Cloud waste climbed to **29% of estimated cloud spend is wasted** in Flexera’s 2026 State of the Cloud Report, snapping a five-year run of steady declines. The report, published in March and still shaping enterprise budget conversations four months later, landed as generative AI moved from pilot projects into production workloads at most large organizations.
Flexera polled 753 cloud decision-makers worldwide and found that 17% of organizations exceeded their public cloud budgets over the past year. Seventy-six percent of large enterprises now spend more than $5 million a month on public cloud, and 58% run generative AI as a cloud service, up from 50% a year earlier. Those numbers turn a familiar FinOps story into a sharper one: spending discipline that had been improving for years just went into reverse.
The timing matters for what comes next. Microsoft reports fiscal Q4 earnings on July 29, and Azure’s growth rate will be one of the clearest signals yet of whether AI infrastructure spending is converting into durable revenue or piling up as the kind of waste Flexera just measured. The last earnings round, covered in our look at Google Cloud’s growth ahead of AWS earnings, already showed all three hyperscalers beating estimates on AI demand. This report is the first hard data suggesting that demand comes with a cost side nobody has fully controlled.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Cloud Waste Hits 29% in Flexera’s 2026 State of the Cloud Report
Flexera has published its State of the Cloud Report for 15 years, and this edition reads like a course correction. Wasted cloud spend rose to 29%, up two points from 27% in 2025, which Flexera describes as the first increase in five years. Waste sat at 28% in 2023, dropped to 27% in 2024, and held there through 2025, before this year’s reversal broke the pattern.
The overspend figure is just as sharp. Organizations exceeded their public cloud budgets by **Organizations are exceeding budgets by 17%**, a gap that adds up fast at scale. A company spending the enterprise-average $5 million a month on public cloud would blow past its annual plan by roughly $10 million if that overshoot held for a full year.
Flexera frames the reversal as a straightforward story: AI adoption is outrunning the cost-governance habits that took years to build. A year earlier, in the 2025 edition of the same report, 84% of respondents already called managing cloud spend their top challenge. Spending discipline was fraying before the waste numbers caught up to the sentiment, a pattern one industry analysis of the 29% figure describes as FinOps still not working three years into the discipline’s mainstream adoption.
Why Cloud Waste Is Climbing Again After Years of Decline
Three trends are converging to push the waste number back up, based on the patterns in Flexera’s data. First, AI services are landing on cloud bills faster than finance teams can tag and attribute them. Fifty-eight percent of organizations now run generative AI as a public cloud service, up from 50% the year before, and 31% describe themselves as still experimenting with GenAI rather than running it in steady state. Experimentation is expensive by nature. Teams spin up GPU instances, test models, and abandon half of what they provision before it ever reaches production.
Second, the metrics enterprises use to judge cloud success are shifting away from pure cost efficiency. Sixty-four percent of organizations now say they measure cloud success by the value it delivers to the business, up 12 points year over year, while cost efficiency as a top metric dropped 6 points. That shift is healthy over the long run, but in the short term it can mean fewer people are watching the bill line by line.
Third, unit economics adoption is still catching up to the workloads it needs to track. Forty-nine percent of organizations now measure cost per service or transaction, up from 40%, which still leaves just over half without a clean way to tie AI spend to a specific product outcome. Without that link, waste stays invisible until the invoice arrives.
Inside the Survey: What 753 Cloud Decision-Makers Told Flexera
Flexera’s 753 respondents spanned company sizes and regions, and the report’s broader findings help explain why waste crept back up even as cloud maturity, on paper, kept improving. Seventy-six percent of large enterprises now spend more than $5 million a month on public cloud, so even small percentage swings in efficiency translate into real money. Seventy-three percent use hybrid cloud, up from the prior year, and **73% use hybrid cloud**, both signs of increasingly complex environments that are harder to monitor end to end.
