Cloud costs and FinOps: pricing models, tagging, budgets and rightsizing
In short: Cloud is billed by use, so cost is an engineering concern, not just a finance one. FinOps is the practice of making spend visible, cutting waste and buying wisely. Tags, budgets with alerts, rightsizing and commitment discounts are the main tools, and unexpected spend can also be a sign of compromise.
How cloud pricing works
You pay for what you use: compute per second, storage per gigabyte-month, requests per million, and data transfer per gigabyte. That flexibility is great for experiments, but it means a forgotten test server keeps charging until someone notices.
| Pricing model | How it works | Best for |
|---|---|---|
| On-demand | Full price, no commitment | Short or unpredictable workloads |
| Commitment (reserved instances, savings plans) | Discount for committing to 1 or 3 years of usage | Steady, always-on workloads |
| Spot / preemptible | Deep discount on spare capacity that can be reclaimed at short notice | Fault-tolerant batch jobs, test environments |
| Free tier | Limited free usage, often for 12 months or always-free amounts | Learning and labs |
The FinOps cycle
FinOps (cloud financial operations) brings engineers, finance and managers together to make cost-aware decisions. The FinOps Foundation describes it as a cycle of three phases.
- Inform: tag resources, split costs by team or project, and show each team its own spend.
- Optimise: remove waste, rightsize, buy commitments for steady load, and use cheaper storage tiers.
- Operate: set policies, budgets and reviews so good habits stick.
Tagging, budgets and alerts
Tags are key-value labels such as Owner=netops, Environment=prod and CostCentre=1234. Without them you cannot tell who owns a resource. Enforce required tags with policy so untagged resources are blocked or flagged.
Budgets send alerts when actual or forecast spend crosses a threshold. A budget with email alerts at 50, 80 and 100 percent takes minutes to set up:
az consumption budget create --budget-name lab-budget --amount 20 --category cost --time-grain monthly --start-date 2026-11-01 --end-date 2027-10-31Tip: a sudden spike in spend, such as GPU instances in a region you never use, can be the first sign of stolen credentials being used for crypto-mining. Treat cost alerts as security signals too.
Rightsizing and common surprises
- Find idle or underused resources using the provider's advisor tool or monitoring data.
- Rightsize: pick a smaller instance size if CPU and memory stay low for weeks.
- Schedule non-production systems to stop at night and weekends.
- Delete unattached disks, old snapshots and unused public IP addresses.
- Reduce data egress with caching, CDNs and private endpoints.
- Commit only to the steady baseline you are sure of.
Avoid cutting security to save money. Turning off logging or backups looks cheap until you need them. Tune log levels and retention instead.
Practice
Your web tier runs at a steady load all year. Which pricing model saves most?
Why are tags essential to FinOps?
A budget alert fires for unexpected compute in an unused region. What might it mean?
Key takeaways
- Cloud is pay-per-use, so engineers influence cost directly.
- FinOps loops through inform, optimise and operate.
- Tags and budgets give visibility and early warnings.
- Rightsize, schedule, clean up and commit to the steady baseline.
- Unexpected spend can be a sign of compromise.
Original content written by Mohamed Fayas for FM TecHub. © 2026 FM TecHub. You're welcome to share a link to this page; please don't copy the text elsewhere without permission.