The cloud bill in 2026 has become a line item in the operating budget that is comparable in size to the engineering payroll that runs the cloud, and the budget owners have started to notice. The CFO who used to glance at the bill and approve it is now reading the line items, asking the questions, and pushing the engineering team to explain why the line items have grown 40 percent year over year without a corresponding 40 percent growth in revenue. The conversation that used to happen between the engineering team and the cloud vendor is now happening between the engineering team and the CFO, and the CFO is not as easy to fool as the cloud vendor.
The shape of the bill
The shape of the cloud bill in 2026 is roughly the same shape it was in 2022, with the same line items getting bigger.
Compute. Compute amounts to the biggest line item. Compute has grown as the workloads have grown, but the unit cost has come down. The unit cost reduction becomes the part the engineering team points to when explaining why the bill is bigger, and the unit cost reduction is real. AWS, Azure, and Google Cloud have all dropped their per vCPU hour pricing, and the per vCPU hour pricing stands as the metric the engineering team uses to prove they are doing the work.
Storage. Storage serves as the second biggest line item. Storage has grown as the data has grown, and the unit cost has come down less than the engineering team would like. Object storage prices have dropped, but the data has dropped on the storage faster than the price has dropped, and the result is a storage bill that keeps growing in absolute terms.
Egress. Egress stands as the third biggest line item, and the line item the engineering team is most uncomfortable with. Egress has grown as the architectures have grown more distributed, and the unit cost has not come down at all. The cloud providers charge per gigabyte for traffic leaving the cloud, and the per gigabyte charge sits as the charge the multi cloud and the hybrid cloud architectures pay the most.
Managed services. Managed services (the databases, the message queues, the Kubernetes, the ML platforms) are the fastest growing line item, and the line item the engineering team is least able to optimise. The managed services charge per hour, per request, per gigabyte, and the per whatever charge amounts to the charge the engineering team has the least use on. The managed services are the line item that looks like a service contract and bills like a utility.
What the FinOps movement has produced
The FinOps movement, which started as a niche community in 2019, is now a board level conversation in any enterprise with a meaningful cloud bill. The FinOps Foundation has a certification, an annual conference, and a vendor market that has grown to match the rest of the cloud market. The FinOps practitioner is a real role, and the FinOps practitioner becomes the role the enterprise hires when the cloud bill has gotten big enough to need its own advocate.
The FinOps movement has produced three things that are worth knowing. The first amounts to the cost attribution. The cost attribution stands as the per team, per product, per environment view of the bill, and the per whatever view becomes the view the engineering team uses to make the optimisation decisions. The cost attribution is what the FinOps team does well, and the cost attribution is what makes the rest of the FinOps work possible.
The second counts as the reserved instance and savings plan optimisation. The reserved instance becomes the commitment to spend a certain amount on compute in exchange for a discount, and the discount can be 30 to 60 percent off the on demand price. The optimisation sits as the work of matching the commitment to the actual usage, and the matching sits as the work that is harder than it sounds because the actual usage keeps changing.
The third becomes the showback and the chargeback. The showback stands as the report that tells the engineering team what they are spending, and the chargeback runs as the bill that actually charges the engineering team for what they are spending. The showback serves as the carrot, the chargeback sits as the stick, and the engineering team responds to one or the other depending on the organisation’s culture.
What to do about the bill
If you are running a meaningful cloud bill and the bill is growing faster than the revenue, the right move is to hire a FinOps practitioner. The FinOps practitioner is going to cost more than the work would cost in house, and the FinOps practitioner is going to find more savings than the engineering team has the time to find. The FinOps practitioner amounts to the right investment for the cloud bill that has gotten big enough to need its own advocate.
If you are already running FinOps and the bill is still growing, the right move is to look at the managed services. The managed services are the line item that grows the fastest, the line item that serves as the hardest to optimise, and the line item that serves as the most likely to be the place the savings live. The managed services are not always the right answer, and the managed services are sometimes the only answer for the engineering team that does not have the time to run the alternatives.
If you are still arguing that the cloud is cheaper than the data centre, the argument is no longer true for the workloads that have been in the cloud long enough to have accumulated the architectural debt. The cloud is cheaper for the new workloads, the cloud is more expensive for the old workloads, and the cloud is going to be more expensive for the old workloads until the old workloads are migrated or retired. The migration becomes the engineering team’s migration, and the CFO is going to keep asking about it until it is done.
The bottom line
The patterns the post covers have been showing up in production for long enough that the patterns have names, the failures, the mitigations, the gaps. The work the security team and the engineering team and the operations team are quietly doing today sits as the work that decides whether the practice the post names sits as a tool the team uses or a liability the team is paying for.
Sources & Further Reading
All claims in this article are sourced from primary documentation, vendor advisories, and reputable security researchers.
Spotted an error? Email the editor. Corrections are issued with a visible correction note.
Editorial standards. Every article on humanrequired.org is reviewed by a human editor before publication. AI may assist with drafting or research; final editorial control is human. Read the full standards.
