Every month the cloud bill lands a little higher, and almost no one can say exactly why. According to the State of FinOps Report 2025, a survey of 861 respondents from large companies together accounting for $69 billion in cloud spend, cutting waste is now the top priority for whoever owns that budget — ahead of closing the books accurately or forecasting next month's spend. The same survey shows the other side: only 15% of those teams actually make decisions by looking at cost. The priority exists on paper; the routine doesn't.
The effect shows up on the invoice. A study by Backblaze with Dimensional Research, run between May and June 2025 among 403 companies storing more than 250 terabytes in the cloud, found that 95% of them had already been hit with a charge they didn't expect — most often tied to data leaving the cloud, the fee charged for moving information out of where it's stored.
For whoever signs off on the budget, that's a number climbing with no decision behind it: a machine left running that nobody uses, a test environment that should have been shut down months ago, backups piling up since forever. In countries outside the dollar zone there's an extra layer: since a large share of the international cloud invoice is billed in US dollars, the exact same usage costs a different amount every time the bill closes — it rises when the exchange rate rises, with not a single extra byte consumed.
Where the bill climbs without anyone deciding
A Coleman Parkes survey commissioned by Harness polled 700 technology professionals — half developers, half engineering leaders — at companies with more than a thousand employees in the United States and the United Kingdom. The picture: 43% of those teams have no real-time visibility into what's sitting idle in the cloud, 39% can't see what was orphaned once a project ended, and 33% can't say what's over- or under-provisioned for what it actually uses.
Even more telling: 55% of developers admitted that reserved-capacity purchases — paying upfront for a discount — are decided by guesswork, with no data behind the number. The result is what the industry has nicknamed a zombie workload: the environment that should have been switched off when the project ended keeps running and billing, because it never errors out, never brings anything down, it just quietly adds to the invoice.
If this happens at large companies with entire teams dedicated to watching cloud spend, it's reasonable to expect the same — or worse — at a smaller company where nobody holds that role at all.
Data egress makes it worse. In the Backblaze study, 55% of companies named the cost of pulling data out of the cloud as the main reason they don't switch providers, even when the current one is no longer the best value. The company stays locked in not by service quality, but by the price of leaving — and while it stays, it keeps paying for what was never reviewed.
Why the usual approach doesn't fix it
The usual approach, at most companies, is to look at the bill only once it arrives — never before. Someone requested a test server six months ago, the project ended, and nobody sent the shutdown order. The environment is still there, billing a small amount every month that on its own doesn't draw attention, but added to ten other forgotten look-alikes turns into an entire line of the budget.
In Brazil, the problem has a structural root. Radar da Nuvem, a report released in 2026 by Samax in partnership with Talentum Ventures, drawing on data from 133 Brazilian technology companies collected in 2024 and 2025, found that a third of them can't say whether they went over their cloud budget in the period measured. In more than 80% of cases, the only side watching that spend is the technical team — development, IT, infrastructure — with finance nowhere near the conversation.
It's a division of labor that doesn't add up: whoever decides to keep an environment running is rarely the one who later explains to leadership why the bill came in higher. Without someone watching both sides — what the operation needs and what it costs — spend climbs by inertia, not by necessity.
Switching cloud providers doesn't fix it on its own either, because leaving itself is expensive, as the Backblaze study showed. What fixes it isn't where the company stores its data. It's having someone with a routine of checking what's running, for what purpose, and since when.
What has to be in place
Controlling a cloud bill that already exists takes routine, not good intentions. Six verifiable mechanisms:
Every resource tagged with who requested it. A server, database, or test environment goes live already carrying the name of the person or team responsible — so when the project ends, there's someone to notify that it can come down.
A monthly bill review with whoever requested the environment, not just the technical team. The invoice passes by whoever will actually use it before it becomes a surprise at the budget meeting.
Shutdown of what isn't used. A forgotten test machine, an idle staging environment, a duplicate copy of the same data: all of it gets hunted down and turned off as routine, not only when someone happens to remember.
An alert when spend crosses a ceiling. A warning fires before the month closes, not after — leaving time to understand the cause and act while the outcome can still change.
A cost estimate before any new environment goes live. The decision to spin something up already comes with the estimated cost, instead of being discovered later, on the invoice.
A review of what backups keep and for how long. Not everything needs to be stored forever; the retention period is also a decision, not a factory default nobody ever revisits.
This is how Skills IT works: it tracks the client's cloud usage, flags what's running without being used, and warns before the invoice closes higher than it should.
The payoff for whoever decides
The payoff of watching the bill as a routine shows up first in the budget. A predictable invoice is one that leadership can actually plan around — without the shock of discovering, at month's close, that the amount rose without anyone having requested anything new.
There's also a time payoff. When shutting down what's unused and reviewing what's running is routine, the IT team stops firefighting the invoice and goes back to what the company actually needs: support, security, the projects that had been on hold waiting for room on the calendar.
And there's a decision-quality payoff. When the cost of spinning up a new environment is estimated beforehand, rather than discovered afterward, whoever approves the budget decides with information — not with last month's invoice as the only reference for the next one.
A starting checklist
Before treating the cloud bill as a closed subject, it's worth running through a simple checklist with whoever handles IT:
- List what's running today. A quick inventory of active servers, databases, and test environments usually turns up things nobody remembered still existed.
- Ask who owns each resource. If the answer is "not sure," that's the first candidate for review or shutdown.
- Set a spending ceiling and an alert. Defining a monthly reference amount, with a warning before it's crossed, takes cost control out of the surprise column and puts it into routine.
- Bring finance into the cloud conversation, even if only once a month. The invoice stops being a subject only the technical team discusses.
- Review what backups keep and for how long. Old copies with no defined expiration date tend to be the easiest item to cut without risk.
