Hey everyone, newbie here 👋. I’ve been trying to wrap my head around the actual costs of running CI/CD pipelines for our small dev team. We’re currently on a single cloud provider, but I keep hearing we should be shopping around.
I stumbled on a few blog posts and reports that try to benchmark the cost-per-minute for compute across AWS, Azure, and GCP. Some even threw in GitHub Actions and GitLab CI minutes. The numbers were all over the place! One said GCP was cheapest for Linux builds, another said Azure had a hidden edge with their free monthly grant.
Honestly, it left me more confused. Has anyone here done a real comparison recently? I'm especially curious about mid-tier teams doing maybe 5,000-10,000 build minutes a month. Did you factor in egress costs or the price of managed runners vs. self-hosted?
I’d be so grateful for any real-world insight or even a glimpse at a sanitized invoice. Just trying to avoid a nasty surprise on our first big cloud bill 😅.
The obsession with cost-per-minute benchmarks is missing the forest for the trees, especially for a small team. Those blog posts are often just marketing fodder.
For your 5,000-10,000 minute range, the real variable isn't the listed compute price. It's whether you're using ephemeral self-hosted runners on a spare VM you already have running versus a managed service with per-minute billing. A single small, long-lived VM on your current provider for self-hosting runners will almost certainly crush the cost of any managed offering, and you stop worrying about minute-by-minute rates entirely. The egress cost panic is usually overblown unless you're moving terabytes of artifacts daily.
A sanitized invoice won't help you, because my setup won't look like yours. The nasty surprise comes from letting pipeline duration bloat, not from picking the 'wrong' cloud. Start with what you have, run it for a month, and see where the money actually goes.
monoliths are not evil
You're right about self-hosted being cheaper, but you've skipped over the main headache: the hidden labor cost.
That spare VM you mentioned isn't free to manage. Someone has to patch it, secure it, handle scaling when ten builds kick off at once, and monitor its uptime. If that's your senior dev's time, your cost-per-minute just doubled.
Show me an actual bill with the ops hours line-item. Then we can talk.
show me the bill