In October 2019, Adobe shut off all its services in Venezuela — overnight, without warning, following a US executive order. No transition period. No alternative. Every Venezuelan business, government office, and freelancer that depended on Adobe’s cloud simply lost access.

Three years later, the pattern repeated at a larger scale. When the Russia sanctions hit in 2022, Microsoft suspended sales, Oracle ceased cloud operations, and SAP halted distribution. Entire countries discovered what dependency on US software means when the political wind changes.

Europe is not Venezuela. It is not Russia. But every European government that runs its administration on Microsoft 365 has accepted the same architecture of dependency — and the same licence terms that make a cutoff technically possible.

Four European institutions decided, at different times and for different reasons, not to wait for that day. Their experiences — spanning two decades, 200,000+ workstations, and very different outcomes — are the most instructive evidence available on what it actually takes to break free from proprietary lock-in on the desktop.

Why This Matters Now: Software Licences as Geopolitical Weapons

For a long time, digital sovereignty was an abstract concept — important but not urgent. That has changed.

In 2022, the Russia sanctions demonstrated what happens when US technology companies shut down services and licences under political pressure. Microsoft suspended the sale of new products and services in Russia. Oracle ceased cloud operations. SAP halted distribution. Adobe had already shut off all services in Venezuela in 2019 — overnight, without warning, following an executive order. GitHub restricted access for developers in sanctioned countries. And when the US placed Huawei on the Entity List, Google was forced to cut off access to Android services — an operating system running on billions of devices.

These cases concern countries that were subject to sanctions. But they establish a structural fact: US software licences are governed by US law, including export control and sanctions provisions — which means licence continuity depends on a foreign jurisdiction’s policy decisions.

The standard licence terms of Microsoft, Oracle, Google, and virtually every US software vendor contain clauses requiring compliance with US export control laws and sanctions regimes. This is not conspiracy theory — it’s written in the EULAs that every organisation accepts upon installation.

An uncomfortable question arises for any organisation whose infrastructure depends on licences governed by a foreign jurisdiction: what happens if the relationship between that jurisdiction and yours changes? The licence terms that enabled shutoffs in Venezuela and Russia exist in every EULA that European organisations have signed. The probability differs; the mechanism is identical.

A caveat is in order: Europe is a NATO ally and the EU is the largest trade partner of the United States — not a sanctioned adversary. US technology companies have an enormous economic interest in maintaining the European market. The probability of a blanket licence withdrawal against EU member states is considerably lower than it was for Russia or Venezuela. But the argument for digital sovereignty does not rest on the most likely scenario — it rests on resilience against unlikely but high-impact disruptions.

That this dependency is now recognised as a problem at the highest political level was demonstrated at the Munich Security Conference in February 2025. Chancellor Friedrich Merz stated in his opening speech: “Nobody forced us into the excessive dependency on the United States in which we recently found ourselves. This lack of autonomy was self-inflicted.” And further: “Competition policy is security policy, and security policy is competition policy. That is precisely why we want to be drivers of progress in future technologies.”

Translated into operational terms: the concentration risk was not imposed — it resulted from procurement decisions that optimised for short-term convenience over long-term resilience. Whether these words will be followed by concrete procurement policy remains open — at the time of the speech, no specific measures were announced.

And anyone who thinks they can simply keep alternatives “in reserve” and activate them in an emergency misunderstands the reality: a migration takes years, not weeks. Schleswig-Holstein is planning for 5–7 years. The Gendarmerie took two decades. If licences are revoked overnight — as happened in Venezuela — and the entire administration runs on Microsoft 365, it’s not just IT that stops. Government stops. A licence cutoff plus emergency migration is a scenario no administration can survive.

The argument, therefore, is not: we should be aware of alternatives. The argument is: we need to have already migrated — or at least be far enough along that a licence withdrawal is a manageable problem, not a total shutdown.

Digital sovereignty doesn’t mean rejecting all US software. It means having reduced the dependency to the point where a disruption doesn’t trigger a catastrophe.

Four institutions — in Munich, Schleswig-Holstein, Paris, and Rome — have attempted exactly that. Their experiences are worth studying in detail.

