For five straight years, the share of cloud spending that companies simply threw away had been shrinking. That streak just broke. Flexera’s 2026 State of the Cloud report, its 15th annual edition, puts estimated wasted spend on infrastructure and platform services at 29% this year, up two points from 27% in 2025. It is the first increase since Flexera started tracking the trend down, and the survey’s own data points to a single culprit: AI workloads that bill in bursts nobody budgeted for.
The report surveyed 753 cloud decision-makers worldwide, and the numbers underneath the headline are just as telling as the 29% figure itself. Seventeen percent of organizations blew through their public cloud budget in the past year. Seventy-six percent of large enterprises now spend more than $5 million a month on public cloud. And global public cloud spending is on pace to cross $1 trillion in 2026, a jump of more than 21% year over year, according to G2’s analysis of the market. Cloud waste in 2026 is no longer a line item finance quietly manages, it is a boardroom metric, and a market of new tools is racing to fix it.
Cloud Waste Jumps to 29%, Ending a Five-Year Decline
The core finding is simple to state and harder to fix. “After five years on the downtrend, wasted cloud spend on IaaS and PaaS increased to 29% this year,” Flexera wrote in its analysis of the 2026 data. That two-point jump erases roughly a third of the ground gained over the prior half-decade of steady FinOps improvement, and it lands at the exact moment enterprises are pouring record sums into cloud infrastructure.
What makes the reversal notable is that it happened despite, not because of, weaker cost discipline. Flexera’s survey shows FinOps practices spreading faster than ever: more dedicated teams, more unit-economics tracking, more executive attention. The waste crept up anyway. That combination, rising maturity paired with rising waste, is the clearest signal yet that traditional cost governance was built for a workload pattern AI has already broken.
Inside Flexera’s 753-Respondent Global Survey
Flexera’s State of the Cloud report has run for 15 years, making it one of the longest continuous benchmarks of enterprise cloud behavior. The 2026 edition polled 753 cloud decision-makers across industries and company sizes, asking about spend, governance, tooling, and where budgets are actually going. That scale gives the 29% waste figure more weight than a single-vendor telemetry snapshot, since it reflects self-reported estimates from the people who own the budgets rather than usage data pulled from one platform’s billing API.
A few data points from the same survey round out the picture. Sixty-eight percent of organizations rank cloud cost optimization as their top cloud initiative, meaning the priority has not slipped even as waste ticked up. Sixty-three percent now run a dedicated FinOps team, up from 51% in 2024, a 12-point jump in two years. And responsibility for cost management is spreading beyond finance: Software Asset Management teams’ involvement rose from 6% to 15% year over year, while business units’ role climbed from 20% to 25% over the same period.
Why AI Workloads Are Breaking the Old Cost Playbook
Traditional cloud cost governance assumes workloads that run on a predictable schedule: a web server that scales with daily traffic, a batch job that runs overnight, a database that grows in a straight line. AI inference and training do not behave that way. A single fine-tuning run can spike GPU consumption for six hours and then vanish for a week. An agentic workflow can fan out into dozens of parallel model calls that nobody scoped in advance. Flexera’s report ties the 29% waste figure directly to that shift, noting that cloud-based AI workloads are surging and driving cost complexity that older forecasting models were never designed to handle.
The clearest evidence of how fast this changed is in how FinOps teams track AI spend at all. According to G2’s 2026 aggregation of cloud cost data, the share of FinOps teams actively tracking AI costs jumped from 31% in 2024 to 63% in 2025 to 98% in 2026. That is close to universal adoption of AI cost tracking in just two years, yet the waste number still climbed. Tracking spend and controlling it turned out to be two different problems, and most organizations solved the first one before they solved the second.
Unpredictable Billing Meets Predictable Budgets
Part of the mismatch is structural. Annual budgets get set months in advance, based on historical usage patterns. AI workloads do not respect that cadence. A product team that ships a new agent feature in March can double its inference bill by April, long before anyone revisits the forecast. Flexera’s report frames this explicitly: “cloud-based AI workloads are surging causing an increase in wasted cloud spend (29%) for the first time in five years,” according to Flexera’s own summary of the findings. Budgets built for infrastructure spend are colliding with consumption patterns built for experimentation.
