From Washington: In a development with direct implications for Australian businesses and IT departments, new data shows Microsoft's decade-long push to retire Windows 10 is finally bearing fruit — though the picture behind the headline figures is considerably messier than Redmond would prefer.
Statcounter's latest figures, covering the full month of February 2026, put Windows 11 at 72.57% of the global Windows desktop market, with Windows 10 now accounting for just 26.45%. As recently as December 2025, Windows 11 held only a slim majority at 50.73%, meaning the newer operating system has gained roughly 22 percentage points in just two months. The Register, which first reported on the Statcounter figures, noted the numbers appear consistent with Microsoft's own claim from earlier this year that Windows 11 has surpassed one billion users globally.
The surge is not, by most accounts, a story of consumer enthusiasm. Microsoft ended free mainstream support for Windows 10 in October 2025, and the financial arithmetic since then has been unambiguous. Enterprises that want to keep security patches flowing must enrol in the Extended Security Updates (ESU) programme. Microsoft's own documentation confirms the ESU programme launched in November 2025 on an annual subscription basis, with pricing starting at US$61 per device for the first year and doubling each subsequent year. For organisations managing hundreds or thousands of endpoints, the sums add up quickly.
Esben Dochy, Principal Technical Evangelist at IT asset tracking firm Lansweeper, told The Register that few organisations are eager to pay for ESU licences, but migration is rarely as simple as pressing an upgrade button. "Many organizations will migrate, but a non-trivial subset will rely on ESU as a safety net because their constraints are less about 'deciding to upgrade' and more about validating dependencies and coordinating operational downtime," he said. That observation is particularly relevant for Australian enterprises running legacy software tied to specific hardware configurations, where vendor certification processes can stretch migration timelines by months.
The server side of this story is, if anything, more concerning for IT administrators. According to Lansweeper's monitoring data, Windows Server 2016 still accounts for 20.3% of all servers it tracks. Microsoft has announced that Windows Server 2016 will also reach end of support, and an ESU programme will eventually apply there too. Dochy noted that server migrations carry far greater operational risk than desktop upgrades, since any downtime affects entire services rather than individual workstations.
There is also a niche but telling story in the Windows 10 Long-Term Servicing Branch (LTSB) figures. The 2016 LTSB release, designed for specialised deployments such as point-of-sale systems, kiosks, and embedded industrial equipment, represents just 0.5% of all Windows devices but makes up 19.8% within the LTSC/LTSB category. These are systems where upgrade gates are set not by IT preference but by vendor certification, peripheral driver support, and tightly scheduled change windows.
It is worth being clear-eyed about what Statcounter's data can and cannot tell us. The figures are derived from tracking code installed on more than 1.5 million websites globally, meaning they reflect web-browsing activity rather than a direct census of installed operating systems. Business and embedded systems are less likely to generate web traffic, so the numbers almost certainly overstate Windows 11's penetration in enterprise and industrial environments. Statcounter also revises its figures as more data arrives.
For Australian organisations, the calculus is straightforward even if the execution is not. Paying US$61 per device in year one, US$122 in year two, and US$244 in year three to stay on an unsupported operating system is, at scale, a far more expensive proposition than funding an orderly migration to Windows 11 or Windows Server 2025. The Australian Signals Directorate has long listed prompt patching and supported operating systems as core elements of its Essential Eight cybersecurity framework, and running unpatched or ESU-dependent systems creates compliance headaches that extend well beyond the licence bill.
The fair case for going slowly is real, though. Organisations running critical infrastructure, healthcare systems, or manufacturing controls cannot simply accept the risk of a botched upgrade mid-operation. Regulatory obligations, vendor support contracts, and the genuine complexity of dependency mapping all make the IT manager's job harder than external observers tend to appreciate. Dismissing the laggards as negligent misses the point; many are stuck because of structural constraints, not indifference.
The honest read of these numbers is that Microsoft's deadline strategy is working exactly as designed: financial pressure is shifting behaviour where gentle encouragement failed for years. Whether that pressure lands equitably on all types of organisations is a different question. Small businesses without dedicated IT staff face the same ESU pricing as large corporations with dedicated migration budgets, and the asymmetry matters. A pragmatic policy response from government would involve clear guidance and, potentially, targeted support for critical sectors where legacy dependencies are highest and upgrade resources are thinnest.