Cloud spend is the new power bill – but who is watching the meter?

In yester year, your old datacentre bill arrived once a month, whether hosted in by your business or via a private cloud service, and the cost rarely surprised you. You were also likely to make long term CAPEX commitments, which made forecasting a relatively trivial matter across that period.

The reality now

Today, the invoice is a plethora of cloud dashboard resources scrolling in real time, and it's trending up.

Billing for cloud services such as compute, storage and software is, in a way, similar to how utilities like electricity or water services are charged.

Pay-as-you-go, the name for this usage-based pricing, is a model where you only pay for the resources you consume, without upfront fees or termination charges.

It is a double-edged sword that can quietly erode margins if nobody is tracking "the meter".

Why this matters to cloud users

Gartner expects end-user cloud spending to increase 21.5% globally to USD$723B in 2025 ¹.

When your cloud costs reach an eight- or nine-figure annual run-rate, every single digit percentage point of waste quietly turns into board-level money.

To give you an idea of how that plays out in practice and why the CIO/CTO (and CFO) need to care, the table from Flexera's 2025 State of the Cloud Report and CRN Annual Public Cloud and Saas Bill in 2024 Report are useful ^{2, 3}:

As both reports show, significant waste is being reported in cloud spend, causing company margins to be quietly squeezed.

Companies that have a high proportional cloud spend to revenue are particularly vulnerable to cloud waste spend further reducing their profit margins, as seen by the $20M spent by the AI/ML company on revenue of $30M. Just 3% of trivial waste causes them $600k of costs and the average 27% of waste is costing them $5.4M. This is causing an 18-percentage point margin drag to the point where the business may no longer be profitable.

Other reports are claiming that waste estimates by executives may be underestimated and the true waste values are in excess of 40% ^{5, 6}.

These include a list of common repeatable mistakes happening within organisations:

• Excessive log/data retention (50% of respondents)
• Incorrect resource sizing/type (49%)
• Costly AI experiments (48%)
• Outdated instance types (48%)
• Misconfigured networking (48%)
• Development systems running during off-hours (47%)

Flexera's report went on to state that 84% of respondents, the highest of all the challenges queried, struggled to manage cloud spend. This is a pervasive problem and not exceptional. Combined with cloud spend expected to increase by 21.5%, many businesses may be at risk of reduced margins, if revenue doesn't increase proportionally.

Why cloud spend is harder than a power bill

Cloud is easily consumed as a utility but managing the spend can be complex. The “flat rate” approach is expensive, driving a need to embrace granular billing, dynamic pricing and complex pricing structures. A multitude of "meters" now exist, with each company implementing their service-specific rules. Some companies charge on a per second basis, others on a per API call and even on token utilisation, common with AI interactions.

As cloud services are increasingly treated as commodities, IT teams have a harder time predicting usage and seasonal variations. New features are constantly released, encouraging teams to run new experiments, improve applications and create new experiences.

On top of these factors, instead of a single pricing lever, there are dozens of switches and levers, such as instance autoscaling, reservations and data transit costs.

Ignoring the cloud meter is risky and even if the unit price drops, the bill can increase with unmonitored usage.

Who should "watch the meter"?

Meter watching is a multi-departmental, multi-business level responsibility. Having control in depth ensures that many eyes are seeing the meter and actioning, where necessary.
Establishing cross-functional FinOps responsibilities is likely to be key in ensuring this a successful venture. This will include:

1. Executive sponsor - usually CIO/CTO with CFO partnership
2. Cross-functional FinOps team (engineering, finance, product, data analysts) accountable for:
   1. Allocation and tagging standards
   2. Budget guardrails and alerts
   3. Show-back or charge-back models
3. Engineering squads – shift-left cost of ownership e.g. alerts in CI/CD
4. Automation – functionality to hibernate idle dev/test, validate tags or reject unapproved SKUs at peer review time or as part of guard rail controls in the cloud environments. These are areas where a decision maker may need help to understand the financial benefits of a technical change made through automation.

Metering and metrics that resonate with boards

Having KPIs that resonate with the board can not only make cost justification easier, it can also help align and simplify the reporting necessary to make end-to-end understanding of the business.
Some examples of these are:

• Alignment of cost to value – by understanding unit costs like $/API call (cost per transaction), $/ML inference (cost per query).
• Potential annual cost savings – from percentage waste e.g. idle + over-allocation can quantify the optimisation gap.
• Confidence in budget and forecasting – forecast variance e.g. plan vs actual.
• Purchase vs usage risk – commit-utilisation e.g. percentage of reserved instance or savings plans.
• Environmental, social and governance goals – carbon per $ helps with alignment.

Five quick wins that you can drive this quarter

Facilities managers adopt an established framework and processes that help them avoid surprises. For example, reading the meter every day can raise warning flags, preventing an outsized end-of-year bill because of a leaky chiller.

Your cloud estate deserves the same discipline. Start a FinOps stand-up, monitor the costs and make watching the meter part of the engineering culture before your CFO flips the breaker.

With all the above in mind and the mention of trivial waste that is affecting your bottom line, here are some quick wins for CIOs/CTOs to achieve early cost savings that can be turned back to profit:

1. Implement methods to identify and track 90% of resources, such as tagging and naming standards, and are associated with an owner and a cost-centre.
2. Kill-switch policies to target stale resources, such as unattached volumes that are older than 30 days in non-production environments. Destroy lab environments after set dates.
3. Rightsizing the top 10 workloads is something that usually nets 10-15% savings immediately.
   
4. Forecasting and commitment strategy refresh to ensure that reserved instances and savings plans are committed to before a new AI project ramp.
5. Monthly FinOps show-back slide in the exec pack, ensuring cloud expenditure is visible and trends and variations are noted throughout the business.

¹ Global cloud spend to surpass $700B in 2025 as hybrid adoption spreads: Gartner
² Flexera 2025 State of the Cloud Report
³ Annual Public Cloud And SaaS Bill In 2024: Report
⁴ EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortisation) is calculated by adding back interest, taxes, depreciation and amortisation expenses to net income.
⁵ New Survey Finds Cloud Waste is On the Rise - Driven by Preventable Mistakes, Inefficiencies, and New AI Initiatives
⁶ Stacklet Cloud Cost & Usage Optimization Survey