Sustainability tracking shows a similar governance gap by region. Forty-seven percent of European organizations have defined programs to track cloud-related carbon footprint, compared with 34% in North America, a split that tracks the regulatory pressure European companies face and that U.S. teams have been slower to match.
| Metric (Flexera 2026 State of the Cloud Report) | 2026 Value | Trend vs Prior Year |
|---|---|---|
| Wasted public cloud spend | 29% | Up from 27% (first rise in 5 years) |
| Average public cloud budget overspend | 17% | Newly highlighted metric |
| Large enterprises spending over $5M/month | 76% | Up |
| Measure cloud success by business value delivered | 64% | Up 12 points |
| Use unit economics to measure cost per service | 49% | Up from 40% |
| Use GenAI as a public cloud service | 58% | Up from 50% |
| Rely on a dedicated FinOps team | 63% | Reported for the first time at this level |
| Have a Cloud Center of Excellence (CCOE) | 71% | Up |
| Use hybrid cloud | 73% | Up |
The picture Flexera paints is an industry that built the org charts and dashboards for disciplined cloud spending, then handed those same teams a new, harder-to-predict category of cost before they had fully absorbed the last one.
The Five-Year Trend: Cloud Waste From 2023 to 2026
Reading the trend year by year makes the 2026 reversal easier to interpret. Waste fell from 28% in 2023 to 27% in 2024, then held flat at 27% in 2025, matching the period when FinOps Foundation membership and dedicated FinOps teams were both climbing fast. That two-year plateau looked, at the time, like early proof that FinOps practices were working.
| Year | Cloud Waste / Recoverable Spend | Source |
|---|---|---|
| 2023 | 28% | Flexera State of the Cloud, 12th annual edition |
| 2024 | 27% | Flexera State of the Cloud, 13th annual edition |
| 2025 | 27% | Flexera State of the Cloud, 14th annual edition |
| 2026 | 29% | Flexera State of the Cloud, 15th annual edition |
| Best-practice benchmark | 20%-30% recoverable with rigorous FinOps | McKinsey estimate, cited in industry FinOps coverage |
That McKinsey range gives the 2026 number useful framing. At 29% waste, most enterprises are sitting near the bottom of the recoverable range Flexera and outside analysts describe, not below it, which means the tools and habits to close the gap already exist. The open question is whether AI spending is growing faster than FinOps teams can extend their existing playbook to cover it.
AI Workloads Are Rewriting the Cloud Cost Equation
Every category of cloud spend Flexera tracks looks different once AI enters the picture, and GPU-backed workloads are the clearest example of why cost management built for a pre-AI cloud stack is struggling to keep up.
GenAI Adoption Jumps to 58% of Cloud Users
Generative AI has gone from an emerging line item to a mainstream one inside a single survey cycle. Flexera found 58% of organizations now consume GenAI as a public cloud service, up from 50%, while 31% are still in the experimentation phase and 29% are running traditional machine learning workloads. Security and compliance ranked as the top concern tied to AI initiatives, ahead of cost, suggesting many organizations are still prioritizing governance and risk over spend control when they design AI deployments.
Why AI Inference Costs Are Harder to Tag and Forecast
Traditional cloud cost tools were built around relatively predictable units: virtual machines, storage buckets, database instances. GPU-backed AI inference doesn’t fit that mold as neatly. Usage can spike with a single feature launch, a model can run cheaply for months and then need retraining on more expensive hardware, and multiple teams often share the same underlying cluster without clean cost boundaries between them. A basic tagging policy, like the query below, is the starting point most FinOps teams use to attribute spend to a team and workload before trying to optimize it.
aws ce get-cost-and-usage
--time-period Start=2026-06-01,End=2026-07-01
--granularity MONTHLY
--metrics "UnblendedCost"
--group-by Type=TAG,Key=team Type=TAG,Key=workloadThat kind of query only works if resources were tagged correctly at creation, which is precisely the step most AI experimentation skips under deadline pressure. It’s a small technical gap with an outsized effect on the 29% waste figure.
FinOps Goes Mainstream: 63% of Enterprises Now Run a Dedicated Team
FinOps stopped being a niche discipline some time ago, and Flexera’s 2026 numbers confirm it. Sixty-three percent of organizations now rely on a dedicated FinOps team, a level of institutional buy-in that would have been unusual five years ago when the practice was still mostly associated with a handful of cloud-native companies. The FinOps Foundation describes the discipline’s scope expanding well past its original mandate. As the Foundation puts it in its State of FinOps 2026 materials, “FinOps is no longer defined by cloud cost management alone, it’s become the method for identifying and communicating technology value across AI, SaaS, licensing, private cloud, and data center throughout the organization where needed,” according to the FinOps Foundation.
