FinOps · Cloud Cost Management
FinOps (Cloud Financial Operations)
What is FinOps?
- Putting the bill in front of the people who spend
- A repeating cycle of visibility, optimization, and operation
- No tag on the bill, no owner
What FinOps changes is who looks at the bill. In the on-premise era, buying a server fixed the cost once, and finance only had to track depreciation. In the cloud, every button a developer clicks incurs a cost. When finance pays at month-end but engineers decide spending every day, no one takes responsibility for the total. FinOps closes that gap: the people who spend see the cost, and finance understands the technical decisions.
In practice, three steps repeat. First is visibility: tag resources to break the bill down by team and service. Next is optimization: shut down test servers left on overnight, delete unused storage, and move resources with confirmed long-term use to committed-use discounts (reserved instances, savings plans). Last is operation: set per-team budget caps and unusual-spend alerts, and review the numbers in the same place every month. The common experience in the field is that a large share of cloud cost comes from resources that are running but that no one is using.
Training and HR teams are not exempt from FinOps. Once you move an LMS or HR system to the cloud, charges rise and fall with usage, and generative AI tools are billed per token. If you budget next year as a flat monthly figure, a year-end training rush or a surge in AI usage can double or triple the bill. It is safer to record usage assumptions instead of a flat fee, and to secure access to monthly usage reports in the contract.
The most common mistake in practice is treating FinOps as a 'cost-cutting project.' If you clean up servers once, cut 20%, and call it done, the same waste returns in the same places six months later. Conversely, if you push only for savings and block even the dev team’s experimental resources, delayed launches of new services cause a bigger loss. It is also common for so many resources to go untagged that half the bill sits under 'unclassified,' and in that state you can’t even tell who should cut. A bill that doesn’t name the responsible team means FinOps hasn’t started.
Seolyeong ParkGenerative AI and AX training expert with 17 years of hands-on experience and 200+ sessions, making AI usable right away even for non-specialists
Seungtae ParkPractical AX instructor who teaches by building AI business development and marketing automation hands-onThe concept emerged from practitioners in the late 2010s, as usage-based cloud billing became widespread and left a gap in cost accountability between technology and finance teams. The FinOps Foundation was founded in 2019 to standardize the terminology and practice framework, and it has since joined the Linux Foundation and serves as an industry standard.
Consider a hypothetical case. Company A, an online education platform with 120 employees, saw its monthly cloud bill double in a year, from KRW 14 million (≈ US$10,500) to KRW 28 million (≈ US$20,900). Revenue had grown only 30%, so the CEO put it plainly: 'If we can’t find where it’s leaking, there’s no investment approval next year.'
- When Manager Lee from finance and the development team lead opened the same screen for the first time, 46% of the bill was untagged, with no way to tell which team it belonged to. They agreed to tag every resource with a team label within a month.
- Once tagging was in place, they found 11 test servers created by a developer who had since left, running for ten months. The development team lead said, 'There is no excuse for this,' and shut them all down that evening.
- Eight production servers that had run for over a year were switched to committed-use discounts, saving KRW 3.1 million (≈ US$2,320) a month. Meanwhile, the auto-scaling budget for the year-end enrollment rush was actually increased.
- Once the company set monthly budget caps per team and an alert for spending 20% above the prior week, the education content team was flagged in its first month for video-conversion jobs and moved them to overnight hours.
- Three months later, the monthly bill had fallen from KRW 28 million (≈ US$20,900) to KRW 19 million (≈ US$14,200), and uncategorized items dropped from 46% to 3%. The CEO approved KRW 120 million (≈ US$89,800) for next year's AI features.