Latest / The Tech Career Podcast with Fexingo: Engineering Jobs, Interviews, and FAANG Career Strategy / How FAANG Engineers Choose Between GCP AWS and Azure
Transcript
- Lucas: You join a FAANG team as a senior engineer. You're assigned to build the next version of an internal data pipeline. First question: which cloud do you use? Most engineers assume the answer is already decided — but it's not, and the choice tells you a lot about how FAANG actually operates. Luna: I think a lot of people outside the industry imagine every FAANG is all-in on one cloud. But it's way more fragmented than that. Lucas: Right. Look at Google. You'd assume Google uses GCP for everything. But I've talked to engineers there who build services on AWS because certain GCP products just weren't mature enough a few years ago. And at Amazon, there's a famous internal mandate called 'dogfooding' — teams are strongly encouraged to use AWS, but even then, some teams run workloads on GCP for specific AI tools like Vertex AI. Luna: So the cloud choice at FAANG isn't top-down. It's often a bottom-up decision by the engineering team. Lucas: Exactly. And that's why understanding how FAANG engineers make that choice is useful for anyone preparing for a cloud engineering interview, or even just deciding which cloud to specialise in. Today we're going to walk through the three main clouds — AWS, Azure, GCP — and look at the real factors that drive the decision inside a FAANG team. Luna: Let's start with market share, because that's the baseline most people know. Lucas: Sure. As of mid-2026, AWS still leads with around 34 percent of the cloud infrastructure market. Azure is at 23 percent, GCP at 11 percent. But those numbers don't tell you which cloud is best for a given project. Inside a FAANG company, the calculus is completely different than for a startup or a bank. Luna: Because FAANG teams are building at a scale where even small differences in cost or latency translate into millions of dollars. Lucas: Exactly. And one of the biggest hidden factors is lock-in. Not vendor lock-in in the abstract sense — but specific product lock-in. Take Google's BigQuery. It's incredibly fast for analytics, and it's serverless, so you don't manage infrastructure. But once you build your entire data warehouse around BigQuery, migrating off GCP is brutally expensive, both in engineering time and data transfer costs. I've seen estimates that a full cloud migration for a mid-size FAANG service costs around ten million dollars and takes 18 months. Luna: So if a team picks GCP for BigQuery, they're basically committing to GCP for the next five years. Lucas: Right. And that's not necessarily bad — if BigQuery gives you a real advantage. But the team needs to be aware of that lock-in. Now compare that to AWS. AWS's strength is breadth. They have over 200 services. The lock-in risk there is more about operational lock-in — your team learns AWS's way of doing things, and the switching cost is cognitive. Your engineers become experts in Lambda, DynamoDB, and S3, and moving to another cloud means retraining everyone. Luna: We should talk about Azure too, because it's the default for a lot of enterprise faang adjacent companies, but inside pure FAANG it's less common. Lucas: That's true. Azure's killer feature is integration with Microsoft's enterprise stack — Active Directory, Office 365, SQL Server. So if your FAANG team is building a product that serves enterprise customers, Azure can be the pragmatic choice. But for most consumer-facing FAANG services, Azure is a distant third. The developer experience on Azure has historically been worse — the CLI is less intuitive, the documentation is fragmented. That's changing, but it still lags. Luna: So let's say I'm a senior engineer about to start a new project at a FAANG. What's my framework for choosing? Lucas: First, look at your workload's primary requirement. If you need the most advanced machine learning tools — especially large language model training — GCP's TPUs and Vertex AI are genuinely ahead. Google has been investing heavily in custom AI hardware for years. If you need the broadest set of services and the fastest iteration, AWS is usually the safest bet. If you need deep enterprise integration, Azure wins. Luna: Second factor: your team's existing expertise. A team that already knows AWS will be more productive from day one. Lucas: That's huge. I once spoke to an engineer at a FAANG who told me their team chose GCP purely because the tech lead had a strong preference for Kubernetes, and GKE — Google Kubernetes Engine — was considered the best managed Kubernetes at the time. That decision cascaded: they built their entire architecture on GCP, and now the team is deeply gcp specialised. Two years later, they're hiring for GCP experience specifically. Luna: So the cloud choice becomes a self-reinforcing loop. The initial technical decision shapes the team's future hiring and career paths. Lucas: Exactly. And that's why, for someone preparing for FAANG interviews, I think the most strategic move is to become very good at one cloud — but also have working knowledge of the other two. Because you'll almost certainly work in a multi-cloud environment at FAANG. Even if your primary platform is AWS, you'll interact with services running on GCP or Azure. Luna: Are there any specific cloud services that are so dominant that they almost force a platform choice? Lucas: Great question. AWS Lambda popularised serverless, but now both GCP Cloud Functions and Azure Functions are competitive. DynamoDB is a