There's a number that comes up in every cloud project. Usually too late. Usually at the CFO's desk. And usually higher than expected.
The cloud bill.
I've worked with Siemens, Siemens Energy, and Volkswagen over six years. Each of these organizations eventually asked the same question: Why is the cloud more expensive than promised? The answer is almost always the same. And it rarely has anything to do with AWS pricing.
The Misconception
Most cloud migrations start with a promise: Cloud is cheaper than on-premise. You only pay for what you use. Pay-as-you-go. Elastic. Efficient.
That's true. In theory.
In practice, here's what happens: A company migrates its workloads to the cloud. The first few months look fine. Then costs grow. Slowly at first, then exponentially. After a year,
the CFO asks why the cloud bill is twice as high as planned.
The answer: Because nobody used the cloud any differently than the data center.
Same architecture, same patterns, different bill.
The Three Cost Drivers Nobody Talks About
1. Architecture Debt
The biggest cost factor in the cloud isn't an AWS fee. It's the architecture.
If you lift a monolithic application into the cloud, you pay cloud prices for an on-premise architecture. The result: cloud pricing for an architecture that wasn't built for the cloud.
At Siemens Energy, we replaced an SAP-based HR data platform with a cloud-native solution. The old solution cost over 1,000,000 EUR per year in licensing and operating costs. The new solution: under 40,000 EUR. Not because AWS is cheaper than SAP. Because the architecture is tailored to the cloud. Serverless, event-driven, no over-provisioning.
800,000 EUR in savings per year. Not through a better contract with AWS.
Through better architecture.
2. Missing Governance
In an enterprise with 1,500 AWS accounts, here's what happens: Teams provision resources for a project. The project ends. The resources keep running. Nobody shuts them down because nobody knows whether they're still needed.
At Siemens, we implemented cross-account governance. Automated checks that identify unused resources. Policies that prevent teams from creating resources without tagging. Budget alerting per account, not per organization.
This isn't rocket science. It's discipline codified in code. But it requires someone who builds this discipline in from the start. Not as a FinOps afterthought after the first cost shock.
3. The Fear of Serverless
Enterprise clients have a complicated relationship with serverless. They understand the concept. They see the cost advantages. And then they choose containers because they feel more familiar.
Containers aren't wrong. But they're more expensive than serverless when the workload fits. A Lambda function that runs five times a day costs almost nothing. An ECS container running 24/7 that processes a request five times a day costs 90 EUR per month. Multiply that by 200 services, and the comfort argument becomes a six-figure cost factor.
For the HR Data Hub, we went all-in on serverless. Lambda, Step Functions, S3, DynamoDB. No permanently running instances for workloads that don't need them. The result: operating costs under 40,000 EUR per year for a platform that processes data from 150+ countries.
What We Recommend to Our Clients
First: Architecture before migration. Before you invest a single euro in the cloud, invest in the right architecture. Lift-and-shift is the most expensive path to the cloud.
Second: Governance from day one. Not as an afterthought. Tagging policies, budget alerts, automated resource cleanup. Before the first workload is deployed.
Third: Be honest about serverless. Not every workload is suited for Lambda. But more are than most enterprise architects assume. The cost savings are real and significant.
Fourth: Cost as an architecture metric. In every Architecture Decision Record, we document the cost implications. Not as a footnote, but as a decision criterion. The cheapest solution isn't always the best. But ignoring costs is always the worst.
The Key Insight
The cloud isn't automatically cheaper. It can be. Dramatically. 800,000 EUR per year dramatically. But only if the architecture is right, governance is in place, and teams think cloud-native, not cloud-migrated.
This requires upfront investment. In architecture, in knowledge, in making the right decisions from the start. If you cut corners here, you pay later. Not to AWS. To yourself.
