Windows usage statistics 2026: what the numbers really mean

First, understand what the statistics measure

Market-share figures look precise because they arrive with two decimal places. Their meaning is less absolute. Statcounter says Global Stats is built from aggregated data covering more than five billion page views per month across more than 1.5 million websites. When a browser requests a page, technical signals can indicate the operating system and version in use. Those observations are aggregated into traffic shares.

The important unit is therefore browsing activity, not a unique person, a physical computer, a paid licence or a company seat. A heavily used office laptop may generate many page views. A factory controller, laboratory instrument or offline workstation may generate none. Shared computers, virtual machines, privacy controls, unusual browser strings and unclassified traffic can also affect what appears in the dataset. Statcounter’s own methodology FAQ is essential context for every number below.

For that reason, “Windows represented 56.61% of worldwide desktop web traffic in June 2026” is supportable. “56.61% of all computers worldwide run Windows” is not. The latter silently converts page-view share into an installed-base census. Likewise, the Windows-version chart describes the distribution within measured Windows desktop traffic; it does not tell us how many Windows licences were sold or how many devices exist.

These limitations do not make the data useless. A very large, consistently collected traffic dataset is excellent for showing direction, relative scale and differences between locations. It becomes most useful when paired with operational evidence: your device inventory, application catalogue, support contracts, security posture and user requirements. Good planning starts with the public trend, then validates it against local facts.

Worldwide desktop operating-system share in June 2026

In Statcounter’s worldwide desktop operating-system data, Windows led with 56.61% of measured traffic. “Unknown” was unusually substantial at 21.45%, followed by OS X at 11.89%, macOS at 4.48%, Linux at 4.36% and Chrome OS at 1.21%.

The first conclusion is straightforward: Windows still dominates observed desktop browsing. Its share is more than twice the “Unknown” category and far ahead of each named alternative. That scale matters to software publishers, support teams and organisations selecting broadly compatible business tools. Windows remains the default desktop target for a great deal of commercial software and organisational infrastructure.

The second conclusion requires restraint. “Unknown” at 21.45% is not an operating-system competitor and should not be rebranded as one. It reflects traffic that Statcounter did not attribute to a named platform in this table. Nor should OS X and macOS automatically be merged without understanding the source’s classifications. The safest editorial practice is to reproduce the categories as reported and explain what the dataset can and cannot establish.

For procurement teams, overall OS share is background rather than a buying specification. A design department may have a legitimate macOS workflow; an engineering team may depend on Linux; a school may use Chrome OS; and a line-of-business application may require Windows. Popularity can indicate ecosystem depth, but compatibility, security and total operating cost decide whether a platform fits a particular workload.

Windows 11 is the majority, but Windows 10 is still material

The more consequential 2026 story is happening within Windows. According to Statcounter’s worldwide Windows version chart for June 2026, Windows 11 reached 69.92% and Windows 10 stood at 28.10%. Windows 7 retained 1.67%, while Windows XP, Windows 8.1 and Windows 8 registered 0.17%, 0.09% and 0.03% respectively. As with the OS chart, these are shares of measured web traffic, based on page views rather than a count of installed machines.

Windows 11’s majority position indicates that the migration has moved beyond early adopters. New hardware has shipped with Windows 11 for years, compatible Windows 10 systems have had an upgrade path, and the end of standard Windows 10 support has sharpened the reason to move. For developers and IT teams, testing Windows 11 as the normal production environment is no longer forward-looking; it is basic coverage.

Still, Windows 10’s 28.10% means slightly more than one in four observed Windows desktop page views came from that version. Any organisation serving a broad Windows audience should expect it to remain present. Internally, the number is a prompt to identify why Windows 10 devices remain: incompatible processors, TPM or Secure Boot constraints, a peripheral that lacks supported drivers, a validated application stack, a delayed replacement cycle, or simple project backlog. The public statistic cannot identify the reason; only an inventory can.

