Cloud Spending Statistics 2026: How Much Businesses Spend
Worldwide public cloud spending is forecast to pass $1 trillion in 2026, and cloud infrastructure providers booked more than $143 billion of revenue in the second quarter alone, with generative AI cited as the main driver of that growth. Inside individual companies, the picture is far less orderly. In Flexera's 2026 State of the Cloud survey, 85% of respondents named managing cloud spend a top challenge, and respondents estimated that 29% of IaaS and PaaS spending is wasted. Finance teams are asked to approve budgets that engineering decisions can change overnight, AI workloads add costs that are hard to attribute, and a market growing at more than 20% a year leaves little room for forecasting mistakes. This guide separates the major 2026 datasets, explains what each one actually measures, and turns them into benchmarks, budgeting methods and cost-control practices for businesses deciding how much to spend.
TL;DR
IDC forecasts that worldwide spending on public cloud services will exceed $1 trillion in 2026, with growth above 21%. That figure covers SaaS, PaaS and IaaS, and SaaS is more than half of it.
Gartner forecasts $287.3 billion of worldwide IaaS spending in 2026 (up 29.3%), and AI-optimized IaaS is a $42.3 billion subset of that total. The two numbers must not be added together.
Synergy Research Group estimates $143.4 billion of cloud infrastructure services revenue in Q2 2026 alone, up 43% year over year. That is quarterly provider revenue, not annual customer spending.
There is no honest average cloud bill for all businesses. In Flexera's survey of 753 respondents, 76% of large enterprises reported spending more than $5 million per month, which would equal more than $60 million a year if sustained (calculation).
Survey respondents estimate that 29% of IaaS and PaaS spending is wasted, and 98% of FinOps Foundation respondents now manage AI spend. Both are self-reported survey results, not audited totals.
Global public cloud spending is forecast to exceed $1 trillion in 2026, according to IDC, while Gartner forecasts $287.3 billion for IaaS alone. Organization-level bills vary widely: in Flexera's 2026 survey, 76% of large enterprises reported spending more than $5 million per month. No credible universal average cloud bill exists for all businesses.
Table of Contents
Cloud Spending Statistics 2026 at a Glance
Cloud spending statistics for 2026 describe at least four different things: end-user spending across all public cloud services, end-user spending on infrastructure only, provider revenue for infrastructure services, and spending reported by survey respondents inside organizations. The table lists the strongest verified figure for each, with its scope, period and source, so that no number is read as something it is not.
Metric | 2026 figure | Growth/change | Exact scope | Period | Source |
|---|---|---|---|---|---|
Public cloud services spending, worldwide | More than $1 trillion | More than 21% | End-user spending on SaaS, PaaS and IaaS across 53 countries | 2026 forecast | IDC, Mar 4, 2026 |
IaaS spending, worldwide | $287.3 billion | +29.3% (from $222.2 billion in 2025) | Infrastructure as a service only, end-user spending | 2026 forecast | Gartner, Aug 10, 2026 |
AI-optimized IaaS, worldwide | $42.3 billion | +96% (from $21.5 billion) | Subset of IaaS built for AI workloads; not additive | 2026 forecast | Gartner, Aug 10, 2026 |
Sovereign cloud IaaS, worldwide | $80 billion | +35.6% | Sovereign IaaS only; a separate forecast, not additive | 2026 forecast | Gartner, Feb 9, 2026 |
Cloud infrastructure services revenue | $143.4 billion | +43% year over year | IaaS, PaaS and hosted private cloud revenue earned by providers, one quarter | Q2 2026 estimate | Synergy Research Group, Jul 30, 2026 |
Large enterprises spending over $5 million per month | 76% of large enterprises | Top spending tiers rose | Survey respondents (753 total) | 2026 survey | Flexera, 2026 |
Estimated wasted IaaS and PaaS spend | 29% | Up from 27% | Respondent estimate, not an audit | 2026 survey | Flexera, 2026 |
FinOps teams managing AI spend | 98% | Up from 63% in 2025 | 1,192 survey respondents | Survey released Feb 2026 | FinOps Foundation, 2026 |
Worldwide IT spending (context) | $6.37 trillion | +14.2% | All IT categories, including devices and services | 2026 forecast | Gartner, Jul 27, 2026 |
The headline totals differ because they answer different questions. IDC's Worldwide Software and Public Cloud Services Spending Guide measures what buyers spend across SaaS, PaaS and IaaS, and IDC says SaaS alone accounts for more than half of the total. Gartner's $287.3 billion covers IaaS only, so it relates to one category inside IDC's broader scope, although the two firms define IaaS differently. Synergy counts revenue earned by providers of IaaS, PaaS and hosted private cloud in a single quarter, so its trailing-twelve-month figure of about $500 billion is a provider-side measure and not a calendar-year buyer forecast.
Taken together, these figures suggest a market in which infrastructure and AI platforms grow fastest while SaaS remains the largest block of spending. For business planning, the implication is to quote every number with its definition. Readers who want adoption and market context beyond spending can continue with Articsledge's cloud computing statistics guide, and those who need a plain-language definition of the technology can start with what cloud computing is.
How Much Are Businesses Spending on Cloud in 2026?
At the market level, IDC forecasts that businesses worldwide will spend more than $1 trillion on public cloud services in 2026. At the company level, spending ranges from small subscriptions to eight-figure monthly bills, and no source reviewed for this article supports a single average for all businesses.
Worldwide market spending
IDC reports that the United States will account for $647 billion of 2026 public cloud spending, Western Europe for $255 billion and Asia/Pacific excluding Japan and China (APeJC) for $84 billion. Banking, software and information services, and retail are the three largest spending industries, together about 25% of the total. These are forecasts of end-user spending, and they describe the market, not the bill of any single company.