That broader mandate lines up with Flexera’s finding that 64% of organizations now judge cloud success by business value rather than raw cost efficiency. It also helps explain why a rising waste percentage hasn’t triggered the kind of budget panic it might have five years ago. Teams increasingly treat some inefficiency as an acceptable cost of moving fast on AI, as long as the resulting product delivers value.
That tolerance has limits, though. AgamiSoft, a cloud cost management publication, frames the core FinOps mandate plainly: “FinOps is the organizational practice of managing cloud spending as a shared financial responsibility between engineering, finance, and business teams,” AgamiSoft notes in its 2026 FinOps overview. When 29% of spend is wasted, shared responsibility starts to mean shared blame, and that’s typically when a CFO asks for a formal FinOps function if one doesn’t already exist.
AWS vs Azure vs Google Cloud: Native Cost Tools Compared
All three major hyperscalers ship cost management tools at no extra charge, which raises an obvious question: if the tools are free and available, why is waste still climbing? AWS offers Cost Explorer for usage analysis alongside Cost Optimization Hub, which centralizes savings recommendations across linked accounts. Microsoft bundles Azure Cost Management + Billing directly into the platform, covering cost analysis, budgets, alerts, and recommendations. Google Cloud pairs its billing reports with Recommender, which surfaces rightsizing and other savings suggestions automatically.
| Capability | AWS | Microsoft Azure | Google Cloud |
|---|---|---|---|
| Primary tool | Cost Explorer + Cost Optimization Hub | Cost Management + Billing | Cost Management / Recommender |
| Cost visibility and reporting | Yes | Yes | Yes |
| Budgets and alerts | Yes | Yes | Yes |
| Automated optimization recommendations | Yes, via Cost Optimization Hub | Yes | Yes, via Recommender |
| Native multi-cloud normalization | No, AWS spend only | No, Azure spend only | No, Google Cloud spend only |
| Additional license fee | None | None | None |
The practical limitation, based on how Flexera’s respondents describe their environments, isn’t a missing feature so much as scope. Native tools are built to explain one provider’s bill. Seventy-three percent of organizations now run hybrid cloud, and a meaningful share run multiple public clouds alongside it, so the single-provider view from any one hyperscaler’s dashboard misses a large part of the total cost picture by design. Our earlier comparison of Bedrock, Azure AI Foundry, and Vertex AI pricing runs into the same problem: three genuinely different pricing models that no single native dashboard reconciles automatically.
Third-Party FinOps Platforms Fill the Gaps
That single-provider blind spot is the business case for the third-party FinOps market. A 2026 review of FinOps tooling from Platform Engineering groups the field into enterprise-scale platforms like IBM’s Apptio Cloudability, Flexera One, and Broadcom’s CloudHealth, developer-focused tools like Datadog Cloud Cost Management, Cast AI, and CloudZero, and newer, more specialized entrants like Finout, ServiceNow Cloud Cost Management, Antimetal, and Pump.co.
Kubernetes cost visibility is its own sub-category worth calling out, since containerized workloads split cost across shared clusters in ways that are notoriously hard to attribute. We covered the setup process in our guide to setting up Kubecost, one of the more widely deployed open-source options for that specific problem, and its adoption has grown alongside Kubernetes becoming a bigger share of enterprise infrastructure spend.
Platform Engineering’s guidance for teams evaluating this market draws a clear line. “Purchase solutions for Reporting/Recommendations (first 2 Rs), these are mature commodities,” the publication advises, while cautioning that “the first 3 Rs provide no lasting impact without Retain (4th R), achieved through shared responsibility frameworks across Development, Finance, Product, Security, and Governance teams,” according to Platform Engineering’s FinOps tools evaluation. In other words, buying a tool solves the visibility problem. It doesn’t solve the organizational problem Flexera’s survey keeps surfacing.
Market Impact: What Rising Cloud Waste Means for Vendors and Buyers
For the hyperscalers, rising waste is a double-edged signal. In the short term, waste is still revenue. An idle GPU instance or an oversized database still shows up as a paid line item on AWS, Azure, or Google Cloud’s books, regardless of whether the customer is getting value from it. That dynamic is part of why cost optimization has historically shipped as an opt-in feature rather than a default setting.
