How businesses can address Cloud Bill Creep

5 min read
Read with AI

Open in

ChatGPT Claude Perplexity

This page

Copied to clipboard
How businesses can address Cloud Bill Creep

Over the last decade, cloud services have expanded to include offerings like workflow automation, data analytics and machine learning. It’s a sprawling, always-on collection of services, teams, providers and AI tools that can no longer be defined by a few predictable, centrally managed workloads.

Keeping track of what the cloud mix is costing organisations has become harder, with reports estimating that 29% of organisational cloud spend is now wasted. If a third of the office supplies budget went missing every year, someone would probably notice. Cloud spend tends to get a pass.

For businesses that depend on scalable cloud infrastructure, having a reliable way to manage costs without overspend eroding the benefits of cloud adoption is critical. It starts with knowing what services or tools they have running, why they have them and if they are still doing what they’re supposed to do.

What is causing cloud bill creep

The cause is usually something ordinary but easy to miss. Resources nobody owns keep accruing charges, dev and test environments are left running all weekend, and snapshots and logs pile up because nobody gave them an expiry date. An unplanned spike in data transfers can add to the total, while autoscaling that scales up quickly is often in no hurry to scale back down.

Budget alerts that go off every month and are routinely ignored are another common feature, and one that says more about an organisation's culture than the bill itself. At that point it’s less an alert and more a monthly newsletter. The biggest red flag, however, is spend that cannot be tied back to a product, a service or a customer. If nobody can say what it is for, nobody can track ROI.

For that reason, unit cost is a more useful measure than the total invoice. Depending on the business, that could be cost per customer, per transaction, or per AI inference. The total should rise as the business grows, while unit cost shows whether the organisation is becoming more efficient, or less.

I don’t think a high bill is the most useful warning sign. The one I’d worry about is a bill that’s growing faster than the business it supports. If spend keeps increasing while customer numbers and transactions stay flat, it’s important to understand why. This often happens after a migration or product launch with old and new systems running side by side, or when paying for extra capacity. We see this a lot after migrations. The old system was due to be switched off “next quarter,” and three quarters later it’s still running. The customers and revenue they were meant to support take time to arrive, and sometimes never do.

Find out what the money is for

Spend that can’t be explained is where cloud creep can start to take hold. Recognising and addressing cloud creep begins with investigating the organisation’s five largest cloud cost categories.

The next step is to ask three questions about each: who owns it, what does it support, and is the cost in line with the value? Then look at how usage changes over time, which creates a baseline without needing a huge transformation programme.

Unit cost and unexplained spend help to highlight where the problem lies, but don’t often reveal what’s safe to switch off. Although the instinct is to go after whatever looks most expensive, what looks idle might be a failover environment, or it might exist for a compliance reason you can’t see from the console.

Teams also need to address the low-risk, reversible items first by shutting down non-production environments at night and on weekends and cleaning up orphaned resources. This involves cutting storage, snapshots and logs that are kept far longer than anyone needs, and right-sizing based on actual usage, not on what someone might have estimated three years ago.

After that, costs should be a part of normal delivery, coming up in architecture reviews, product planning and engineering decisions, the same way performance, security and availability do. This process works best as a habit rather than an annual clean-up.

Share responsibility across finance, product and engineering teams

Keeping these costs under control requires making cost management part of everyday decisions, with finance, product and engineering teams sharing responsibility. It’s likely that if you hand it to one team as an infrastructure problem, it will not get solved, whereas a shared FinOps approach gives those teams a common language.

Instead of finance telling engineering to cut costs, they discuss the spend needed to deliver a specific service or customer experience. Sometimes the answer is to reduce capacity. Sometimes it is to spend more on resilience or automation, where the business case holds. Either way, the answer only emerges when the right people are in the room.

This shift is already underway. Research found that 63% of organisations have FinOps teams and 71% run a Cloud Centre of Excellence. These structures bring finance, engineering, business units and governance closer together. Yet 85% still call managing cloud spend a top challenge. So having a FinOps team clearly isn’t enough on its own. Someone still has to own the decisions.

Use AI to help manage cloud costs

Even when ownership is shared across teams, visibility can be clouded as no single team can keep an eye on every resource in a large estate. AI can help close that gap, though it does cut both ways. Research points to AI workloads as one of the drivers behind the recent rise in wasted cloud spend, yet the same technology is good at spotting patterns a human might miss. It can flag unusual spend, surface underused resources, forecast demand, and turn technical data into information that product and finance teams can actually use.

What AI shouldn’t do is make the final call. A tool might recommend shutting down an underused server without knowing it’s the failover, or that a regulator requires it to be there. This is where the ownership work pays off: AI raises the flag, and the resource owner, who knows what it supports, checks the context before anything changes.

Where to start

If your cloud bill has been creeping, I wouldn’t start with a new tool or a big program. I’d start with a few things you can do this month:

Pull your five largest cost categories and put a name next to each one.

Ask what each one supports and whether the cost still matches the value.

Pick one unit cost metric that makes sense for your business and track it monthly alongside the total.

Switch off non-production environments outside working hours and clean up anything nobody claims.

Put cost on the agenda for your next architecture review and planning session.

None of this is glamorous, but it’s where most of the savings are. Cloud makes it very easy to create capacity, much faster than governance can keep up. That’s the price of speed and flexibility, and it’s usually worth paying. The trick is knowing what you’re paying for.