Large-enterprise spending
Flexera's 2026 State of the Cloud report, a survey of 753 cloud decision-makers and users, found that 76% of large enterprises spend more than $5 million per month on public cloud. As arithmetic, $5 million a month equals more than $60 million a year if spending stayed at that level for twelve months. Flexera's 2021 report found that 36% of enterprises spent more than $1 million per month, but Flexera itself notes that the audience segments and thresholds differ, so the two results are a directional comparison and not a like-for-like trend.
What the data says about smaller businesses
No current, representative average cloud bill for small businesses appears in the primary sources reviewed. IDC's guide covers eight company sizes, but its public release does not publish values by size. The FinOps Foundation's 2026 survey includes 20% SMB respondents (fewer than 1,000 employees), 33% enterprise (1,000 to 9,999) and 47% large enterprise (10,000 or more), yet it reports practices, not typical bills. Flexera's findings that 73% of organizations run a hybrid cloud estate and that GenAI use is universal among respondents apply to a survey population that leans toward cloud-active organizations, which is why they should not be generalized to every small firm. Articsledge's cloud adoption statistics article covers usage patterns in more detail.
Why one universal average misleads
A single average would blend businesses that differ on nearly every cost driver:
Workload intensity, architecture and engineering maturity
Revenue model, especially whether the product itself runs in the cloud
SaaS footprint, AI and GPU consumption, storage and data transfer
Observability, security, backup and uptime requirements
Commitment discounts and the share of infrastructure still on premises
For business planning, the implication is to replace "what is the average?" with "what do comparable organizations with similar workloads spend per unit of output?" The budgeting and benchmarking sections show how to answer that from an organization's own usage.
Global Public Cloud Spending in 2026
IDC forecasts that worldwide spending on public cloud services will surpass $1 trillion in 2026, growing more than 21%, and expects that spending to double by 2029. Gartner's September 2025 forecast abstract likewise projected 21.3% public cloud services growth in 2026 and a market of $1.48 trillion by 2029, although the two firms' dollar totals rest on different definitions.
IDC's March 4, 2026 release describes where the money goes. SaaS remains the largest deployment category at more than half of spending, while PaaS is the fastest-growing at more than 37% in 2026, driven by demand for AI platforms, data platforms and cloud-native development. Within secondary markets, IDC says investment concentrates in IaaS, enterprise resource management and security software. The five-year growth rate exceeds 20% in the Middle East and Africa, Latin America, APeJC, Central and Eastern Europe and Western Europe.
As a calculation, a market above $1 trillion after growing 21% implies a prior-year base of roughly $830 billion (rounded). Vintage and definitions matter: Gartner's November 2024 forecast put 2025 worldwide public cloud end-user spending at $723.4 billion, under a different taxonomy and an earlier forecast, so the two should not be compared directly.
IDC also names risks to its outlook, including regulatory fragmentation, skills shortages and cloud cost optimization pressures. The last of these is the subject of much of this article. Regional sovereignty requirements shape spending in Europe in particular, a theme covered in the trends section and in Articsledge's guide to sovereign cloud.
Cloud Spending by Company Size
Credible sources describe organization size through definitions, sample composition and spending thresholds, not through published average bills. The table summarizes what is actually reported, and what is not.
Segment | Definition used by source | What the source reports | What is not reported | Source |
|---|---|---|---|---|
SMB | Fewer than 1,000 employees | 20% of FinOps Foundation respondents; AI priorities rise alongside foundational FinOps work | An average or typical cloud bill | FinOps Foundation, 2026 |
Enterprise | 1,000 to 9,999 employees | 33% of respondents | A spending distribution | FinOps Foundation, 2026 |
Large enterprise | 10,000 or more employees (FinOps Foundation); Flexera uses its own definition | 47% of FinOps respondents; teams at organizations managing $100 million or more in annual spend average 8 to 10 practitioners plus 3 to 10 contractors | Whether results represent all large enterprises | FinOps Foundation, 2026 |
Large enterprise (monthly spend) | As defined by Flexera | 76% spend more than $5 million per month on public cloud | The remaining 24% by band; survey is not audited | Flexera, 2026 |
Headcount is a weak proxy for cloud cost. A cloud-native B2B SaaS company with a few hundred employees can run more compute, storage and data transfer than a far larger organization whose systems remain mostly on premises, because its product is the workload. Revenue model, architecture, data volume, uptime requirements and AI usage typically explain more of the bill than the number of employees. The FinOps Foundation notes that smaller organizations balance AI alongside foundational work, while larger organizations increasingly treat AI as a dedicated domain, which suggests the same cost categories matter at every size even when the scale differs.
For business planning, the implication is to benchmark against organizations with a similar business model and workload profile, then use size only as a secondary filter.
Cloud Spending by Service Model: IaaS, PaaS, SaaS and Beyond
Businesses pay for cloud through three main service models: infrastructure, platforms and software. IDC says SaaS accounts for more than half of public cloud spending in 2026, PaaS is the fastest-growing model, and Gartner forecasts IaaS at $287.3 billion.
Service model | What the business pays for | Typical billing pattern | Current verified data point |
|---|---|---|---|
Virtual machines, storage, networking | Usage-based, with optional commitments | Gartner forecasts $287.3 billion worldwide in 2026, up 29.3% | |
Managed databases, application and AI platforms | Usage-based or tiered | IDC: fastest-growing deployment model, over 37% growth in 2026 | |
Business applications | Per seat, per usage or tiered subscription | IDC: more than half of public cloud spending; Flexera: the largest share of respondents report monthly SaaS spend of $200,001 to $500,000 | |
Hosted virtual desktops | Per user | No current 2026 figure verified for this article | |
AI-optimized IaaS | GPU and accelerator infrastructure | Usage-based or reserved capacity | Gartner forecasts $42.3 billion in 2026; a subset of IaaS |
SaaS-heavy and infrastructure-heavy businesses produce very different bills. A company that mostly buys applications sees cost scale with seats and tiers, and the control problem is license sprawl: Flexera found that 8% of respondents are not tracking SaaS costs, up from 5% the year before. A company that builds on infrastructure sees cost scale with consumption, and the control problem is allocation and utilization. Many businesses carry both. Articsledge's guide to SaaS management software covers tooling for the first problem, and the sections below cover the second.
