The Number Nobody Wants to Look At
Cloud spending is the one line item that grows faster than headcount, faster than revenue, and faster than anyone can comfortably explain to a CFO. The pitch when you moved to AWS or GCP was operational efficiency. The reality, for most engineering organisations, is a bill that doubles every eighteen months with no clear explanation for why.
The uncomfortable truth is that most of that growth isn't demand. It's waste.
Flexera's annual State of the Cloud report consistently puts cloud waste at 30–35% of total spend for the average organisation. In practice, for teams that haven't done a deliberate infrastructure audit in the last twelve months, the number is often higher. The waste isn't dramatic — it's not someone spinning up a GPU cluster and forgetting about it. It's a hundred small inefficiencies that compound invisibly.
Where the Money Goes
Idle and Underutilised Resources
The most common culprit is infrastructure that was provisioned for a workload that no longer exists, or that was right-sized for peak load and runs at 8% utilisation the other 22 hours of the day.
Development and staging environments are the classic example. They were built to mirror production. They run around the clock. They sit idle every night, every weekend, and every public holiday. A properly scheduled environment stops costing money when nobody is using it.
The same pattern appears with load balancers pointing at nothing, RDS instances attached to decommissioned services, S3 buckets storing data that nobody has accessed in two years, and EC2 instances left running because nobody was sure if they could safely be stopped.
Over-Provisioned Instance Families
When a service was first deployed, the engineering team picked an instance size that felt comfortable. Comfortable usually means "bigger than we need, so we won't get paged at 3am." That's a reasonable decision at launch. It becomes a cost problem when the instance family is never revisited.
Modern cloud providers offer detailed CPU, memory, and network utilisation metrics. For most services that have been running for more than six months, the data to right-size them accurately exists — it just isn't being looked at.
Unoptimised Data Transfer Costs
Data egress is one of the most consistently underestimated line items in a cloud bill. Traffic leaving a region, traffic between availability zones, traffic from services that aren't using a CDN — these aren't free, and the cost accumulates in ways that billing dashboards don't surface clearly.
An architecture that moves data efficiently — keeping frequently accessed data close to compute, using CDN edge caching appropriately, structuring inter-service communication to avoid unnecessary cross-region hops — can cut data transfer costs substantially without changing any application behaviour.
Reserved Capacity That No Longer Matches Reality
Reserved instances and committed use discounts are excellent tools for predictable workloads. They become problems when the workload changes and the commitment doesn't. A one-year reserved instance purchased for a service that was refactored onto serverless infrastructure eight months ago is pure waste for the remaining four months.
What an Infrastructure Audit Actually Involves
A serious cloud cost audit isn't about cutting corners or degrading performance. It's about aligning what you're paying for with what you actually need.
The process starts with a complete inventory of running resources mapped against actual utilisation data over a representative period — not just the last 24 hours, but the last 30 to 90 days, capturing peak load, weekend patterns, and batch job windows.
From there, the work is systematic: identify resources that can be scheduled off, right-size instances where utilisation data supports it, consolidate services where it makes architectural sense, convert on-demand spend to reserved or committed pricing where workloads are stable, and implement budget alerts and anomaly detection so new waste doesn't accumulate silently.
The result isn't just a lower bill. It's infrastructure that's easier to reason about, easier to maintain, and cheaper to scale.
What You Should Do Before the Next Invoice Arrives
Pull up your cloud billing dashboard and look at the top 10 line items by spend. For each one, ask: do I know exactly what this is? Do I know whether it's being used? Do I know if it's sized correctly?
If the answer to any of those is "not really," you have an audit to do.
The teams that run lean, efficient cloud infrastructure didn't get there by accident. They got there by making deliberate decisions about what they're running and why — and by revisiting those decisions regularly as their systems evolve.
If your cloud bill doesn't feel like it reflects the value you're getting, it probably doesn't. We've helped organisations reduce infrastructure spend by 35–60% through structured audits — without touching application performance.
Get in touch and we'll start with an honest assessment of where your current spend is going and what a leaner architecture would look like.