Cloud bills creep. A NAT gateway here, an oversized database there, a managed service you bought on a Marketplace plan that credits don’t cover — and suddenly you’re paying for infrastructure nobody scoped.
Here are five changes we make on almost every engagement that pay for themselves quickly.
1. Right-size before you reserve
Reserved instances and savings plans are great — after you know your real baseline. Buy them too early and you’ve locked in waste. Measure two to four weeks of actual utilisation first, then commit only to the floor.
2. Watch the Marketplace line
Third-party SaaS bought through your cloud Marketplace is not covered by sponsorship or free credits — it bills straight to your card. A single managed message broker or monitoring add-on can quietly cost more than the rest of your stack combined. Audit it.
3. Kill idle NAT and load balancers
NAT gateways and load balancers bill per hour and per GB, whether or not anything flows through them. Orphaned ones from torn-down environments are one of the most common sources of silent spend.
4. Move stateless workloads to scale-to-zero
Marketing sites, internal tools, and bursty workers rarely need an always-on node. Platforms that scale to zero between requests turn a fixed monthly cost into near-nothing.
5. Self-host what you can operate
Managed services are worth it when the operational burden is real. But a broker or cache your team can run inside an existing cluster moves the cost into compute you’re already paying for.
None of this requires heroics — just someone looking at the bill line by line. If you want a second set of eyes on yours, get in touch.