Because AI-optimized IaaS is part of IaaS, a reader should treat the $42.3 billion as a closer look at one slice of the $287.3 billion and never add the two.
Cloud Infrastructure Spending and Market Growth
Synergy Research Group estimates that worldwide spending on cloud infrastructure services reached $143.4 billion in the second quarter of 2026, up 43% year over year and the fastest growth in eight years. Trailing-twelve-month revenue reached about $500 billion.
Synergy, now part of TechInsights, reported the figures on July 30, 2026. The increase was more than $43 billion from Q2 2025, and the market has roughly doubled in size over 11 quarters of accelerating annual growth. Public IaaS and PaaS services account for most of the market and grew 47% in the quarter. The United States grew 49%, above the worldwide rate, and Synergy's chief analyst said generative AI-specific cloud services are growing at 165% year over year. Nine neocloud companies are now among the top 40 providers by cloud infrastructure revenue.
Scope matters. Synergy's category covers IaaS, PaaS and hosted private cloud revenue earned by providers. It excludes SaaS, so it is narrower than IDC's public cloud total, and it measures provider revenue, not what buyers report spending. A single quarter should not be multiplied by four and presented as an annual market, which is why the trailing-twelve-month figure is the better annual reference. Readers who want the infrastructure layer explained can use Articsledge's overview of cloud infrastructure and of hosted private cloud.
For business planning, the implication is that forecasts have been revised upward during the year: Gartner raised its 2026 worldwide IT spending forecast from $6.08 trillion in October 2025 to $6.37 trillion in July 2026. Unit prices, capacity and commitment terms therefore deserve explicit attention in budgets.
AWS, Microsoft Azure, Google Cloud and Oracle Spending Indicators
Provider financial results show how fast each company's cloud business is growing, but they are vendor revenue and not customer spending, and each company defines its cloud segment differently. The figures below are not directly comparable.
Provider | Reported indicator | Period | Growth | What the figure includes |
|---|---|---|---|---|
AWS (Amazon) | $42.2 billion AWS segment sales; $169 billion annualized run rate | Q2 2026 (quarter ended June 30) | +37% | All AWS segment sales; run rate is Amazon's annualized description, not a forecast |
Microsoft Cloud (Microsoft) | $59.3 billion; about $214 billion for fiscal 2026 per CEO remarks | Q4 FY2026 (quarter ended June 30); fiscal year ended June 30, 2026 | +27% (quarter) | Azure and other cloud services plus Microsoft 365 commercial cloud, Dynamics 365, LinkedIn and related offerings; not Azure alone |
Azure (Microsoft) | Azure revenue above $100 billion in fiscal 2026, a first | Fiscal year ended June 30, 2026 | Azure and other cloud services +43% in Q4 | Azure only, as Microsoft reports it |
Google Cloud (Alphabet) | $24.8 billion revenue; $8.8 billion operating income; $514 billion backlog | Q2 2026 (quarter ended June 30) | +82% | The Google Cloud segment, which includes Google Cloud Platform alongside other enterprise offerings; not GCP alone |
Oracle total cloud (Oracle release) | $11.6 billion | Fiscal Q1 2027 (quarter ended August 31, 2026) | +62% | Cloud infrastructure plus cloud applications |
Oracle cloud infrastructure | $7.4 billion | Fiscal Q1 2027 | +121% | IaaS only; Oracle also reported $4.2 billion of cloud applications (SaaS), up 10% |
Three cautions apply. First, reporting periods and fiscal calendars differ, so quarterly figures should not be summed into a market total. Second, backlog and remaining performance obligations, such as Google Cloud's $514 billion and Oracle's $664 billion, are contracted revenue not yet recognized and are not spending in 2026. Third, a provider's revenue divided by its customers does not yield an average customer bill, because customer counts, sizes and mixes are not disclosed on a comparable basis.
Market share is a separate measure. Synergy estimated Q2 2026 shares of 28% for Amazon, 20% for Microsoft and 15% for Google, and Articsledge's cloud market share analysis covers that data in full. Readers who want a service-level view of the largest provider can use the Amazon Web Services guide.
What Drives a Company's Cloud Bill?
A cloud bill is the sum of many separately metered services, and the largest lines are usually compute, storage, databases, data transfer and managed services. Costs outside the provider invoice, such as tooling and engineering time, belong in a full total-cost view.
Cost category | Main billing driver | Why costs grow | Common optimization lever |
|---|---|---|---|
Compute | Instance size, hours and purchase model | Overprovisioned instances, always-on capacity, traffic growth | Rightsizing, autoscaling, commitments, spot capacity |
GPUs and accelerators | Accelerator type, hours and reserved capacity | AI training and inference demand, idle reserved capacity | Scheduling, utilization tracking, smaller or cheaper models |
Storage | Gigabytes stored, tier and request volume | Snapshots, backups and data that is never deleted | Lifecycle policies, tiering, retention limits |
Provisioned capacity, throughput and replicas | Overprovisioned clusters, high-availability replicas | Capacity reviews, reserved capacity, query tuning | |
Data transfer and networking | Gigabytes leaving a region, zone or provider | Cross-region traffic, internet egress, multicloud designs | Placement of services, caching, traffic architecture |
Managed services | Per-request, per-hour or per-feature pricing | Convenience adoption, overlapping tools | Service reviews, consolidation |
Invocations, duration and memory | Chatty architectures, cold-start workarounds | Function tuning, batching | |
Kubernetes and containers | Cluster nodes, control plane and add-ons | Low node utilization, many small clusters | Bin packing, cluster consolidation, resource requests |
SaaS licenses | Seats, tiers and usage | Unused seats, duplicate tools | License rationalization, renewal reviews |
Security tools | Data scanned, endpoints and events | Coverage expansion, overlapping products | Tool consolidation, scoped scanning |
Observability (telemetry) | Log volume, metrics and trace retention | Verbose logging, long retention | Sampling, tiered retention |
Backup, disaster recovery and resilience | Replicated data and standby capacity | Multi-region designs, availability zones | Tiering recovery objectives by workload |
Support and compliance | Percentage of spend or flat tiers; audit tooling | Spend growth lifts percentage-based support | Plan reviews, scoped compliance zones |
Operations labor | Engineering and FinOps time | Complexity, tool sprawl | Automation, platform standards |
Architecture shapes several of these lines at once. Teams adopting cloud-native infrastructure gain elasticity but also take on more metered components, so cost visibility has to be designed in. This section is not a pricing tutorial, because list prices change and negotiated rates differ, but the table shows where to look first.