The Five-Year Trend Line, in Flexera’s Own Numbers
The table below pulls together the year-over-year metrics Flexera and its data partners published alongside the 2026 report. Not every metric has a matching prior-year figure in the public data, so this table sticks to numbers that were explicitly reported with a comparison point, rather than estimating gaps.
| Metric | Prior Period | 2026 | Change |
|---|---|---|---|
| Wasted IaaS/PaaS cloud spend | 27% (2025) | 29% | +2 points, first rise in 5 years |
| Orgs with a dedicated FinOps team | 51% (2024) | 63% | +12 points in 2 years |
| FinOps teams tracking AI spend | 31% (2024) | 98% | +67 points in 2 years |
| SAM team involvement in cost management | 6% (prior year) | 15% | +9 points year over year |
| Business unit involvement in cost management | 20% (prior year) | 25% | +5 points year over year |
| Orgs exceeding their public cloud budget | n/a | 17% | New metric in 2026 report |
| Large enterprises spending >$5M/month on cloud | n/a | 76% | New metric in 2026 report |
Read across the rows and a pattern emerges: every governance metric moved in the right direction, and the waste number moved in the wrong one. That is the paradox at the center of this year’s report, and it is why Flexera’s language shifted from cost-cutting to cost complexity. “Cloud waste remains stubbornly high (29% estimated wasted spend), driven by growing complexity across hybrid cloud, SaaS and AI environments,” the company wrote in its five-year trend analysis.
A Wave of AI-Aware FinOps Tools Launched in September Alone
The market’s response arrived almost immediately, and it arrived in a cluster. In the two weeks before and after Flexera’s report circulated widely, at least five vendors announced new or expanded AI cost governance products, each pitching a slightly different angle on the same problem: nobody can see what their AI workloads actually cost until the bill arrives.
| Vendor / Product | Announced | Focus | Cloud Coverage |
|---|---|---|---|
| CostPerform Cloud & AI | Sept. 10, 2026 | Visualizing and allocating combined AI and cloud costs | Multi-cloud |
| Stacklet Cloud AI FinOps Benchmark | Sept. 10, 2026 | Tested controls defining AI cost governance | AWS, Azure, Google Cloud |
| FinOpsly AI Cost Governance | Sept. 5, 2026 | Unified control across GPUs, data pipelines, licenses | Multi-cloud |
| AICost.ai (CloudIntelligence.ai) | Sept. 4, 2026 | Decision-intelligence platform for agentic AI spend | Multi-cloud |
| Aerie by AidenAI | Sept. 8, 2026 | Azure-native cost optimization and governance | Azure only |
CostPerform’s announcement is notable for its pricing move: the company and its partner OnPoint Advisory are temporarily waiving fees on an initial FinOps assessment, a sign of how aggressively vendors are trying to land customers while the Flexera numbers are still fresh in procurement conversations. Stacklet took a different approach, publishing a benchmark of tested controls rather than a product, effectively trying to become the reference standard for what good AI cost governance looks like across all three major clouds at once.
FinOpsly’s pitch, described in a report on its launch, goes further than infrastructure billing: it tries to consolidate GPU rental costs, data pipeline spend, and software licensing into one governance layer, betting that the real waste hides in the gaps between those categories rather than inside any single cloud bill.
Market Impact: A Cost Management Software Market Racing to Catch Up
Behind the product launches sits a market that is expanding fast enough to absorb all of them. G2’s 2026 statistics roundup puts the cloud cost management software market on track to reach $19.27 billion by 2033, growing at a 17.6% compound annual rate. That is a market being built almost entirely around a failure mode: companies spending more than $1 trillion on cloud in 2026 while wasting close to a third of it, an equation that makes even an expensive FinOps tool look cheap by comparison.
The urgency shows up in how vendors are pricing their new AI tools. Several of the September launches lead with free assessments, benchmarks, or limited-time trials rather than sticker prices, a pattern that suggests vendors expect the Flexera numbers to do the selling and want to remove friction before a prospect’s enthusiasm cools. It also suggests a market still figuring out what customers will actually pay for AI-specific cost governance versus what they already get bundled into existing FinOps platforms.
How AWS, Azure, and Google Cloud Price Their Way Into the Problem
Part of the waste story is self-inflicted by the cloud providers themselves, through pricing structures that shift faster than most FinOps processes can track. Microsoft’s own pricing update for Azure AI Foundry, effective September 1, 2026, raised prices for EU Data Zone deployments to 20% above global pricing and pushed other non-US regional deployments 7% to 16% higher, according to Microsoft’s own foundry pricing documentation. Data residency, in other words, now carries a direct AI cost premium that many procurement teams had not modeled.
