The cloud was supposed to free companies from having to buy infrastructure in advance. In many organisations, however, it has resulted in a bill whose amount is difficult to predict and link to business performance.
The problem is not that the cloud is inherently expensive. It arises when resource consumption grows faster than the value the company generates from them. As long as the CFO sees a rising bill and the CIO sees growing infrastructure needs, neither party knows whether the rise in costs is justified.
The cloud guarantees flexibility, not savings
Traditional infrastructure required the purchase of servers and licences in advance. The cloud has replaced this model with resources that are deployed on demand and billed according to usage.
However, this does not mean it is always cheaper. Its main benefits are speed, access to technology and the ability to respond to changing demand. Savings only materialise when a company is able to align its usage with its actual needs.
The economics of the cloud work best with seasonal, rapidly growing or unpredictable workloads. They are less obvious in the case of systems that utilise a similar level of computing power around the clock. In such cases, a company may end up paying for flexibility that it hardly ever uses.
Therefore, the right question is not: ‘Is the cloud cheaper?’, but: ‘Which applications justify the cost of its flexibility?’.
Cost is designed in alongside the architecture
The cloud bill is rarely limited to virtual machines. It also includes databases, storage, backups, monitoring, security, licences, support and data transfer.
It is precisely these additional services that often account for the difference between the cost estimate prepared before migration and the actual invoice. On AWS, costs can depend on the direction of data flow, the region, the availability zone and the network components used. Poorly designed communication between parts of a single application can therefore generate fixed charges, even though the computing power itself has been well optimised.
For businesses, this means that costs do not only arise at the end of the month. They are factored into the service right from the start, when choosing the architecture, the level of redundancy, the method of data replication and the duration of log retention.
Migrating existing systems without changing how they operate can be particularly costly. If an application was over-provisioned in an on-premises data centre, it may continue to use excessive resources after migration. The difference is that every unused unit now becomes a visible operational cost.
The cloud therefore does not automatically fix inefficient infrastructure. It can only change the way in which a company pays for that inefficiency.
A high bill does not always indicate a problem
The most common mistake is to assess the cloud’s efficiency solely on the basis of total expenditure. A 20 per cent increase in the bill may indicate a lack of control over resources. However, it may also result from serving 50 per cent more customers.
Cost alone does not allow us to distinguish between these situations.
That is why it is crucial to measure the cost per customer, transaction, order, API request or other business outcome. The FinOps Foundation points out that unit economics creates a common language for finance, technology and product teams. Costs can be measured, amongst other things, per customer, transaction, service, resolved case or revenue generated.
If the total bill is rising but the cost per transaction is falling, the company is likely benefiting from economies of scale. If, on the other hand, the number of customers is increasing and the cost of serving them is also rising, the infrastructure is becoming less and less efficient.
It is this deterioration in unit cost – rather than exceeding a certain amount on an invoice – that indicates the point at which the cloud ceases to improve the company’s economics.
FinOps cannot start with an invoice
Companies recognise the problem but still struggle to solve it. According to Flexer, 63 per cent of the organisations surveyed already have a FinOps team, yet the estimated proportion of wasted cloud expenditure has risen to 29 per cent. Cost management remains the main cloud challenge for 85 per cent of respondents.
This shows that a dashboard, report or centralised team alone is not enough. FinOps is ineffective if it only comes into play after the end of the month and is reduced to searching for underutilised resources.
The greatest impact is felt earlier: during application design, database selection, defining the required level of availability and planning for scalability. Cloud expenditure arises from code, infrastructure configuration and product decisions, not from the finance department.
CFOs and CIOs need a common metric
The CFO brings to the discussion the budget, margin targets, sales forecasts and an assessment of the risks associated with long-term commitments. The CIO understands resource utilisation, security requirements, application characteristics and the technical implications of cost-cutting.
Only by combining these perspectives can one determine whether a particular expenditure is a waste or a price paid for value. Redundancy increases costs but mitigates the impact of failures. Monitoring costs money but reduces the time taken to detect problems. Test environments do not directly generate revenue, but they reduce the risk of production errors.
The aim, therefore, is not the lowest possible bill. It is the best balance between cost, value and risk.
Discussions between the CFO and the CIO should focus on whether the cost per customer is falling, which products generate the highest expenditure, who is responsible for their cost-effectiveness, and whether paid resources actually support sales, quality or security.
In this way, the question ‘why has the bill gone up again?’ is replaced by a far more important one: ‘has the additional expenditure improved the product’s performance?’.
Not every application belongs in the public cloud
Sometimes optimisation means moving a stable, heavily used system to a private infrastructure or a hybrid model. This does not mean the entire cloud strategy has failed.
The Uptime Institute points out that high costs are the main reason for moving some workloads away from the public cloud, but the scale of this shift away from the cloud is often exaggerated. Companies are sticking with hybrid environments and selectively deploying applications where their characteristics best suit the cost model.
The cloud retains its advantage where rapid scaling, experimentation and access to advanced services are key. Private infrastructure may be better suited to stable systems with high and predictable usage. The decision should therefore not apply to the entire organisation, but to a specific workload.
Costs should rise more slowly than value
Rising expenditure is not a problem if the infrastructure supports a proportionally higher volume of sales, number of customers or scale of operations. A warning sign only appears when the cost of serving a single business unit begins to rise.
Mature cloud management is not about minimising every item on the invoice. It is about checking whether additional expenditure translates into greater scale, a better product, lower risk or faster time-to-market.
The cloud ceases to reduce costs when its usage grows faster than its business value. CFOs and CIOs should therefore jointly monitor the cost per customer, transaction or service, rather than assessing cloud strategy solely on the basis of the monthly bill.
