Most organisations don’t have a single, obvious cloud waste problem. They have thirty small ones.
Flexera’s annual State of the Cloud report has consistently found that organisations waste around 28–35% of their cloud spend. That figure has stayed stubbornly high even as cloud usage matures, which tells you something important: the waste isn’t a beginner’s mistake. It’s structural. It builds up over time as infrastructure grows, teams change, and nobody has a clear ownership over the bill.
Here’s where it usually hides.
Over-provisioned compute
This is the biggest category by far. When teams size virtual machines and databases, they tend to provision for peak load — and then rarely revisit that decision. The result is servers running at 10–20% utilisation month after month. Cloud providers give you the tools to right-size these workloads; the bottleneck is almost always time and organisational will, not technical complexity.
The practical fix is to pull utilisation data for your core compute resources over a 30-day window and look for anything consistently below 40% CPU and memory. That’s where you start.
Development and test environments running 24/7
A production environment needs to be up around the clock. A development or testing environment almost certainly doesn’t. Leaving dev instances running overnight and over weekends is one of the most common and fixable sources of waste — automated scheduling can typically reduce their running hours by 65% or more without anyone noticing.
Forgotten storage and snapshots
Object storage is cheap per gigabyte, which is exactly why it accumulates unchecked. Old database snapshots, backup files that outlived their retention policies, log data nobody is reading — these add up. The same applies to unattached storage volumes left behind when their associated compute instances were deleted.
A storage audit usually surfaces more than people expect.
Data transfer costs
Moving data within the same cloud provider is often free or very cheap. Moving it between regions, between cloud providers, or out to the internet is not. Data transfer is one of the trickier cost categories because it’s hard to attribute — it doesn’t map neatly to a single team or workload. Architectures that weren’t designed with data locality in mind can generate significant transfer costs without anyone realising it.
Missed commitment opportunities
Cloud providers offer substantial discounts — typically 30–60% — for committing to a resource for one or three years. Reserved instances and savings plans make sense for workloads that run continuously and are unlikely to change significantly. Most organisations underuse these, partly from genuine uncertainty about future needs, and partly because someone needs to own the decision and nobody does.
A commitment analysis against your actual workload patterns usually identifies meaningful savings that require no infrastructure changes at all.
What to do with this
The honest answer is that a meaningful cloud cost review takes a few days of focused work: pulling the data, analysing usage patterns, mapping costs to workloads, and prioritising the findings by effort and impact. It’s not technically complex, but it does require someone who knows where to look and has the time to look properly.
If you’d like to understand what’s in your specific bill, that’s exactly what our Cloud Cost Review covers.