Munich: A Political Defeat, Not a Technical One

The story begins on 28 May 2003. Munich’s city council votes by a large majority to migrate its 14,000+ government PCs from Windows NT and Microsoft Office to Linux and open source software. It’s a bombshell. Steve Ballmer, then CEO of Microsoft, flies to Munich personally to offer special pricing to Mayor Christian Ude (SPD). Ude declines.

The LiMux project — named after its custom Linux distribution — is implemented over the next decade. By 2013, the migration is essentially complete. The city puts the total cost at approximately €23 million. An internal study concludes that a comparable Windows deployment would have cost around €34 million.

What Actually Went Wrong

LiMux’s technical problems were real but manageable. Maintaining a custom distribution was resource-intensive. Some specialised applications had no Linux version. Compatibility with Microsoft Office documents was imperfect — a problem made worse by external partners who continued to send .docx and .xlsx files.

But LiMux didn’t fail on technology. It failed on politics.

In 2014, Dieter Reiter (SPD) is elected as the new mayor. During the campaign, he had promised to “fix the city’s IT problems” — a thinly veiled signal for a return to Microsoft. In 2016, Microsoft moves its German headquarters from Unterschleißheim to Munich’s Schwabing district. Accenture is hired to evaluate the IT landscape — the same Accenture known as a Microsoft partner. The recommendation: switch back to Microsoft.

In 2017, the city council votes to return to Windows and Microsoft Office. The estimated cost of the reverse migration: between €49 and €100 million — several times the original LiMux investment.

The Lesson

Munich demonstrates that a technically successful migration can be destroyed politically if it lacks institutional anchoring. One council vote can reverse another. A new mayor can overturn an entire city’s IT strategy — if the migration rests solely on a political mandate rather than on institutional foundations.

Since 2020, under a new coalition (Greens/SPD/Volt), Munich has been pursuing a more open approach again. An Open Source Program Office (OSPO) was established. But the lost years and the lost money cannot be recovered.

Schleswig-Holstein: Not Repeating the Mistake

If Munich is the example of how not to do it, Schleswig-Holstein is the attempt to do it right.

In 2021, under Digital Minister Jan Philipp Albrecht (Greens), Germany’s northernmost state adopts a fundamental decision: break free from Microsoft dependency. But not as a big-bang effort — as a systematic, multi-year transformation. The approach is built on six pillars:

  1. LibreOffice replacing Microsoft Office
  2. Linux replacing Windows (in later phases)
  3. Nextcloud for file synchronisation
  4. openDesk for the collaboration platform
  5. Element/Matrix for messaging
  6. Open-Xchange for email and calendar

The Key Difference: Applications First

What distinguishes Schleswig-Holstein from Munich is the sequence. Instead of starting with the operating system — the most visible and emotionally charged change — the state started with applications. LibreOffice first, then email, then the rest. The operating system comes last.

This strategy has a practical advantage: when users are already working with LibreOffice and Open-Xchange, switching the operating system changes almost nothing about their daily workflow. The dreaded “productivity dip” is significantly smaller.

Numbers and Progress

By the end of 2025, the LibreOffice migration was complete on roughly 80 % of the 30,000 workstations. The email migration to Open-Xchange — covering more than 44,000 mailboxes and 110 million emails and calendar entries — was fully implemented in October 2025. The Linux migration is in its pilot phase.

The financial picture is clear: €15 million saved in licensing fees so far, of which €9 million has been reinvested directly into open-source development. In June 2025, the state established a dedicated Open Source Programme Office (OSPO), led by Nicole Beckendorf, to coordinate the strategy long-term — and has begun cross-border collaboration with Thüringen and Estonia (via X-Road).

Political Anchoring

Perhaps the most important factor: the migration survived a change of government. When the CDU replaced the Greens in the state government in 2022, the new CDU Digital Secretary of State Dirk Schrödter didn’t just continue the course — he publicly defended it. In a widely noted letter, he explained why dependency on a single vendor poses a risk to public administration.

This cross-party anchoring is exactly what Munich lacked. In Schleswig-Holstein, the open source strategy is not a question of left or right, but a pragmatic decision about the future of the state’s IT.

Honest About Problems

Schleswig-Holstein doesn’t hide the difficulties. Judges need special document templates that haven’t been fully implemented in LibreOffice. The police union initially resisted. Certain specialised software — CAD systems, forensic tools — will foreseeably remain on Windows.