AWS moved in the opposite direction on some fronts while tightening others. Bedrock’s on-demand inference pricing for certain models dropped by as much as 80% in a July 2026 update, even as EC2 Capacity Blocks reserved for machine learning workloads rose roughly 15% across regions earlier in the year. That combination, cheaper inference paired with pricier dedicated capacity, is exactly the kind of shifting target that makes static budgets obsolete within a single fiscal quarter.
Historical Context: How Cloud Waste Became a Boardroom Metric
Cloud waste has been on Flexera’s radar since long before AI made it fashionable. In the early years of the State of the Cloud report, waste estimates hovered well above where they sit today, and the industry response was largely tactical: rightsizing instances, killing idle resources, buying reserved capacity. Those tactics worked, which is why the number spent five years trending downward. FinOps as a discipline formalized around exactly that playbook, and it matured into a recognized job function with its own certifications and a nonprofit foundation behind it.
What changed in 2026 is the type of resource being wasted. Rightsizing a virtual machine is a solved problem. Rightsizing a GPU cluster serving a large language model, where demand can be near zero one hour and saturated the next, is not. The same discipline that drove waste down for half a decade was built around resources that scale linearly and predictably. AI workloads do neither, and that gap is precisely what shows up as the 2-point reversal in this year’s numbers.
Competitive Comparison: Native Cloud Tools vs Third-Party FinOps Platforms
Every major cloud provider ships its own cost management console: AWS Cost Explorer, Azure Cost Management, Google Cloud’s Billing reports. All three have added AI-specific cost breakdowns over the past year. The limitation is structural rather than technical: none of them show spend across the other two clouds, and most enterprises with more than $5 million a month in cloud spend run workloads on at least two providers.
Where Native Tools Hold Up
Native tools remain the fastest way to get granular, real-time billing data for a single cloud, and they are free or bundled into existing support contracts. For a company running entirely on one provider, that is often enough. The gap opens up the moment AI workloads span providers, which is increasingly common given how differently AWS, Azure, and Google price GPU capacity and model inference.
Where Third-Party Platforms Win
Independent platforms like the ones launched this September exist specifically to unify that view, and to add the allocation and chargeback logic that native consoles handle poorly. The tradeoff is cost and integration overhead: a third-party FinOps platform requires connecting to every cloud’s billing API, mapping tags consistently across providers, and, in most cases, paying a subscription on top of the cloud spend it is meant to reduce.
What This Means for Engineering Teams Day to Day
For engineers, the practical shift is in what gets tagged and reviewed before code ships, not just after the bill arrives. Teams that treated cost tagging as an afterthought are the ones most likely to show up in next year’s waste numbers, because AI-specific line items, model endpoint calls, vector database storage, GPU rental hours, do not always inherit tags the same way a standard compute instance does. A simple resource tagging policy, enforced at deploy time rather than audited after the fact, is one of the few controls that scales with unpredictable AI usage.
{
"Version": "2026-09-01",
"RequiredTags": ["team", "project", "environment", "ai-workload"],
"EnforcementMode": "deny-on-missing-tag",
"AppliesTo": [
"ec2:Instance",
"bedrock:ModelInvocation",
"sagemaker:TrainingJob",
"gpu:ReservedCapacity"
]
}That kind of enforced tagging will not close a 29% waste gap on its own, but it is the prerequisite for every FinOps tool listed above. None of the September product launches can allocate cost accurately against untagged resources, which means the tooling race described earlier only pays off for organizations that fix tagging discipline first.
What Flexera’s Data Actually Shows, in Its Own Words
It is worth returning to Flexera’s own framing, because the company has been measuring this exact metric for 15 years and its language this year is more cautious than triumphant. “After five years of decline, wasted cloud spend increased slightly to 29%, reflecting growing cost complexity from AI and new IaaS and PaaS services,” the company noted on its official 2026 State of the Cloud report page. The word complexity recurs across nearly every section of the report, more than waste or cost alone, which tells you where Flexera thinks the real problem sits.
Outside commentators reached a similar conclusion from a different angle. A Forbes Tech Council piece published September 8, 2026 argued that the next phase of cost control will have to be automated rather than manual, framing agentic AI itself as a tool for finding savings rather than only a source of new spend, a bet that several of the vendors above are already making with their product launches.