AI Is Reshaping Cloud Spending in 2026
Gartner forecasts that worldwide spending on AI-optimized IaaS will grow 96% in 2026 to $42.3 billion, and that spending on running models (inference) will pass spending on building them (training) for the first time. AI is a large and fast-growing slice of infrastructure spending, not a replacement for total cloud spend.
Gartner's August 10, 2026 release puts 2025 AI-optimized IaaS spending at about $21.5 billion and projects $66.1 billion in 2027, a 56.5% increase. Within the 2026 figure, Gartner forecasts $23.3 billion for inference and $19 billion for training, so inference is 55% of the total, rising to 59% in 2027. As a calculation, $42.3 billion is about 14.7% of Gartner's $287.3 billion IaaS forecast, and because it is a subset it adds nothing to that total.
Inference matters for budgeting because it is an ongoing, variable expense. Training is a project with an end date, while inference runs every time a customer or employee uses a feature. Cost therefore follows usage volume, model choice, prompt and context length (tokens), and response time targets, and it also draws on storage and networking. GPU and accelerator capacity is a further variable: reserved capacity that sits idle still costs money, while on-demand capacity may be scarce. Articsledge's guides to AI inference and model training explain the technical difference.
Other 2026 indicators point the same way. Synergy estimates that generative AI-specific cloud services grew 165% year over year in Q2 2026. In Flexera's survey, all respondents use GenAI in some capacity, 45% use it extensively, and use of GenAI as a public cloud service rose from 50% to 58%. Flexera also notes that cloud-based AI services often cost more than traditional options and use different pricing metrics, which makes forecasting harder.
The FinOps Foundation found that 98% of respondents now manage AI spend, up from 31% in 2024. Their top challenges are visibility into AI costs, allocating AI costs to business units and determining AI value, and their most requested tool feature is granular monitoring of tokens, LLM requests and GPU utilization. For wider context, see Articsledge's generative AI statistics, enterprise AI statistics and AI implementation cost guides.
How Much Cloud Spending Is Wasted?
Respondents to Flexera's 2026 State of the Cloud survey estimate that 29% of IaaS and PaaS spending is wasted, up from 27% the year before and the first increase in five years. This is a survey estimate, not an audited loss rate, and no company should assume it wastes exactly 29%.
Waste is spending that delivers no business value, and it takes recognizable forms: idle or oversized cloud instances and over-provisioned resources, unattached storage volumes and forgotten snapshots, development and test environments left running, duplicate or unused SaaS seats, excessive logging, avoidable data transfer, poorly tagged resources that nobody owns, underused GPUs, and commitments that are either missing for steady workloads or too large and inflexible. Flexera attributes the 2026 rise to added cost complexity from AI and new PaaS and SaaS offerings.
As an illustration of scale only, 29% of a hypothetical $1 million monthly bill is $290,000 a month, or $3.48 million over 12 months (calculation). Because the survey figure is an average of self-estimates, a given organization may sit well above or below it.
Indicator | Value | Scope | Source |
|---|---|---|---|
Estimated wasted IaaS and PaaS spend | 29% (up from 27%) | Respondent estimate | Flexera, 2026 |
Managing cloud costs is a top challenge | 85% | Survey respondents | Flexera, 2026 |
Optimizing cloud costs is the top initiative | 68% | Survey respondents | Flexera, 2026 |
Use any one commitment discount per provider | Fewer than half | More than half still use on-demand pricing | Flexera, 2026 |
Not tracking SaaS costs | 8% (up from 5%) | Survey respondents | Flexera, 2026 |
Use unit economics to understand cost per service | 49% (up from 40%) | Survey respondents | Flexera, 2026 |
The FinOps Foundation adds a useful nuance: mature practitioners report diminishing returns from traditional optimization after the largest waste is removed, leaving many smaller opportunities. For business planning, the implication is that a first optimization pass often pays off quickly, while later gains depend on allocation, forecasting and architecture decisions.
Why Cloud Budgets Run Over Forecast
Cloud budgets run over because consumption is variable, ownership is spread across teams and pricing is complex. Flexera and the FinOps Foundation both identify forecasting and cost visibility as persistent difficulties, particularly for AI and SaaS spend.
The main causes are:
Variable consumption that follows traffic, data volume and release cycles, not a fixed schedule
Rapid product growth and engineering self-service, where teams can launch resources without a budget conversation
AI workloads with new pricing metrics, such as tokens, and uncertain adoption
Pricing complexity, including egress, observability and cross-region transfer charges that are easy to overlook
Underused or mis-sized commitments, which raise unit costs or create paid-for capacity that sits idle
Mergers and acquisitions, accidental multicloud and decentralized purchasing
SaaS seat sprawl and renewals that bypass central review
Forecasting methods that extrapolate last month's bill without modeling drivers
Flexera reports that fewer than half of organizations use any single commitment discount with each provider, and that distributed ownership, hybrid complexity, SaaS growth and AI all contribute to waste and unpredictability. The FinOps Foundation finds that the most prioritized capabilities across technology categories are allocation, forecasting, budgeting and planning, which suggests that teams are building the foundations before they optimize. Its respondents also rank pre-deployment architecture costing among the top requested tool features, which points to estimating cost before a service ships.