Windows 7’s 1.67% is small beside the two current giants, but it is operationally meaningful. Legacy systems often cluster rather than distribute evenly. One business may have none; another may rely on several machines attached to specialist equipment. Traffic data does not prove that these devices are industrial, offline or application-bound, so those explanations must remain hypotheses. What the figure does show is residual online activity that cannot be dismissed as zero.

Windows 8.1 and XP are smaller again, but percentage alone is not a security assessment. An unsupported system with access to sensitive data or a production network can create disproportionate risk. The practical response is not panic or an automatic purchase. It is discovery, isolation where necessary, ownership of exceptions and a funded plan.

Europe, Spain, Germany, France, Italy and Portugal

A worldwide average can hide meaningful regional differences. Statcounter’s June 2026 pages show Windows 11 ahead of Windows 10 throughout the selected comparison: Europe, Spain, Germany, France, Italy and Portugal. Each page is based on browsing activity in Statcounter’s network, not national device registration.

Europe as a whole was modestly ahead of the worldwide figure: 70.86% for Windows 11 versus 26.35% for Windows 10. Germany and France sat above the European Windows 11 level, at 74.61% and 72.34%. Spain, Italy and Portugal were below the European Windows 11 average, although Windows 11 still held a clear majority in each market.

Germany led this selected group for Windows 11 adoption, recording 74.61% Windows 11 and 21.70% Windows 10. France followed at 72.34% and 24.69%. Portugal had the lowest Windows 11 share of the five national markets, 65.13%, and the highest Windows 10 share, 33.62%. Germany’s 52.91-point gap was the widest in the table; Portugal’s 31.51-point gap was the narrowest, while still leaving Windows 11 well ahead.

These differences are descriptive, not explanations. They may reflect hardware renewal, consumer behaviour, organisational migration, traffic composition or other factors, but the chart does not isolate causes. It would be a mistake to conclude that every Portuguese business is further through migration than every Italian one. Country traffic shares are population-level signals; individual estates can differ dramatically.

Regional data is most valuable for setting expectations. A software provider serving these markets should test primarily on Windows 11 without abandoning Windows 10 coverage where customers still use it. A multinational IT team can use the differences to ask better questions: Are country offices following the same refresh policy? Are old devices concentrated in one business unit? Do local applications or procurement cycles explain a slower transition? The answers should come from internal evidence, not assumptions attached to a national percentage.

Gaming moves faster: the Steam comparison

The Steam Hardware & Software Survey for May 2026 provides a useful contrast. Among participating Steam systems, Windows accounted for 93.85%, Linux for 3.99% and macOS/OSX for 2.16%. Across all surveyed systems shown, Windows 11 64-bit had 69.76%, Windows 10 64-bit 23.99% and Windows 7 64-bit 0.07%.

Steam’s Windows 11 result, 69.76% across all participating systems shown, sits close to Statcounter’s 69.92% share within Windows desktop traffic, while Steam’s Windows 10 figure is lower at 23.99%. This apparent proximity must not obscure different denominators. Steam surveys an opt-in population associated with a gaming platform; Statcounter observes page views across a broad website network. Gaming users are more likely to value recent graphics support, current drivers, modern security features and newer hardware. Their upgrade profile is naturally different from that of a mixed population that includes offices, schools, public terminals and legacy business workloads.

The denominators also differ. Steam’s 93.85% is Windows’ operating-system share among participating Steam systems, while Statcounter’s 56.61% is Windows’ share of measured desktop web traffic. The 69.76% Steam figure concerns Windows 11 64-bit across all surveyed systems shown, while Statcounter’s 69.92% concerns Windows 11 within measured Windows desktop traffic. These values should never be added, averaged into a “better” market share or used interchangeably.

Steam’s value lies precisely in segmentation. If you sell games, gaming PCs or enthusiast components, its audience may resemble your customers more closely than general web traffic does. If you manage an accounting firm, it probably does not. Choose evidence by workload and audience. A statistic becomes misleading when it is accurate for one population but presented as universal.