Cloud Cost Management Market at a Glance
| Data Point | Figure | Source |
|---|---|---|
| Global public cloud spend, 2026 | >$1 trillion, +21% YoY | G2 2026 cloud cost statistics |
| Estimated wasted cloud spend, 2026 | 29% | Flexera State of the Cloud 2026 |
| Cloud cost management software market by 2033 | $19.27 billion (17.6% CAGR) | G2 2026 cloud cost statistics |
| Enterprises spending >$5M/month on cloud | 76% of large enterprises | Flexera State of the Cloud 2026 |
| Orgs exceeding their cloud budget in the past year | 17% | Flexera State of the Cloud 2026 |
| FinOps teams tracking AI spend, 2026 | 98% (up from 31% in 2024) | G2 / Flexera 2026 data |
5 Predictions for Cloud Cost Management Through 2027
- Waste keeps climbing before it falls again. Expect Flexera’s 2027 report to show a third straight year outside the old downward trend, since AI adoption is still accelerating faster than tagging and forecasting practices can adapt.
- AI cost governance becomes a checkbox in vendor RFPs. With five products launched in a single September week, enterprise buyers will start treating dedicated AI cost controls as a baseline requirement rather than a differentiator by mid-2027.
- Cloud providers add usage caps and budget alerts as default features. Expect AWS, Azure, and Google Cloud to compete on built-in guardrails for AI spend, partly to blunt the appeal of third-party platforms like the ones launched this month.
- Consolidation hits the new wave of FinOps startups. Not every vendor that launched an AI cost product in September 2026 will still be independent a year from now; the market is too crowded for five point solutions to all survive on the same problem.
- Data residency pricing premiums spread beyond Azure. Following Microsoft’s EU Data Zone increase, expect AWS and Google Cloud to introduce comparable regional AI pricing tiers as regulatory pressure on data location grows.
How Engineering and Finance Teams Can Cut Waste Now
None of the fixes here are exotic, which is itself part of the frustration in this year’s numbers: the tools existed before the waste went up. Enforcing tags at deploy time, as shown above, is the cheapest control available and the one most often skipped under deadline pressure. Setting hard budget alerts on AI-specific services, rather than blanket account-level alerts, catches runaway inference costs before they compound across a billing cycle. Reviewing GPU reservation commitments quarterly instead of annually matches the review cadence to how fast AI pricing actually moves, based on the swings seen in Bedrock and EC2 Capacity Block pricing this year alone.
The organizations already running dedicated FinOps teams, now 63% of respondents, are not immune to the waste increase, which suggests that having a team is necessary but not sufficient. The gap is in giving that team AI-specific visibility rather than folding AI spend into the same dashboards built for compute and storage. That is precisely the gap the September product wave is trying to fill, and it is also the gap any team can start closing internally before buying another subscription.
Frequently Asked Questions
What is the current cloud waste percentage in 2026?
Flexera’s 2026 State of the Cloud report estimates wasted cloud spend on IaaS and PaaS at 29%, up from 27% in 2025. It is the first year-over-year increase in five years of previously declining waste.
Why did cloud waste increase after five years of decline?
Flexera attributes the reversal primarily to AI workloads, which consume compute in irregular, hard-to-forecast bursts that traditional budgeting and rightsizing practices were not designed to handle.
How many organizations exceeded their cloud budget in 2026?
Seventeen percent of organizations surveyed by Flexera said they exceeded their public cloud budget over the past year, despite rising investment in FinOps teams and processes.
What percentage of FinOps teams now track AI spend?
According to G2’s 2026 analysis, 98% of FinOps teams now track AI spend, up sharply from 63% in 2025 and just 31% in 2024, making AI cost tracking nearly universal in two years.
Which companies launched new cloud cost management tools in September 2026?
Five vendors announced AI-focused cloud cost products in early-to-mid September 2026: CostPerform, Stacklet, FinOpsly, AICost.ai, and AidenAI’s Aerie for Azure. Each targets a different slice of AI and multi-cloud spend governance.
How big is the cloud cost management software market?
G2’s 2026 statistics put the cloud cost management software market on track to reach $19.27 billion by 2033, growing at a compound annual rate of 17.6%, driven largely by demand for AI-specific cost controls.
Do native AWS, Azure, and Google Cloud tools track AI costs well enough on their own?
Native tools handle single-cloud billing well but do not unify spend across providers. Most enterprises spending more than $5 million a month on cloud run workloads across at least two providers, which is the gap third-party FinOps platforms are built to close.
Will cloud waste keep rising in 2027?
Based on the current trajectory of AI adoption outpacing tagging and forecasting practices, analysts and industry commentary suggest waste could climb further before the new generation of AI-aware FinOps tools brings it back down.
