Where companies most often overpay in the cloud
A cloud bill rarely grows because of one big mistake. It grows because of dozens of small ones - instances nobody scales down, disks left over from deleted servers, licences nobody cancels. Here are the most common sources of wasted spend on AWS and Azure.
FinOps often gets reduced to "switch to cheaper instances." In reality it's mostly a visibility problem - companies don't know exactly what they're paying for, because nobody checks regularly. Here are six places where money leaks out most often.
Six most common sources of overspend
Oversized instances
Servers sized "to be safe" and never revisited. Actual CPU/memory utilisation is often a fraction of what the instance is provisioned for.
Orphaned resources
Disks, snapshots and IP addresses left behind by deleted instances that keep getting billed - because nobody cleaned them up.
Missing reservations / savings plans
Workloads that have run continuously for years still pay on-demand prices, when a long-term commitment would cost noticeably less.
No autoscaling
Capacity set for peak load and left there 24/7 - even though nighttime or weekend traffic drops to a fraction of that.
Duplicate and unused licences
SaaS tools and licences that survived a team reorg or migration - nobody actively uses them, but they're still billed.
Cross-region and cross-account data transfer
An architecture that unnecessarily moves data between regions or accounts pays for transfer that could be designed away.
The common thread behind most of these: nobody checks them regularly. A one-off audit finds the current state, but without visibility (tagging, per-team/per-project reporting) the problem creeps back within months.
How it's usually fixed
- Audit actual utilisation - compare provisioned capacity against real load over the last 30-90 days, not what someone assumed when it was set up.
- Clean up orphaned resources - often the fastest win, since it requires no architecture change.
- Tagging and reporting - so it's visible which team or project is paying for what; without it, optimisation happens blind.
- Review commitments - reservations/savings plans for stable workloads, autoscaling instead of fixed capacity for variable ones.
The typical savings range I find with clients is up to 20% of cloud spend, with no impact on performance or availability. Environments that have never been audited tend to land at the higher end of that range.
When to tackle it
The best time is before the cloud bill starts to hurt - a quarterly check prevents small inefficiencies from compounding into a big problem. If an audit has never happened, even a one-off review usually finds quick wins within a week or two.
Want to know exactly where you're overpaying?
On a free consultation we'll walk through your environment and I'll tell you where I see the biggest opportunity to save.
Book a free consultation →