Cost Cloud Build minutes by machine type after cache savings and the free tier, and check concurrency queueing.
A faster build machine costs more per minute but runs for fewer minutes, so the bill is the product of the two — an e2-highcpu-8 at roughly five times the rate is only worse if it delivers less than a five-fold speed-up. Caching cuts minutes directly, and the free monthly allotment is subtracted afterwards, which is why small teams often pay nothing at all. Build queueing wastes engineer time rather than money, and the queue model here shows how close the concurrency limit is to saturation before developers start complaining. Confirm the per-minute rate for each machine type in the Google Cloud pricing calculator, and remember the free allotment is per billing account.
Cloud Build Cost
billed minutes = builds × (build time ÷ machine speed-up) × (1 − cache saving) − free minutes; cost = billed minutes × the machine's per-minute rate.
billed minutes = builds × (build time ÷ machine speed-up) × (1 − cache saving) − free minutes; cost = billed minutes × the machine's per-minute rate. A faster build machine costs more per minute but runs for fewer minutes, so the bill is the product of the two — an e2-highcpu-8 at roughly five times the rate is only worse if it delivers less than a five-fold speed-up. Caching cuts minutes directly, and the free monthly allotment is subtracted afterwards, which is why small teams often pay nothing at all.
Build queueing wastes engineer time rather than money, and the queue model here shows how close the concurrency limit is to saturation before developers start complaining. Confirm the per-minute rate for each machine type in the Google Cloud pricing calculator, and remember the free allotment is per billing account.
This calculator takes 11 inputs: Builds per month, Average build time, Build machine type, Price per build-minute on the default machine, Price per build-minute on e2-highcpu-8, Price per build-minute on e2-highcpu-32, Free build-minutes per month, Build time removed by caching, Speed-up on the larger machine, Concurrent build slots, Peak builds submitted per hour. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.
When the build genuinely parallelises — large compilations, big test suites — a 32-vCPU machine can pay for itself and return engineer time as well. For a build that spends most of its time pulling dependencies over the network, the extra vCPUs sit idle and you pay the higher rate for nothing.