The state’s response is not “everything must be Linux” but rather: “No new Microsoft licences where alternatives exist.” Pragmatism over dogmatism.

The French Gendarmerie: 20 Years, 103,000 Machines, No Hype

While Munich made headlines and Schleswig-Holstein published strategy papers, the French Gendarmerie has done something that overshadows both: it has carried out the largest Linux desktop migration in Europe — quietly, over two decades, and it’s still running today.

The Timeline

  • 2004: Begins with Firefox (replacing Internet Explorer) and OpenOffice (replacing Microsoft Office)
  • 2005–2007: Further application migrations, internal evaluation of Linux distributions
  • 2008: First GendBuntu installations on workstations (a custom Ubuntu-based distribution maintained in-house)
  • 2014: Approximately 70,000 workstations on GendBuntu
  • June 2024: 103,164 workstations, 97 % running GendBuntu

The numbers are impressive. But the methodology is even more so.

Why It Works

Incremental migration. The Gendarmerie chose the same approach that Schleswig-Holstein would later adopt: applications first, then the operating system. By the time the first Linux desktops were rolled out in 2008, users had already been working with Firefox and OpenOffice for four years. Switching the operating system was merely the final step.

In-house IT team. Lieutenant-Colonel Xavier Guimard, later Colonel, drove the migration with technical expertise and military determination. The internal team maintains GendBuntu, adapts it to the Gendarmerie’s needs, and provides support. There is no external dependency on a distributor.

Military structure. In a military organisation, decisions are binding. There is no mayoral election that can reverse course. When the order is “we migrate to Linux,” the migration happens. This is an advantage that cannot be directly transferred to civilian government — but it demonstrates that the problem with failed migrations is rarely the technology.

Cost efficiency. The Gendarmerie puts the savings at approximately 40 % of the total cost of ownership (TCO) compared to a Windows equivalent. This figure includes not just licence costs but also maintenance, support, and hardware lifecycles — Linux systems can run longer on older hardware.

What You Shouldn’t Conclude

The Gendarmerie migration proves that Linux works at scale. But it would be wrong to treat it as a blueprint for every organisation. The internal IT team is large and highly qualified. The military command structure eliminates political uncertainty. And even the Gendarmerie has kept 3 % of its systems on Windows — for applications that have no Linux version.

The Italian Military: Office Migration Without an OS Change

Not every migration needs to overhaul the desktop down to the operating system. In 2015, the Italian Ministry of Defence began migrating its approximately 150,000 workstations to LibreOffice and the Open Document Format (ODF) — without changing the operating system. The “LibreDifesa” project covers the office suite only.

This approach is less radical but no less relevant. Microsoft Office licences for 150,000 workstations represent a significant cost factor. And switching to ODF as the native document format creates the prerequisite for future freedom of choice — including the operating system. It is exactly the first step that Schleswig-Holstein also took.

The Italian example shows: you don’t have to go the whole way at once. Simply switching the office suite to LibreOffice and the data formats to ODF significantly reduces dependency on a single vendor — and opens doors that were previously closed.

Patterns of Success

The four cases differ in scale, structure, and ambition — but the migrations that worked share a common logic.

Start with the applications, not the operating system. Both Schleswig-Holstein and the Gendarmerie migrated office software and email first, then tackled the OS. Munich did both in parallel, which multiplied complexity and gave critics a broader attack surface. By the time the Gendarmerie rolled out its first Linux desktops in 2008, users had been working with Firefox and OpenOffice for four years. The OS switch barely registered.

Think in years, not quarters. The Gendarmerie took two decades. Schleswig-Holstein plans for five to seven years. Every attempt to compress a migration into a political term has either failed or produced brittle results.

Anchor the decision beyond a single government. Munich’s reversal proves that a city council vote can undo a decade of work. Schleswig-Holstein survived a change from Green to CDU leadership because the migration was framed as pragmatic infrastructure policy, not an ideological project. The Gendarmerie’s military command structure eliminated political uncertainty entirely.

Build internal expertise before you need it. Every successful migration was driven by an internal team that understood the new stack. Outsourcing that competence to consultants — who may have their own vendor relationships — is how Munich ended up with an Accenture evaluation that recommended switching back to Microsoft.