The most recent budget-variance figure in Flexera's published material is historical: in its 2021 report, respondents said public cloud spend was over budget by an average of 24%. A comparable 2026 variance figure was not available in the sources reviewed, so this article does not present one. Articsledge's cloud governance guide covers the policies that keep ownership clear.
FinOps Statistics and Cloud Cost Management in 2026
FinOps, the practice of managing cloud and technology costs through collaboration between engineering, finance and business teams, has become mainstream in 2026 and is moving from cutting costs toward managing technology value. Flexera found that 63% of respondents have FinOps teams and 71% have a Cloud Center of Excellence or similar.
The FinOps Foundation's sixth annual State of FinOps survey, released in February 2026, drew 1,192 respondents whose organizations represent more than $83 billion in annual cloud spend. Its findings show scope expanding well beyond public cloud:
98% of respondents manage AI spend, up from 63% in 2025 and 31% in 2024
90% manage or plan to manage SaaS, 64% licensing, 57% private cloud and 48% data center, and 28% are including or planning to include labor costs
78% of FinOps practices report to the CTO or CIO organization, up 18 points since 2023, while 8% report to the CFO
Practitioners with VP-level or C-suite engagement show two to four times more influence over technology selection decisions
81% use centralized enablement (60%) or hub-and-spoke models (21%), and teams stay lean even at large scale
The Foundation also changed its mission statement from advancing the people who manage the value of cloud to the value of technology, and it reports that governance, forecasting, organizational alignment and scope expansion together outweigh optimization alone as priorities.
Flexera's data show the same shift from savings to value. In its 2026 survey, 64% of organizations measure cloud progress by value delivered to business units, up 12 points, while the share using cost efficiency fell 6 points. FinOps team responsibility for governing IaaS and PaaS rose from 38% to 45%, and 49% of respondents now use unit economics, up from 40%.
Unit economics means measuring cloud cost against a business outcome. Useful examples are cloud cost per customer, per transaction, per API call, per AI inference or per dollar of revenue. A bill that rises 30% is a concern if revenue is flat and a sign of healthy growth if cost per customer is falling. Articsledge's cloud operating model guide shows where these practices sit in a wider operating structure.
What Should a Business Budget for Cloud?
No credible source supports a universal recommended percentage of revenue or IT budget for cloud, so a business should build its budget from its own usage. The framework below turns billing data into a defensible annual number.
Expected annual cloud budget = baseline workload cost + growth-driven variable usage + storage and data-transfer costs + managed services + security and observability + AI workloads + support and compliance + contingency − realistic commitment and discount savings
To populate each term from actual usage:
Export at least 12 months of billing data by service, account and team, ideally in a consistent format, and fix tagging gaps first. A well-structured cloud landing zone makes this easier.
Separate steady demand from variable demand. Baseline workloads that run continuously are the candidates for commitments, while variable usage should follow a driver such as customers or transactions. Articsledge's guide to the cloud workload explains how to classify them.
Convert growth plans into unit costs, for example expected new customers multiplied by cloud cost per customer.
Model AI separately: expected requests, tokens per request, model price and the share of traffic that will use AI features, plus any training or experimentation.
Add storage, data transfer, managed services, security, observability, SaaS licenses, support and compliance as their own lines.
Set contingency from the organization's historical forecast error, not from a rule of thumb.
Subtract only the savings from commitments that actually apply to steady usage, net of the risk of under-utilizing them.
The following hypothetical example shows the arithmetic only. The amounts are illustrative inputs, not industry averages or benchmarks.
Budget component (hypothetical) | Monthly amount |
|---|---|
Baseline workload cost | $40,000 |
Growth-driven variable usage | $8,000 |
Storage and data transfer | $6,000 |
Managed services | $10,000 |
Security and observability | $7,000 |
AI workloads | $9,000 |
Support and compliance | $5,000 |
Subtotal | $85,000 |
Contingency (10% of subtotal) | $8,500 |
Commitment savings (25% on $30,000 of steady baseline) | −$7,500 |
Net monthly budget | $86,000 |
Annual budget (12 months) | $1,032,000 |
Three illustrative scenarios show how the same formula behaves in different organizations. They describe emphasis and risk, not typical spending.
Scenario | Likely dominant cost lines | Biggest forecasting risk | Budgeting approach |
|---|---|---|---|
Smaller digital business | Compute, storage, SaaS subscriptions, a few managed services | Unmonitored growth and SaaS sprawl | Monthly reviews, simple tagging, budget alerts, commitments only for clearly steady usage |
Growing SaaS or mid-market company | Compute, databases, data transfer, observability, new AI features | Usage growing faster than revenue; telemetry costs | Cost-per-customer budgets, showback, quarterly commitment reviews |
Large cloud-intensive enterprise | Multi-account compute, data platforms, GPUs, resilience, support | Complexity, AI demand and decentralized purchasing | Product-level budgets, a FinOps team, chargeback, scenario modeling, negotiated agreements |
Teams that want tooling support can review Articsledge's overview of budget planning software, although the method matters more than the product.
How to Benchmark Your Cloud Spending
A useful benchmark compares an organization with peers that resemble it, or with its own history, using ratios tied to business output. No credible source reviewed provides a general cloud-cost-to-revenue standard, so none is offered here.
Peer comparisons are only meaningful when peers share most of the following traits:
Company size and revenue
Business model, especially whether the product runs on cloud infrastructure
Workload profile and growth rate
Geography and compliance requirements
Architecture, including hybrid and multicloud footprint
Within the organization, these ratios can be measured and tracked over time:
Cloud cost per customer, per employee, per transaction or per workload
Cloud cost as a percentage of the revenue from the digital product it supports
Infrastructure cost per transaction
AI inference cost per task or per conversation
Gross-margin impact for SaaS businesses, since hosting is a cost of revenue
These are internal measures and not industry standards. Their value lies in trend and variance: a falling cost per customer with rising total spend usually signals healthy scaling, while a rising unit cost points to waste or an architecture problem. Survey data from Flexera and the FinOps Foundation can benchmark practices, such as how many organizations use unit economics (49%) or have a FinOps team (63%), but they do not benchmark bills. Teams planning a move can also compare unit costs before and after a cloud migration to confirm that expected savings appeared.