How to use the numbers without letting them drive the decision

Usage statistics are most useful as a benchmark. If your organisation still has 70% of its Windows estate on Windows 10, the gap from the worldwide traffic pattern is a reason to investigate, not proof of failure. Your fleet may contain long-lived equipment, regulated applications or a refresh programme timed around contracts. Conversely, an estate that is 95% Windows 11 can still carry serious risk if the remaining 5% supports critical operations and nobody owns it.

Separate three layers in every report. The first is the external signal: Statcounter or another clearly named source, month, geography, population and metric. The second is your measured state: unique managed devices, active users, application sessions or another denominator you can define. The third is the decision: upgrade, replace, contain or accept temporarily. Keeping those layers visible prevents a public traffic percentage from being mistaken for an internal compliance target.

Trend direction matters too, but this guide deliberately uses the latest complete month rather than a partial July reading. Monthly values can move, and small country-level changes may reflect traffic composition as well as genuine migration. Avoid building a purchasing forecast from one decimal movement. Review several periods when setting capacity or support policy, while using the July snapshot here as a consistent cross-market comparison.

Executives generally need a concise dashboard: supported versus unsupported devices, Windows 11 eligibility, migration wave, exception count, risk owner and expected completion date. Technical teams need the underlying records. Finance needs replacement and ESU costs over the full transition, not just this quarter. Users need notice, tested workflows and a clear support route. The same statistic should lead to different information for each audience, while the source and denominator remain unchanged.

Finally, do not equate prevalence with safety. A widely used release can be misconfigured; a niche workload can be well isolated and controlled. Market data tells you where ecosystems and user populations are moving. Security depends on support status, patching, configuration, identity, network design, backups and operational discipline. Popularity is one planning input, never a compensating control.

Windows 10 after end of support: function is not support

Microsoft ended standard support for Windows 10 on 14 October 2025. Its official support notice makes an important distinction: Windows 10 PCs continue to work, but they no longer receive normal security updates, feature updates or technical support. A device booting successfully in 2026 therefore says nothing about whether its operating system is within ordinary support.

For consumers who need more time, the Windows 10 Consumer Extended Security Updates programme can provide critical and important security updates through 13 October 2026. ESU does not deliver new features or general technical support. With that date only months beyond this article’s 4 August 2026 editorial cutoff, consumer ESU should be treated as a short transition window, not a new long-term lifecycle.

For organisations, Microsoft’s Windows 10 ESU documentation describes coverage for up to three years after end of support, licensed annually with increasing annual cost. Eligibility, deployment and licensing should be verified for the organisation’s exact environment. ESU can reduce security exposure while a controlled migration proceeds; it does not remove the need to modernise, and it does not cure unsupported applications, aging hardware or missing vendor support.

A sensible exception process records the device owner, business reason, data handled, network exposure, compensating controls, ESU status and retirement date. High-risk legacy workloads may need segmentation, restricted internet access, application allow-listing, enhanced monitoring or replacement. The right measure depends on context, but “it still works” is not enough.

A practical Windows audit and migration plan

Start with inventory rather than purchasing. Discover physical PCs, virtual desktops, shared devices, kiosks and specialist machines. Record the Windows edition and version, build, activation status, hardware model, processor, RAM, storage, firmware mode, TPM state, owner, location and last active date. Flag systems that are unsupported, rarely connected or absent from management tools; they are often the devices that surprise a migration project.

Next, test Windows 11 eligibility against Microsoft’s official hardware requirements. The baseline includes a compatible 64-bit processor, 4 GB of RAM, 64 GB of storage, UEFI with Secure Boot capability and TPM 2.0. Meeting minimums is not the same as offering a good working experience, so assess workload needs, battery condition, storage headroom, memory pressure, warranty and expected remaining life as well.

