When a cloud bill surprises the board, the instinct is to negotiate with the provider. The real leverage is upstream, in decisions made long before the invoice.
Most runaway cloud spend traces back to architecture, not pricing: over-provisioned services running at the peak all month, data moving across boundaries that quietly meter every gigabyte, environments left on because nobody owns turning them off, and workloads sized for a spike that happens twice a year.
Design for the trough, scale for the spike
The elegant cloud architecture doesn’t pay for peak capacity all the time. It sits small at the baseline and scales cleanly when demand arrives. Getting there is a design exercise — autoscaling that actually works, storage tiers matched to access patterns, and a hard look at where data flows and what that movement costs.
A right-sized architecture saves more than any discount you’ll ever negotiate.
Make cost visible to the people who create it
- Tag and attribute spend so teams see the cost of their own decisions.
- Right-size continuously — provisioning is a habit, not a one-time task.
- Kill idle environments automatically; don’t rely on someone remembering.
- Treat data egress as a first-class design constraint, not a line item you discover later.
Done well, cost optimisation isn’t about spending less on a worse system. It’s about paying for what you actually use — which usually means a cleaner, more resilient architecture as a side effect.