How Businesses Can Reduce Cloud Spending Without Slowing Growth
The most reliable way to reduce cloud spending is to start with visibility and ownership, remove obvious waste, then optimize rates and architecture. The goal is business value per dollar, not the lowest possible bill, and every measure below carries a tradeoff.
Allocation and tagging. Assign every cost to a team, product or customer. Without ownership, no other action holds. The tradeoff is upfront discipline from engineering teams.
Idle and orphaned resources. Remove unattached volumes, stale snapshots and forgotten environments. The risk is deleting something still needed, so confirm ownership first.
Rightsizing and autoscaling. Match capacity to demand using cloud automation. Aggressive settings can hurt latency during spikes.
Scheduling non-production environments. Shut down development and test systems outside working hours.
Commitments. Reserved instances, savings plans and committed-use discounts lower unit prices for steady usage. Flexera finds that fewer than half of organizations use any single commitment discount per provider, but over-commitment creates paid-for idle capacity.
Spot or preemptible capacity. Suitable for fault-tolerant, interruptible work; unsuitable for latency-sensitive services.
Storage lifecycle and database optimization. Tier or expire old data and review database sizing and replicas.
Egress and network architecture. Keep chatty services close together and reduce cross-region or cross-provider traffic.
Telemetry retention. Sample and tier logs; extremely short retention can weaken incident response.
SaaS rationalization. Remove unused seats and duplicate tools at renewal.
GPU scheduling and inference optimization. Raise accelerator utilization and match model size to task difficulty, while testing quality after each change.
Budgets, anomaly alerts, showback and chargeback. Make cost visible to the people who can change it. Chargeback can create friction if allocation rules are disputed.
Unit economics and architecture reviews. Review the highest-cost services against business outcomes, using the cloud-native architecture patterns appropriate to each workload.
Codifying guardrails with infrastructure as code and sharing cost standards through DevOps practices keeps savings from decaying. The FinOps Foundation reports that many organizations are asked to fund AI through optimization savings, so sustained efficiency also enables growth.
AWS vs Azure vs Google Cloud Cost Decisions
No cloud provider is universally cheapest. The lowest cost for a business depends on workload, region, discounts, existing agreements and architecture, and on how much the team already knows about each platform.
These factors shape the real price difference between providers:
List price versus negotiated price, since large customers often hold private pricing
Committed-use discounts and how well they match steady demand
Enterprise agreements, including existing Microsoft or Oracle software commitments that can change effective prices
Support plan costs, which may be a percentage of spend
Data-transfer patterns, because egress and cross-region traffic differ by architecture
Workload fit, including managed database, analytics and AI services
AI and GPU capacity availability, which can matter more than price per hour
Multicloud complexity, including duplicated tooling and skills
Switching costs, such as data movement, retraining and re-architecture
A fair comparison prices the same workload on each platform with the same availability target, then includes the labor and migration costs of the move. Articsledge's cloud-first strategy guide covers the decision framework, and the cloud market share article explains how each provider's position differs without implying a price ranking. Provider market share does not indicate which platform is cheaper for a particular company.
Cloud Spending Trends to Watch Through 2027 and Beyond
Forecasts point to continued double-digit growth in public cloud, faster growth in AI infrastructure and a rising share of inference. Every figure below is a forecast and can be revised.
Forecast | Figure | Scope | Source |
|---|---|---|---|
Public cloud services spending | Expected to double by 2029; strong double-digit growth through 2029 | End-user spending on public cloud services | IDC, Mar 2026 |
Public cloud services spending | $1.48 trillion by 2029; 21.3% growth in 2026 | Worldwide public cloud services (abstract) | Gartner, Sep 2025 |
AI-optimized IaaS | $66.1 billion in 2027, up 56.5%; inference 59% of the total | Subset of IaaS | Gartner, Aug 2026 |
Sovereign cloud IaaS | $80 billion in 2026; Europe forecast to surpass North America in 2027 | Sovereign IaaS only | Gartner, Feb 2026 |
Worldwide IT spending | $6.37 trillion in 2026, revised up from $6.08 trillion in October 2025 | All IT categories | Gartner, Jul 2026 |
Several operational trends accompany these numbers. Gartner estimates that sovereign IaaS will shift about 20% of current workloads from global to local providers. Oracle said it delivered 850 megawatts of additional data center capacity in its fiscal first quarter, a sign of how fast provider capacity is expanding. FinOps is widening from cloud to SaaS, licensing, data center and AI, with platform engineering teams joining cost conversations earlier, and Flexera's 49% unit economics adoption and 73% hybrid estates suggest that governance will shape outcomes as much as price. Articsledge's cloud computing trends and cloud modernization guides cover the technology side, and AI agents vs SaaS covers software spending.
Methodology and How to Read Cloud Spending Statistics
This article prioritizes primary research and first-party disclosures, and it treats every statistic according to what it measures. Reputable sources can publish different totals for valid reasons.
Source hierarchy. Official releases and abstracts from IDC, Gartner and Synergy; Flexera and FinOps Foundation publications; and company earnings releases and filings. Full Gartner and IDC reports are paywalled, so only public releases and abstracts were used, and some company figures come from earnings material reproduced by news outlets.
Forecasts versus actuals. IDC and Gartner figures for 2026 are forecasts. Synergy's quarterly figures are estimates of reported results, and company results are reported.
End-user spending versus vendor revenue. IDC and Gartner measure what buyers spend; Synergy and the providers measure what vendors earn.
Quarterly versus annual data. A quarter is not multiplied by four to create an annual figure. Synergy's trailing-twelve-month revenue is used instead.
Survey limits. Flexera surveyed 753 cloud decision-makers and users, and the FinOps Foundation surveyed 1,192 practitioners. Both are self-reported and lean toward cloud-active organizations, so they describe respondents and not all businesses.