Longer term, the incentive flips. Enterprises that feel burned by a 29% waste rate and a 17% budget overshoot tend to respond in one of two ways. They either centralize purchasing power with a single preferred vendor to negotiate better committed-use discounts, or they diversify further to avoid concentration risk, which is part of why 73% of organizations already run hybrid cloud. Either response reshapes the sales motion hyperscalers depend on for predictable growth.
For the FinOps tooling market, the numbers are a tailwind. A field that already includes large platform plays like Flexera One, IBM’s Apptio suite, and Broadcom’s CloudHealth, alongside faster-moving challengers like CloudZero, Vantage, and Cast AI, now has fresh, well-publicized evidence that the problem it addresses is getting worse rather than better. Cloud trend coverage from InformationWeek’s 2026 outlook points to the same pressure point, flagging network and interconnect optimization as a growing enterprise priority alongside raw compute cost.
For enterprise buyers, the practical impact is more immediate. Budget cycles built around 2025 assumptions are already out of date. A finance team that modeled cloud growth on last year’s numbers, without a specific line for AI compute, is the team most likely explaining a variance to its board this quarter.
Historical Context: From Lift-and-Shift to the FinOps Era
Cloud cost management has moved through roughly three phases since public cloud went mainstream in the early 2010s. The first was pure migration: enterprises moved workloads off on-premises hardware and treated the resulting cloud bill as an acceptable cost of agility, with little formal oversight. The second, running roughly from 2018 through 2023, was the rise of FinOps as a named discipline, driven by the FinOps Foundation’s founding and the spread of showback and chargeback models that made individual teams accountable for what they spent.
The third phase, which Flexera’s 2026 report effectively documents the start of, is what the report itself calls the “value era,” where the question shifts from how much a team spent to what that spending produced. That framing is more mature, but it can also mask waste in the short term, since a team delivering clear business value has more room to justify an inefficient implementation than a team that can only point to a cost-efficiency metric.
Cloud trend analysis from ShapeBlue’s look at the trends defining 2026 describes this same convergence of cloud strategy and business value as one of the year’s defining shifts, not an isolated Flexera finding. Twenty-nine percent waste, in that light, isn’t a failure of FinOps as a discipline. It’s a sign the discipline’s boundaries expanded faster than its tooling did.
The Competitive Landscape: Multi-Cloud, Hybrid, and CCOE Adoption
The complexity driving waste up is structural, not just a matter of discipline. Seventy-three percent of organizations now run hybrid cloud, up from the prior year, and 71% have established a Cloud Center of Excellence or an equivalent governance function. Both numbers describe an environment that looks more mature on paper, with dedicated structures and cross-functional ownership. Yet that same complexity is exactly what makes end-to-end cost visibility harder to maintain.
A Kubernetes cluster running autoscaled workloads across multiple node pools can shift cost dramatically based on scheduling decisions alone. Tools built for that specific problem, compared directly in our Karpenter vs Cluster Autoscaler vs KEDA benchmark, exist precisely because underlying infrastructure choices determine a meaningful share of the final bill before any FinOps team even reviews it. Messaging and streaming infrastructure carries a similar hidden cost profile. The choice between a managed queue and a self-run streaming platform, the kind of trade-off we measured in SQS vs Kafka, can swing both raw cost and the operational overhead needed to monitor it.
None of this is unique to any single provider. It’s the byproduct of enterprises spreading workloads across more environments, more services, and more architectural patterns than the previous generation of cost tools was designed to track in one view.
5 Predictions for Cloud Cost Management in 2027
Based on the trajectory in Flexera’s data and the broader market signals from this year, expect the following over the next 12 months:
- Cloud waste stays elevated before it improves. AI workload provisioning is unlikely to slow down faster than FinOps teams can build tagging and attribution practices around it, so the 2027 report will likely show waste holding near or above 29% rather than snapping back toward 27%.
- Native hyperscaler tools add AI-specific cost attribution. With 58% of organizations running GenAI as a cloud service, AWS, Microsoft, and Google Cloud all have a clear incentive to build GPU- and model-level cost breakdowns directly into Cost Explorer, Azure Cost Management, and Recommender rather than ceding that ground entirely to third-party platforms.
