Five dollars for the map
Last weekend my family and I went to a pumpkin patch. Great corn maze, the kind that takes a solid forty minutes if you’re honest about it. Admission covered the maze. There was no map at the entrance. Why would there be? You’d just paid to get lost.
About twenty minutes in, at the point where every path looks like the one you just came down, there’s a kid at a folding table. Five dollars for a map. I started laughing, because I’d seen this business model before, in fact, I’d helped build it.
Think about what that five dollars is. The farm designed the maze. The farm chose how many dead ends it has. The farm knew exactly where people give up and then put the table there. You don’t buy the map when you’re deciding whether to go in. You buy it once you’re committed, disoriented, and the sun’s getting low. Nobody’s mad about it at a pumpkin patch. Being lost is kind of the point, the map is a joke you’re in on.
Now do the same thing with an enterprise cloud bill, and call the map FinOps.
I’ve been in pricing and packaging discussions at IBM, VMware, and Broadcom. In those discussions the phrase “let’s make it easier for the customer to understand what they’re paying for” almost never came up. The phrase “let’s make it easier for the customer to leave” NEVER came up. Not because the people in those rooms were villains. But because opacity is a growth strategy, and everyone’s bonus plan knew it.
So when the industry grew a whole profession along with a foundation, certifications, and a conference circuit to help customers decode their own bills, I recognized it immediately. FinOps is the customer paying five dollars for a map after the vendor built the maze. The difference is that the corn maze costs you an afternoon. The cloud maze costs you a department.
And this year, the map stopped working.
The reversal
Flexera’s 2026 State of the Cloud Report, its fifteenth edition, surveyed 753 cloud decision-makers. The headline finding: estimated wasted cloud spend rose to 29% of IaaS and PaaS budgets, up from 27%. It’s the first increase in five years. Every prior year of the trend went the right direction.
The rest of the report goes on to say that 63% of organizations have an established FinOps team. 71% run a Cloud Center of Excellence. 85% say managing cloud costs is their single top challenge even ahead of security. Cloud financial management has never been more widely staffed, more mature, or more executive-visible. And the waste number still went the wrong way.
The FinOps Foundation’s own State of FinOps 2026, drawn from 1,192 practitioners representing more than $83 billion in annual cloud spend, tells you where the fire started. Two years ago, only 31% of FinOps teams managed any AI spend. Today it’s 98%. FinOps has expanded to cover SaaS (90% of teams), software licensing (64%), private cloud (57%), and data center spend (48%). More than a quarter of teams now include labor costs as a responsibility. The Foundation formally changed its mission statement this year from managing the value of cloud to managing the value of technology.
Translation: the scope of things that need a cost specialist to interpret is growing faster than the specialists.
What broke it
AI, obviously. But we should be precise about why and how AI broke it.
Inference doesn’t fit a reservation model. It bills in tokens, requests, and GPU-seconds, in bursts nobody budgeted for, across shared models that multiple teams draw from with no clean way to attribute the cost. Unit prices for a given model fall. Total bills rise. The cost then compounds downstream through the CI/CD, testing, and deployment tooling that all now carries an AI meter of its own.
Every one of those properties is a variant of something the cloud already did to you. Per-second billing that looked cheap until you multiplied it. Egress you didn’t know you’d pay until you tried to leave. Reserved capacity that saved money right up until your architecture changed. AI didn’t invent the opacity it simply used it as a foundation and added new pricing units.
The FinOps practitioner quoted in the Foundation’s report said it better than I can: “dashboards are yesterday’s table stakes. Reactive. What’s needed is proactive, real-time, automated.” He’s right. He’s also describing a job the platform should be doing, not a job his team should be expanding into.
The uncomfortable part
Here’s what most people don’t want to say out loud about FinOps: it’s a tax.
Every dollar of FinOps tooling, every headcount on a cloud cost team, every hour an engineer spends tagging resources so the allocation report reconciles is money spent because the vendor’s bill couldn’t answer the question “what did I pay for?” on its own. It’s a growth industry because the complexity that feeds it is a growth strategy for the people selling the underlying compute.
I don’t say that to diminish the practitioners. The FinOps people I’ve met are some of the sharpest operators in the enterprise, and 78% of them now report into the CTO or CIO because their organizations finally understood the stakes. They’re doing essential work.
But essential work that exists only because a vendor made it necessary is still a cost the vendor imposed on you. The dirty secret isn’t that cloud is expensive. It’s that the expense of understanding cloud is a second product line, and you’re the one paying for it.
What “not needing FinOps” would actually require
I’m not going to pretend cost governance disappears. What I’ll argue is that it should be a property of the platform, not a department.
→ Pricing you can compute on a napkin. A handful of units — CPU, GPU, memory, storage — at a rate that doesn’t change based on region, commitment, or the phase of the moon. If you can’t estimate next month’s bill from this month’s workload without a spreadsheet, then pricing is the problem.
→ Spend limits the platform actually enforces. Not alerts. Not a Slack message after the fact. A boundary the infrastructure refuses to cross, set once, honored always.
→ Descaling that happens without a review cycle. Waste is idle capacity you’re still paying for. The platform should release it the moment demand drops, not the month the FinOps report flags it.
→ A bill that answers “what did I pay for?” in one page. Line items a CFO reads without a translator. Nothing to reconcile because nothing was obscured.
Notice what’s missing from that list: a tagging strategy, a cost-allocation tool, a center of excellence, a certification. That’s the point. When the platform is legible, the discipline built to decode illegibility has nothing to decode.
Why we built ContextOS this way
This is the position we took in the Cloud Pricing post, and the Flexera reversal is exactly why we took it.
ContextOS prices on a handful of units you already understand — CPU, GPU, memory, storage — at published rates, on a statement you can read without training. Spend limits are enforced by the platform, not suggested by it. When a Context’s load drops, the platform descales it and stops billing for what it released. The intelligence that a FinOps team would otherwise be hired to supply: what’s running, what it costs, whether it should still be running; is built into the layer that runs it.
We are not building a better FinOps dashboard. Plenty of good companies are, and the Flexera numbers suggest the market for them is only going to grow. We’re building the platform that makes the dashboard unnecessary, where “what am I paying for?” has an answer before you ask it.
Five years of falling waste just reversed with record FinOps investment behind it. That should tell you the map isn’t the problem. The maze is.
ContextOS is now available. Let’s talk
Sources
- Flexera, 2026 State of the Cloud Report, March 18, 2026 (n=753) — 29% waste (up from 27%), first increase in five years; 63% FinOps teams; 71% CCOE; 85% cost management top challenge; 76% of large enterprises >$5M/month; GenAI 58% (from 50%)
- FinOps Foundation, State of FinOps 2026 (n=1,192, $83B+ cloud spend) — 98% managing AI spend (from 31%); 90% SaaS / 64% licensing / 57% private cloud / 48% data center / 28% labor; 78% report to CTO/CIO; mission statement change; practitioner quote on dashboards