FinOps 2025: How Enterprises Cut $44.5B in Cloud Waste
Enterprises are cutting an estimated $44.5 billion in cloud waste by starting with visibility, automating the easy wins, and then committing. That means rightsizing overprovisioned compute, shutting down idle resources, using reserved instances and savings plans, and giving engineers the real-time cost data fewer than half of teams can access.
Enterprise cloud bills keep climbing, and a growing share of that spend buys nothing at all. Harness's FinOps in Focus 2025 report estimates that 21% of enterprise cloud infrastructure spend, about $44.5 billion this year, is wasted on underutilized resources. The study, a survey of 700 developers and engineering leaders across the US and UK conducted by Coleman Parkes in late 2024, pins the blame on a familiar culprit: the disconnect between FinOps teams and the developers who provision infrastructure. At the same time, the FinOps Foundation's 2025 State of FinOps survey of 861 practitioners representing roughly $69 billion in public cloud spend shows the discipline expanding fast into AI and SaaS budgets.
How much cloud spend is actually wasted?
About one dollar in five. Harness estimates that 21% of enterprise cloud infrastructure spend is wasted on underutilized resources, equivalent to $44.5 billion in 2025, calculated as 21% of an estimated $211.9 billion in cloud system infrastructure spend, using Gartner's worldwide public cloud end-user spending forecast for 2025, according to the report.
Most of it comes down to visibility. Fewer than half of respondents say they have access to real-time data on idle cloud resources (43%), unused or orphaned resources (39%), or over- and under-provisioned workloads (33%). Without a clear picture of requirements, 55% of developers say purchasing commitments are ultimately based on guesswork.
"Cloud infrastructure spend is one of the biggest line items for modern enterprises, right behind salary. Leadership teams should ask themselves if they are comfortable relying on guesswork to manage and optimize this spend," said John Bonney, CFO at Harness, in the report announcement.
Why is cloud waste climbing again?
AI workloads added a volatile new line item before most teams built governance for it, while the old sources of waste kept flowing. Enterprises take an average of 31 days to identify and eliminate cloud waste such as idle, orphaned, or unused resources, and about 25 days to detect and rightsize overprovisioned ones, per Harness's 2025 survey. That lag is expensive when GPU instances, inference API calls, and AI-adjacent SaaS subscriptions are provisioned at speed.
The FinOps Foundation's 2025 data confirms the AI pressure: AI spending is now managed by 63% of respondents, up from 31% the year before, and 97% are investing in multiple infrastructure areas for AI. Planned AI investment skews heavily toward public cloud, with 69% going to SaaS and 30% to data center or private cloud. Practitioners rank AI optimization itself as a much lower priority for now. The immediate need is simply understanding what the spend is.
Developers are optimistic that automation closes the gap: 86% of developers believe AI will enhance their ability to optimize costs within the next year.
What does the working FinOps playbook look like?
Start with visibility, automate the easy wins, then commit. The data shows most teams skip the basics: 71% of developers do not carry out spot orchestration, 61% do not rightsize instances, 58% do not use reserved instances or savings plans, and 48% do not track and shut down idle resources, according to Harness. Only 32% have fully automated practices to enforce cost savings.
| Waste source | Fix | What the 2025 data shows | Effort |
|---|---|---|---|
| Overprovisioned compute | Rightsize instances and containers | 61% of developers don't rightsize; ~25 days on average to detect and fix overprovisioning | Low |
| Idle and orphaned resources | Auto-shutdown schedules for non-production | 48% don't track and shut down idle resources; 31 days on average to eliminate waste | Low |
| On-demand list pricing | Reserved instances, savings plans, spot orchestration | 58% don't use RIs or savings plans; 71% skip spot orchestration | Medium |
| Blind purchasing decisions | Real-time cost visibility, shift-left culture | 55% say commitments are guesswork; only 43% see idle-resource data in real time | Medium |
The cultural fix matters as much as the tooling. "The reality is developers do not often view cost optimization as a priority. This disconnect leads to overprovisioned resources, idle instances, and inefficient architectures that quietly drain budgets," explained Ravi Yadalam, Senior Director of Product Management at Harness. His prescription: shift FinOps left so engineers see how architectural choices affect financial outcomes before deployment, a view 62% of developers say they want, asking for more control over infrastructure costs.
How are FinOps teams handling the AI spending surge?
They are building the visibility layer first and deferring optimization. The FinOps Foundation reports that the capabilities practitioners apply to AI spend sit in the "understand cost" and "quantify value" domains: allocation, data ingestion, reporting, anomaly detection, and planning and forecasting. Optimization activities rank much lower. That order makes sense, since you cannot optimize spend you cannot attribute.
Scale explains the urgency. In the Foundation's 2025 survey, 31% of respondents work at organizations spending more than $50 million a year on public cloud, 20% at organizations spending more than $100 million, and over 20 at organizations spending more than $1 billion a year. At that scale, a 21% waste rate is a nine-figure problem, which is why workload optimization and waste reduction remained a top priority for 50% of practitioner respondents, ahead of full allocation of cloud spending and accurate forecasting.
Where is FinOps headed next?
Beyond public cloud, into what the Foundation calls "Cloud+." A combined 65% of practitioners now manage SaaS spend, with a 25% predicted increase in the next 12 months, while licensing, private cloud, and data centers follow, according to the 2025 State of FinOps. For the next 12 months, implementing governance and policy at scale becomes the top future priority, overtaking workload optimization. The biggest priority climbers are managing AI/ML spend, managing costs beyond public cloud, and getting to unit economics.
Standardization is catching up too: 57% of practitioners plan to use the open-source FOCUS billing specification in the next 12 months. And the practice keeps professionalizing. Investment and tooling saw a 20% jump as the thing practitioners say they most need to hit their priorities, even as the respondent base keeps growing: 861 practitioners this year, up year over year.
Cloud waste is an information problem. Teams that give engineers real-time cost data, automate the shutdown of what nobody uses, and extend FinOps discipline to SaaS and AI spend are the ones turning a $44.5 billion industry leak into a durable advantage.
Related reading: Cloud Security Posture Management: Stop Misconfigurations · SaaS Backup: Protecting the Data Your Cloud Apps Hold
FAQ
How much of enterprise cloud spend is wasted?
Harness estimates that 21% of enterprise cloud infrastructure spend — about $44.5 billion in 2025 — is wasted on underutilized resources, calculated as 21% of an estimated $211.9 billion in cloud system infrastructure spend.
What causes most cloud waste?
A lack of visibility. Fewer than half of respondents have real-time data on idle cloud resources (43%), unused or orphaned resources (39%), or over- and under-provisioned workloads (33%), and 55% of developers say purchasing commitments are ultimately based on guesswork.
Why is AI making cloud waste worse?
AI workloads added a volatile new line item before most teams built governance for it. AI spending is now managed by 63% of respondents, up from 31% the year before, while GPU instances, inference API calls, and AI-adjacent SaaS subscriptions get provisioned at speed.
What does the working FinOps playbook look like?
Start with visibility, automate the easy wins, then commit: rightsize instances and containers, set auto-shutdown schedules for non-production resources, and use reserved instances, savings plans, and spot orchestration. Most teams still skip the basics — 71% don't do spot orchestration and 61% don't rightsize.
Where is FinOps headed next?
Beyond public cloud into "Cloud+": 65% of practitioners now manage SaaS spend, and implementing governance and policy at scale becomes the top future priority for the next 12 months, while 57% plan to use the open-source FOCUS billing specification.

Comments
Post a Comment