Hosting & deployment pricing

Every hosting platform bills on one of three models, and knowing which one you're looking at tells you more than any single number: a flat instance price (you pay for a box, whether or not it's busy), pure metered usage (you pay for exactly what ran, down to the second or the request), or a generous free tier with a steep cliff the moment you cross it. The three comparisons on this page each pit a different pair of these against each other, because the "cheaper" option depends entirely on how spiky your traffic is.

The three billing models, and who they favor

Flat-rate instance pricing (a fixed monthly box) rewards steady, predictable load — you know your bill in advance, and a quiet weekend doesn't refund you anything, but a traffic spike doesn't blow up your invoice either. It's the easier one to budget for as a small team.

Pure metered billing (pay per second of compute, per request, or per build-minute) rewards spiky or low-traffic workloads — a side project that gets hit twice a day can cost close to nothing — but the same model turns a viral moment or a runaway loop into a bill you didn't see coming. If you go this route, know where the spend alert lives before you need it.

Free-tier-then-cliff platforms (generous limits, then a hard jump to a paid tier) are the cheapest way to prototype and the easiest way to get surprised — the jump is rarely gradual, and it's usually triggered by something invisible to the person writing the code: sleep/wake behavior on an idle instance, a bandwidth allowance, or a build-minute cap.

The costs that don't show up on the pricing page

Egress and bandwidth. This is the single most common source of "wait, why is my bill 5x higher than the plan price" across every platform on this page. A generous compute or request allowance says nothing about what you pay to serve a large asset or a busy API to the outside world — that's metered separately, often at a rate that dwarfs the base plan once real traffic shows up.

The control plane you're not managing but are still paying for. On managed Kubernetes specifically, the advertised price is usually just the worker nodes — the control plane, load balancers, and cross-zone data transfer get billed separately and rarely show up in the headline comparison.

Cold starts and idle sleep. A free or cheap tier that "sleeps" an idle service after some minutes of inactivity isn't really free once you count the latency hit on the next real request — worth knowing before you put anything user-facing behind it.

How to actually compare two platforms here

Add up your realistic monthly compute, your realistic bandwidth, and — if it applies — your build minutes, at 2x your current traffic, not today's. The platform that wins at your current scale is not always the one that wins after your next traffic spike, and switching hosting under load is a worse week than picking carefully now. Every comparison below runs these numbers against the vendor's own current pricing, dated at the time it was checked.