Then map applications and dependencies. Identify the software that matters, who uses it, how it is licensed, whether the vendor supports Windows 11, and whether it depends on browser components, drivers, macros, databases, USB devices, printers or security agents. Test representative workflows with real users. A launch screen is not acceptance testing; printing, export, authentication, integrations and performance may expose the actual blocker.

Classify each device into a decision path:

• Upgrade: hardware is compatible, applications are supported and the user experience will be acceptable.
• Replace: the device fails requirements, is unreliable or costs more to retain than to renew.
• Temporarily maintain: a documented dependency blocks migration and supported ESU or strong compensating controls are available.
• Isolate or retire: risk is unacceptable, the workload is obsolete or no responsible support path exists.

Run a pilot before broad deployment. Include ordinary users and difficult edge cases, define rollback and data-recovery procedures, monitor help-desk demand and measure application failures. Roll out in manageable waves, not one heroic weekend. Keep a live exception register and give every exception an owner and deadline.

Finally, verify outcomes: encryption enabled, security tools healthy, updates flowing, backups tested, applications functional and activation valid. Remove old devices from access systems and management consoles, securely erase storage and retain disposal evidence where policy requires it. Migration is complete only when the legacy risk is closed, not when the new machine arrives.

Licence and procurement guidance: buy the right to use, not just a key

Windows procurement should connect four questions: Which edition does the workload require? Is the hardware suitable? What licence rights are being supplied? What evidence will you retain? A product key is a technical activation mechanism; by itself, it does not prove lawful licence rights, ownership history or transferability. Documentation matters.

Before buying, confirm whether users need Windows Home or Pro capabilities, especially domain or Entra-related management, BitLocker features, remote access and organisational policy controls. Check the base licence on replacement hardware, the permitted upgrade path and any volume-licensing terms relevant to your environment. Avoid selecting an edition solely because its key is cheaper.

Second-hand and perpetual licences can be a cost-conscious option when they are properly documented, appropriate to the territory and workload, transferable under the relevant terms, and supplied by a seller with a clear process. They are not a shortcut around compatibility or lifecycle. A legitimate Windows 10 entitlement does not make an unsupported Windows 10 deployment strategically sound, and a Windows 11 licence does not make incompatible hardware compatible.

Licendi presents documented second-hand/perpetual options for buyers who have completed those checks. You can review the Windows licence category, compare available Windows 11 licence options, or assess Windows 10 licence options for a justified legacy requirement. Availability and suitability must be validated at purchase time.

Ask the supplier for the product description, edition, licence type, transfer documentation where applicable, invoice and activation/support process. Keep those records with the device or entitlement register. For larger estates, reconcile purchases, assignments, reassignments and retirements. The cheapest transaction becomes expensive if it cannot survive an internal audit or support request.

A practical procurement brief should therefore state the business workload, number of devices or users, required edition, deployment method, hardware status, target support horizon, evidence required and budget. This changes the conversation from “find a Windows key” to “provide an auditable, supportable Windows outcome.”

What the 2026 data means for decisions

The headline is not complicated: Windows still leads desktop web traffic, and Windows 11 is now the normal Windows version. The decisions underneath it are more nuanced. Windows 10’s remaining share is large enough to demand compatibility and migration planning, but its post-support status makes passive delay increasingly risky. Small residual shares of older versions can conceal concentrated operational dependencies.

Use Statcounter to understand the broad web landscape, Steam to understand a gaming-oriented population and Microsoft documentation to understand lifecycle and requirements. Do not blend those sources into a fictional installed-base total. Then bring the analysis home: inventory your estate, test applications, confirm hardware, document exceptions and obtain licences with evidence appropriate to your use.

The strongest plan is neither “upgrade everything immediately” nor “keep everything that still turns on.” It is a controlled portfolio of upgrades, replacements and tightly governed temporary exceptions. That approach respects budgets without confusing short-term savings with long-term value.

Frequently asked questions