Every startup ends up on one of the big three clouds: AWS, Microsoft Azure, or Google Cloud. Founders often agonise over the choice. Having spent years inside AWS, my honest view is: for most startups, the choice matters far less than how you use it. Here's how to decide without overthinking.
What each is good at
- AWS — the biggest, most mature, widest range of services. The most engineers know it. Sensible default for most startups.
- Azure — strongest if you're a Microsoft shop (.NET, Office 365, enterprise agreements). Often the natural fit for B2B selling into Microsoft-heavy enterprises.
- Google Cloud — strong in data, analytics and machine learning; a clean developer experience.
For a typical web/app startup, all three will do the job comfortably. The differences are at the edges, not the core.
The pricing gotchas
This is where founders actually get hurt — not the sticker price, the surprises:
- Data transfer (egress). Moving data out costs money on all three, and it's easy to overlook until the bill arrives.
- Idle resources. Servers and databases left running "just in case" quietly burn cash.
- Managed-service premiums. Convenient services cost more than rolling your own — usually worth it, but know you're paying for it.
- Free-tier cliffs. Generous free tiers expire or have limits; budget for what happens after.
The bill rarely balloons because you picked the "wrong" cloud. It balloons because nobody's watching the usage.
How to choose without over-thinking
- Already Microsoft-heavy? Lean Azure.
- Data/ML at the core? Google Cloud is worth a look.
- Everything else / not sure? AWS — the safe, well-supported default.
Then stop deliberating and start building. You can move later if you genuinely need to (though migrations aren't free, so don't chop and change casually).
The thing that actually matters
Whichever you pick, the money is saved or wasted in how it's set up — right-sizing, turning off what you're not using, and keeping an eye on the bill. That's an architecture and discipline problem, not a vendor problem. I touched on the broader version of this in scaling startup architecture.
Choosing and setting up cloud infrastructure sensibly is core fractional CTO work — exactly the kind of senior judgement that stops a startup overpaying. Book a free call if you'd like a hand.