Overlap. AI-optimized IaaS is within IaaS, IaaS is within public cloud, and sovereign cloud IaaS is a separate Gartner forecast. Categories are not added together.
Geography, currency and rounding. Figures are in US dollars unless stated, fiscal calendars differ among providers, and numbers are rounded as published.
Calculations. Where a figure is derived, such as annualizing a monthly threshold, it is labeled as a calculation.
Figures reflect the newest releases available on October 13, 2026, and later quarterly results will supersede the quarterly figures used here.
FAQ
How much are businesses spending on cloud computing in 2026?
IDC forecasts that worldwide public cloud services spending will exceed $1 trillion in 2026. Individual bills vary widely: Flexera found that 76% of large enterprises spend more than $5 million per month. No universal average exists for all businesses.
How much will be spent on public cloud in 2026?
IDC forecasts more than $1 trillion in 2026, up more than 21%, with SaaS above half of the total. Gartner forecasts $287.3 billion for IaaS alone. The figures differ because they cover different categories.
How much do large enterprises spend on cloud each month?
In Flexera's 2026 survey of 753 respondents, 76% of large enterprises reported spending more than $5 million per month on public cloud. That equals more than $60 million a year if sustained (calculation). It is a survey threshold, not an average.
What is the average cloud spend for a business?
No defensible universal average exists. Spend depends on workload intensity, architecture, SaaS footprint, AI use, data transfer and discounts. Sources publish thresholds and survey distributions, not average bills, so build a budget from your own usage.
How much cloud spending is wasted?
Flexera respondents estimate that 29% of IaaS and PaaS spending is wasted, up from 27%. It is a self-reported survey estimate, not an audited rate, so an individual organization may waste more or less.
Why are cloud costs increasing in 2026?
AI demand is a major driver: Gartner forecasts AI-optimized IaaS up 96%, and Synergy reports generative AI cloud services growing 165%. Variable consumption, pricing complexity, SaaS sprawl and underused discounts add pressure, according to Flexera and the FinOps Foundation.
How much is being spent on IaaS in 2026?
Gartner forecasts worldwide IaaS spending of $287.3 billion in 2026, up 29.3% from $222.2 billion in 2025. AI-optimized IaaS, forecast at $42.3 billion, is a subset of that total and should not be added to it.
How much is AI adding to cloud spending?
Gartner forecasts $42.3 billion of AI-optimized IaaS in 2026, about 14.7% of its IaaS forecast (calculation). Inference is $23.3 billion and training $19 billion. The figure covers AI-optimized infrastructure only, not all AI-related cloud spending.
Which company has the largest cloud market share?
Amazon Web Services leads Synergy's cloud infrastructure services market with 28% in Q2 2026, ahead of Microsoft at 20% and Google at 15%. Market share measures provider revenue, not which platform is cheapest.
How much revenue does AWS generate?
Amazon reported AWS segment sales of $42.2 billion for the quarter ended June 30, 2026, up 37%, with an annualized run rate of $169 billion. This is vendor revenue, not what the average customer spends.
Is Azure more expensive than AWS?
No blanket answer is supported. Cost depends on workload configuration, region, discounts, existing agreements and architecture. Compare the same workload on each platform with the same availability target, including negotiated pricing and migration costs.
Is public cloud cheaper than on-premises infrastructure?
It depends on utilization, workload stability and total cost of ownership, including staff and refresh cycles. Variable or fast-growing workloads often suit cloud, while steady, high-utilization workloads can favor other models. Flexera reports that 73% of respondents run hybrid cloud.
What is FinOps?
FinOps is a discipline in which engineering, finance and business teams manage cloud and technology costs and value together. In the FinOps Foundation's 2026 survey, 98% of respondents manage AI spend and 90% manage or plan to manage SaaS.
How can a company reduce its cloud bill?
Start with allocation and tagging, remove idle resources, then rightsize, schedule non-production systems and match commitments to steady usage. Review storage, data transfer, telemetry and SaaS licenses, and judge results by cost per business outcome.
How should a company forecast cloud spending?
Build bottom-up from usage: baseline workloads, growth-driven variable usage, storage and data transfer, managed services, security, AI, support and contingency, less realistic discounts. Tie each driver to a unit such as customers or transactions, and review regularly.
What will happen to cloud spending in 2027?
Forecasts point to continued growth. IDC expects strong double-digit public cloud growth through 2029, and Gartner forecasts $66.1 billion of AI-optimized IaaS in 2027, up 56.5%, with inference at 59%. These forecasts may be revised.
Key Takeaways
IDC's more than $1 trillion, Gartner's $287.3 billion and Synergy's $143.4 billion measure different things and cannot be compared directly.
AI-optimized IaaS ($42.3 billion) is a subset of IaaS, and with inference at 55% it behaves as an ongoing variable cost.
No defensible average cloud bill exists; the verified benchmark is that 76% of large enterprises spend over $5 million per month.
The 29% waste figure is a respondent estimate, which still signals that allocation and commitments need attention.
FinOps is shifting from savings to value: 98% manage AI spend and 49% use unit economics.
Budgets should be built from usage and judged by unit costs, not invented percentages.
Actionable Next Steps
Export 12 months of billing data and set a clean cost baseline.
Allocate every cost to a team, product or customer through tagging.
Separate steady demand from variable demand.
Remove idle, orphaned and oversized resources.
Model AI workloads and inference separately, with token and GPU assumptions.
Match commitments to steady usage only.
Set budgets and anomaly alerts for each owner.
Adopt two or three unit economics metrics, such as cost per customer.
Review architecture for egress, telemetry and database costs.
Re-forecast every quarter and compare against actuals.
Glossary
Cloud computing: On-demand computing services delivered over the internet with usage-based pricing.
Public cloud: Shared infrastructure run by a provider and rented to many customers.
Private cloud: Cloud infrastructure dedicated to one organization.
Hybrid cloud: A mix of public cloud with private cloud or on-premises systems.
Multicloud: Use of services from more than one public cloud provider.
IaaS: Rented compute, storage and networking.
