overall, I found the exam easier than the original one. But I would strongly recommend reviewing core cloud architecture concepts and refreshing knowledge of specific product capabilities. Don't be stressed — it's manageable.
One thing I was unsure about was the scoring logic. For example, if there are 50 questions, what is the allowed error rate? And for renewal, I had about 3–4 questions where I wasn't confident in my answer. But overall, the exam felt fair.