specific lock-in — it's a NoSQL database that's deeply integrated with AWS, and it's incredibly fast but expensive at scale. Some FAANG teams have moved away from DynamoDB to Cassandra or CockroachDB for cost reasons. So even within AWS, teams are making choices to avoid lock-in. Luna: We should mention cost modelling. Because the public pricing is almost never what a FAANG team actually pays. Lucas: Absolutely. FAANG companies negotiate massive enterprise discounts. A team might see a 30 to 50 percent discount off list price, plus committed-use discounts for reserving capacity. So the public cost comparison tables you find online are almost irrelevant. What matters is the internal deal your company has with each cloud provider. And that's usually confidential. Luna: But there's also the hidden cost of developer productivity. If one cloud's SDK is much better documented, that saves weeks of engineering time. Lucas: That's the metric that's hardest to quantify but most important. I've seen teams choose GCP purely because of their documentation quality for Kubernetes. And I've seen teams pick AWS because AWS has the most Stack Overflow answers. Developer velocity is a real factor. Luna: So for someone listening who's studying for FAANG interviews, what should they focus on? Should they pick one cloud and become an expert, or learn all three? Lucas: I'd say become an expert in one — probably AWS, because it's the most common — and then learn the equivalents in GCP and Azure well enough to discuss trade-offs in an interview. A system design interview might ask: 'Design a real-time analytics pipeline.' If you immediately jump to AWS Kinesis, that's fine, but if you can also say, 'Alternatively, we could use GCP Pub/Sub with Dataflow, which might be simpler if the team is already on GCP,' that shows real depth. Luna: Another angle: the cloud certification path. Do FAANG engineers value certifications like AWS Solutions Architect? Lucas: Honestly, not that much for senior roles. Certifications are seen as entry-level proof of knowledge. What matters more is hands-on experience. But there's a catch: if you're coming from a non faang background, a certification can help you get past the resume screen. Once you're in the interview, though, they'll ask you real-world questions, not certification trivia. Luna: So the best preparation is actually building something on multiple clouds. Even a small project — like a serverless API that runs on both AWS and GCP — would teach you the differences in cold start times, pricing, and deployment. Lucas: Exactly. And that project would be a great portfolio piece. Speaking of which, we should talk about how Netflix and Pinterest made their cloud choices. Netflix is famously all-in on AWS. They moved their entire streaming infrastructure there starting in 2008. But they also built their own tools on top of AWS — like Chaos Monkey — to handle failures. So they're deeply locked into AWS, but they've also created their own abstractions. Luna: And Pinterest started on AWS but then migrated many services to GCP in 2016, citing better performance for their machine learning workloads. Lucas: Right. That migration took years and was a huge engineering effort. But it paid off for them because GCP's AI tools were a better fit for their recommendation algorithms. So the lesson is: platform choice is strategic. It's not just about today's needs but about where you expect your engineering needs to go in the next three to five years. Luna: That's a good point. Now, something we haven't touched on: the role of open source. How does that factor into the cloud decision? Lucas: Huge factor. Kubernetes is the obvious example — it's open source, but GKE is the most polished implementation. Similarly, TensorFlow is open source, but it's deeply integrated with GCP. AWS has its own open source projects like Firecracker for serverless. FAANG engineers often prefer platforms that offer the best managed services for the open source tools they already use. Luna: So the cloud is not just infrastructure — it's a development platform that shapes your entire engineering culture. Lucas: Exactly. And I think that's the real takeaway from this episode. When you're choosing a cloud as a FAANG engineer, you're not just picking a vendor. You're betting on a set of tools, a community, and a future direction. And that decision will affect your team's productivity, your hiring, and your own career growth for years to come. Luna: Speaking of career growth, one thing I appreciate about this show is that we don't run ads. It means we can talk freely without trying to sell you a cloud certification course or a resume service. Lucas: Yeah, that's intentional. We keep this show ad-free because we think the advice is more useful when it's not being sponsored. If you find value in that, you can support the show at buy me a coffee dot com slash fexingo. No pressure — just a way to keep this going without compromising the content. Luna: Exactly. And we're back to the topic: if you're prepping for a FAANG interview, I'd say the cloud question is a great one to study deeply because it ties together system design, cost awareness, and strategic thinking. Lucas: And remember: the cloud is always changing. What's true today about GCP vs AWS vs Azure might shift next year. The skill that lasts is being able to evaluate trade-offs quickly and pragmatically. That's what FAANG interviewers are really testing. Luna: Thanks for listening. We'll be back next week with another topic that helps you level up your tech career.