- Unit economics becomes a board-level metric. The jump from 40% to 49% adoption in a single year suggests cost-per-transaction and cost-per-feature reporting moves from a FinOps team’s internal dashboard to something CFOs ask about directly, especially for AI-powered product features.
- Third-party FinOps platforms consolidate. A crowded field of enterprise platforms, developer-focused tools, and newer specialized entrants is a classic setup for acquisitions, particularly as larger platform players look to add AI-specific cost intelligence rather than build it from scratch.
- FinOps teams expand their mandate beyond public cloud. Following the FinOps Foundation’s own framing, expect more organizations to formally extend FinOps practices to SaaS licensing and private infrastructure spend, not just public cloud, as the next reported milestone.
How Enterprises Can Cut Cloud Waste Now
The Flexera data points to a fairly consistent playbook, even if execution is where most organizations are still catching up. Tagging and attribution come first. A resource that isn’t tagged to a team, project, or workload is effectively invisible to any cost tool, native or third-party. Unit economics comes second, since the 49% of organizations already measuring cost per service have a much clearer basis for deciding whether a given AI feature is worth its infrastructure bill than the 51% still working from aggregate cost alone.
Governance structures matter more than any single tool choice. The 71% of organizations with a Cloud Center of Excellence and the 63% with a dedicated FinOps team are best positioned to catch waste before it compounds across a full budget cycle, since both functions exist specifically to give someone organizational authority over the numbers.
Matching the tool to the problem beats buying one platform and hoping it covers everything. Kubernetes cost visibility, multi-cloud billing normalization, and AI-specific spend attribution are three different problems with different mature tools attached to each. Treating reporting and recommendations as a commodity purchase while building retention practices in-house, per Platform Engineering’s guidance above, is a reasonable template for deciding where to buy versus where to build.
Related Coverage
- How to Set Up Kubecost: 12 Steps, 60 Min [2026]
- SQS vs Kafka 2026: 1M vs 700K msg/sec, $460 Floor
- Google Cloud Hits 82% Growth Ahead of AWS Earnings [2026]
- Bedrock vs Azure AI Foundry vs Vertex AI: 17x Gap [2026]
- Karpenter vs Cluster Autoscaler vs KEDA: 3x Faster [2026]
- What Is Amazon RDS?
Frequently Asked Questions
What is FinOps?
FinOps is the operating model that treats cloud spending as a shared responsibility between engineering, finance, and business teams, rather than a cost finance manages alone after the fact. The FinOps Foundation describes it as extending beyond public cloud alone into AI, SaaS, and private infrastructure spend.
What is cloud waste and how is it measured?
Cloud waste refers to money spent on cloud resources that deliver no corresponding business value, such as idle instances, oversized databases, or abandoned test environments. Flexera measures it through an annual survey in which cloud decision-makers estimate the share of their own cloud spend that falls into that category.
Why did cloud waste rise in 2026 after five years of decline?
Flexera attributes the reversal largely to AI adoption outpacing the tagging, attribution, and governance practices FinOps teams had built for more predictable cloud workloads like virtual machines and storage.
What is the Flexera State of the Cloud Report?
It’s an annual survey, now in its 15th year, that polls hundreds of cloud decision-makers (753 in the 2026 edition) about cloud spending, adoption trends, and cost management practices across AWS, Azure, and Google Cloud.
How many organizations have a dedicated FinOps team in 2026?
Sixty-three percent, according to Flexera’s 2026 survey, alongside 71% that have established a Cloud Center of Excellence or equivalent governance function.
Are AWS Cost Explorer, Azure Cost Management, and Google Cloud’s Recommender enough on their own?
They cover single-provider visibility well and carry no extra license fee, but none of them natively normalizes spend across multiple providers, which is why third-party FinOps platforms remain common at organizations running hybrid or multi-cloud environments.
What is unit economics in a FinOps context?
It means measuring cost per service, transaction, or customer rather than tracking only the aggregate cloud bill. Flexera found 49% of organizations now use unit economics, up from 40% the year before.
How much of cloud spend can FinOps practices realistically recover?
Estimates cited in industry FinOps coverage, including from McKinsey, put realistic savings from rigorous FinOps discipline at 20% to 30% of cloud spend, a range that roughly brackets Flexera’s reported 29% waste figure for 2026.
