I Stopped Overengineering My Cloud Setup (And You Should Too) | bscale

Look, I get it. You just learned about AWS Lambda and suddenly you want to architect your side project like you're building the next Netflix. I've been there. After 4+ years in tech and running multiple startups, here's what I wish someone told me earlier.

Your cloud choice should match your scale. That's it.

The Real Math Nobody Talks About

Major cloud providers like AWS, GCP, and Azure are stupidly cheap—if you set them up right. But that "if" is doing a lot of heavy lifting.

The time it takes to properly configure IAM roles, set up VPCs, optimize your Lambda cold starts, and figure out why your CloudWatch bill is suddenly $200? That's dev hours you're not spending on your actual product.

Here's my golden rule: If you can afford to hire a dedicated cloud engineer, you should be on the majors. If you can't, you probably shouldn't.

An enterprise with a $100k monthly cloud bill? Yeah, that 1% optimization matters. Your startup burning $10/month on infrastructure? Bro, just ship.

What I Actually Use (And Why)

For Enterprise & Scale → AWS / GCP

I've used both extensively—AWS for serverless architectures and CDK-based infrastructure, GCP for data-heavy projects and Firebase-powered apps. They can obviously do so much more, but here's where I've found them shine.

AWS has the best serverless experience, hands down. The CDK makes infrastructure-as-code almost enjoyable. GCP is arguably the most comprehensive platform out there—banks love it, and Firebase is chef's kiss for devs who prefer working with SDKs.

Use them when: You're past product-market fit, have dedicated DevOps bandwidth, need tight ML/data integrations, or your client specifically requires it (enterprise contracts, compliance, fintech, etc.)

For Actually Shipping Fast → Vercel & Railway

For serverless, Vercel is my go-to. Deploy a Next.js app in seconds, automatic previews, edge functions—it just works. For non-serverless stuff (long-running processes, websockets, background jobs), Railway is unbeatable. Simple CI/CD, autoscaling, postgres in two clicks.

I love both. Dead serious. They have everything you need to scale to 10k DAU without thinking about infrastructure.

Use them when: You're validating an idea, building an MVP, or honestly just want to focus on code instead of YAML files.

Honorable Mentions

Sevalla — Haven't used it personally, but trusted friends vouch for it. Worth checking out.

Render — Same boat. Solid reputation in the dev-first space.

The Complexity Ladder

If you're learning, I'd recommend this progression:

Start with dev-first (Railway, Render) — Ship things. Get dopamine. Learn what infrastructure needs to do.
Graduate to major dev-friendly (Cloud Run, AWS CDK, Amplify) — Training wheels, but you're touching real cloud primitives.
Go full major (raw AWS/GCP/Azure) — When you actually need it, or when you're getting paid to know it.

The majors have dev-friendly options now, but they're still relatively harder to configure than the dev-first providers. Don't let anyone shame you for using Railway. You're not a worse engineer for choosing speed.

The Bottom Line

Cloud knowledge is valuable. I'm not saying don't learn AWS—I'm an AWS Community Builder, I clearly think it matters.

But being good at cloud means committing serious time to it. That's a career path, not a prerequisite for launching your startup.

Build first. Optimize later. Ship the thing.

What's your go-to stack? I'm always curious what other builders are using. Hit me up.