PaaS: Managed platforms for building and running applications.
SaaS: Applications delivered by subscription.
DaaS: Virtual desktops delivered from the cloud.
FinOps: Managing cloud and technology costs through engineering, finance and business collaboration.
Cloud waste: Spending that produces no business value, such as idle or oversized resources.
Rightsizing: Matching resource size to actual demand.
Reserved instance: A discounted commitment to use specific capacity for a set term.
Savings plan / committed-use discount: A discount in exchange for committing to a level of spend or usage.
Egress: Data leaving a provider's network or region, often billed per gigabyte.
Unit economics: Cost measured against a business unit such as a customer, transaction or inference.
Showback: Reporting cloud costs to teams without charging them.
Chargeback: Billing teams internally for the cloud costs they incur.
AI-optimized IaaS: Gartner's category for cloud infrastructure designed for training and running AI models.
Inference: Running a trained AI model to produce outputs.
Total cost of ownership (TCO): The full cost of a system over its life, including labor, licensing and operations.
Sources & References
IDC. "Global Public Cloud Spending to Surpass $1 Trillion in 2026, Driven by PaaS and AI Platform Adoption." IDC, March 4, 2026. https://www.idc.com/resource-center/press-releases/publiccloudspend2026/
Gartner. "Gartner Forecasts Worldwide AI-Optimized IaaS Spending to Grow 96% Through 2026." Gartner, August 10, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-08-10-gartner-forecasts-worldwide-artificial-intelligence-optimized-iaas-spending-to-grow-96-percent-in-2026
Gartner. "Gartner Says Worldwide Sovereign Cloud IaaS Spending Will Total $80 Billion in 2026." Gartner, February 9, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-02-09-gartner-says-worldwide-sovereign-cloud-iaas-spending-will-total-us-dollars-80-billion-in-2026
Gartner. "Gartner Forecasts Worldwide IT Spending to Grow 14.2% in 2026, Totaling $6.37 Trillion." Gartner, July 27, 2026. https://www.gartner.com/en/newsroom/press-releases/2026-07-27-gartner-forecasts-worldwide-it-spending-to-grow-14-point-2-percent-in-2026-totaling-6-point-37-trillion
Gartner. "Gartner Forecasts Worldwide IT Spending to Grow 9.8% in 2026, Exceeding $6 Trillion For the First Time." Gartner, October 22, 2025. https://www.gartner.com/en/newsroom/press-releases/2025-10-22-gartner-forecasts-worldwide-it-spending-to-grow-9-point-8-percent-in-2026-exceeding-6-trillion-dollars-for-the-first-time
Gartner. "Forecast: Public Cloud Services, Worldwide, 2023-2029, 3Q25 Update" (abstract). Gartner, September 26, 2025. https://www.gartner.com/en/documents/6996966
Gartner. "Gartner Forecasts Worldwide Public Cloud End-User Spending to Total $723 Billion in 2025." Gartner, November 19, 2024. https://www.gartner.com/en/newsroom/press-releases/2024-11-19-gartner-forecasts-worldwide-public-cloud-end-user-spending-to-total-723-billion-dollars-in-2025
Synergy Research Group. "Q2 Cloud Market Passes $143 Billion; Highest Growth Rate in Eight Years." Synergy Research Group (TechInsights), July 30, 2026. https://www.srgresearch.com/articles/q2-cloud-market-passes-143-billion-highest-growth-rate-in-eight-years
Flexera. "Cloud cost management isn't just about cutting waste: What 5 years of Flexera's State of the Cloud data reveals." Flexera Blog, September 2, 2026. https://www.flexera.com/blog/finops/cloud-cost-management-trends/
Flexera. "The new era of cloud: What 2026 data tells us about spend, scale and strategy." Flexera Blog, 2026. https://flexera.com/blog/finops/the-new-era-of-cloud-what-2026-data-tells-us-about-spend-scale-and-strategy
Flexera. "Flexera 2026 State of the Cloud Report" (15th annual, 753 respondents). Flexera, March 18, 2026. https://info.flexera.com/CM-REPORT-State-of-the-Cloud
FinOps Foundation. "State of FinOps 2026 Report" (1,192 respondents). FinOps Foundation, Linux Foundation project, February 2026. https://data.finops.org/
Amazon.com, Inc. "Amazon.com Announces Second Quarter Results." Amazon Investor Relations, July 30, 2026. https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-Second-Quarter-Results
Microsoft Corp. "Microsoft Cloud and AI Strength Fuels Fourth Quarter Results." Microsoft Investor Relations and SEC Form 8-K, July 29, 2026. https://microsoft.com/en-us/investor/earnings/fy-2026-q4/press-release-webcast and https://www.sec.gov/Archives/edgar/data/0000789019/000119312526323632/msft-ex99_1.htm
Outlook Business. "Microsoft Reports Strong Q4 as Azure Crosses $100 Billion, Revenue Rises 18%" (reporting CEO remarks on fiscal 2026 Microsoft Cloud revenue). Outlook Business, July 2026. https://www.outlookbusiness.com/corporate/microsoft-reports-strong-q4-as-azure-crosses-100-billion-revenue-rises-18
Alphabet Inc. "Alphabet Announces Second Quarter 2026 Results." SEC Form 8-K, Exhibit 99.1, July 22, 2026. https://www.sec.gov/Archives/edgar/data/0001652044/000165204426000066/googexhibit991q22026.htm
Anadolu Agency. "Alphabet revenue jumps 24% as cloud growth accelerates on AI demand" (Google Cloud operating income and backlog). Anadolu Agency, July 23, 2026. https://aa.com.tr/en/economy/alphabet-revenue-jumps-24-as-cloud-growth-accelerates-on-ai-demand/4006554
Oracle Corporation. "Oracle: Fiscal 1Q27 Financial Results" (release dated September 10, 2026, as reproduced by Storage Newsletter, September 23, 2026). https://storagenewsletter.com/2026/09/23/oracle-fiscal-1q27-financial-results