Accept that not everything will migrate. No successful project converted 100 % of systems. Specialised applications that require Windows will persist. Acknowledging this openly, and planning dual-boot or virtualisation solutions, is not failure — it is the professionalism that makes the rest of the migration credible.

And above all: start with open standards. ODF instead of .docx, CalDAV instead of proprietary calendar systems, IMAP instead of Exchange. If your data is in open formats, you have options. If it’s locked in proprietary formats, you have none — regardless of whether the alternative operating system is technically ready.

Desktop Linux in 2026: What Has Changed

The technical landscape has changed substantially since the early days of LiMux.

Wayland has replaced X11 as the default display server in most distributions. This means better HiDPI display support, better security isolation between applications, and more stable graphics performance.

NVIDIA open-sourced its kernel drivers in 2024 — a turning point for desktop Linux, as NVIDIA graphics cards had long been the biggest hardware challenge.

Active Directory alternatives such as FreeIPA and Samba AD have matured to the point where they can serve as replacements for Microsoft’s directory service in many use cases. For organisations not dependent on AD-specific group policies, this is a realistic option.

Printing — long a weak spot — has been significantly improved through CUPS 3.0 and IPP Everywhere. Driver chaos is largely a thing of the past.

And at the policy level, much has changed too. openDesk, the open source workplace environment developed by the Centre for Digital Sovereignty (ZenDiS), reached version 1.0 in 2024. Denmark has launched a feasibility study on Linux migration. And a European citizens’ initiative for “EU-Linux” is gathering signatures.

What Remains

The four case studies don’t tell a simple story. Munich shows that technical success can be undone by politics. Schleswig-Holstein shows it can be done better — with system, patience, and cross-party consensus. The French Gendarmerie shows that Linux works at scale and saves significant money — under the right organisational conditions. And the Italian military shows that even the first step — switching the office suite — can make a significant difference.

For organisations considering a migration, these experiences suggest concrete steps — ordered by urgency:

Start now (weeks, not months):

  • Switch to open document formats. Save new documents in ODF rather than .docx/.xlsx/.pptx. This costs nothing, requires no infrastructure change, and creates the foundation for every subsequent step. If your data is in open formats, you have options. If it’s locked in proprietary formats, you don’t.
  • Inventory your proprietary dependencies. Which applications require Windows? Which use Active Directory-specific features? Which have no Linux-compatible alternative? You cannot plan a migration you haven’t mapped.

Now that Windows 10 has reached end-of-life (since October 2025):

  • Pilot Linux on the workstations that need it least. Microsoft ended security updates for Windows 10 on 14 October 2025. Every organisation now faces a choice: pay for extended support, upgrade to Windows 11 (which requires newer hardware for many machines), or deploy Linux on hardware that still works. For reception desks, kiosk systems, and single-purpose terminals, Linux is the obvious answer — and the pilot gives your team migration experience before it matters.
  • Deploy LibreOffice organisation-wide. This is the single highest-leverage step. Both Schleswig-Holstein and the Italian military started here. Once LibreOffice is the default, the operating system beneath it becomes a replaceable component.

This year (project-level planning):

  • Build internal competence. Every successful migration required an internal team that understood the new stack. Start training now: Linux administration, Samba AD or FreeIPA for directory services, Ansible or Salt for configuration management. The Gendarmerie built its internal team years before the rollout reached full scale.
  • Secure institutional anchoring. A single political decision is not enough — Munich proved that. Seek cross-party consensus, embed the migration in a multi-year IT strategy, or anchor it in an independent body. The migration must survive a change of government.

Accept and plan for (multi-year horizon):

  • Not everything will run on Linux — and that’s fine. No successful migration converted 100 % of systems. Specialised applications that require Windows will persist. Acknowledging this and planning dual-boot or virtualisation solutions is not failure — it is the professionalism that distinguishes Schleswig-Holstein’s approach from Munich’s.
  • Think in 3–7 year timeframes. The Gendarmerie took 20 years. Schleswig-Holstein plans for 5–7. The organisations that succeed are those that treat migration as infrastructure modernisation, not as a political project with a ribbon-cutting deadline.

The question is no longer whether Linux works on the desktop — the French Gendarmerie has proven that for two decades. The question is whether your organisation has begun building optionality before it’s needed. A migration takes years; a licence disruption takes days. The time to start is while the choice is still yours.

Sources

Topic overview: Operating Systems