I paid my first real attention to cloud cost not in a strategy meeting or on a FinOps dashboard. It was that something that should have been simple had become expensive, and nobody saw it.
That’s how these things normally go. You don’t just wake up to a huge bill out of the blue. It grows out of a long series of reasonable decisions: keep the environment running a little longer, size the VM a little bigger just in case, leave staging up because someone may need it later, do not touch the database because the safer option is to leave it alone.
I have seen this pattern from both sides of the infrastructure work. The platform itself was rarely the expensive part on AWS, on Azure, during migrations, or keeping services alive for real usage. It was the routines around the platform. When we moved workloads from EC2 to Azure VM setups, the cost difference often wasn’t the cloud provider but how deliberately we managed the environment once it was up.
The tools are good, but they are not magic. Cost dashboards tell you what resources are oversized, idle, and wasting money in what environments. They can’t tell you why nobody took ownership of that resource in the first place. That’s the real problem.
Many teams treat cost optimization as a cleanup sprint. So, someone is charged with finding waste, they cut the obvious stuff, the bill comes down a bit and everyone moves on. That will work one time. Twice maybe. But if we keep making decisions about infrastructure the same way, we will get the same waste back.
The most important thing for me is that the cost has to show up early, not late. It should be in the conversation when something is being provisioned, not just when finance starts asking questions. Now is the time for a team creating a new environment, building a new service or choosing a larger instance size “just to be safe” to ask if the choice actually matches the workload.
I’ve also learned that ownership is more important than reporting. A cost report that doesn't go anywhere specific gets ignored. If each team is looking at their own spend, it’s much harder to ignore a simple weekly review, because now the number is owned by someone. The same applies to tagging. Tags will drift if they are only a reminder on a wiki page. They happen if Terraform or deployment code needs them.
That’s why cloud cost control is an engineering discipline more than a finance job. It’s not really about maximizing every dollar out of it. It’s about developing the habit of asking, before you deploy, whether this is the right shape of infrastructure for the problem you really have.
That’s the part most tools can’t do for you.
Top comments (0)