Can we do without it after all? Bavaria withdraws the Microsoft 365 contract worth almost 1 billion euros and creates sovereign jobs. What can be learned from LiMux, Wienux, Schleswig-Holstein, France and Denmark - and which eight levers are decisive this time.
On 27 May 2026, the Bavarian Ministry of Digital Affairs officially confirmed what had already been rumored: the planned framework agreement with Microsoft for the nationwide introduction of Microsoft 365 in the state administration will not be concluded. Instead of investing almost a billion euros in US software licenses, Digital Minister Fabian Mehring (Free Voters) is now building a "sovereign basic workplace" - with solutions from the Center for Digital Sovereignty (ZenDiS), Bavarian in-house developments from the BayernCloud school and open source components. This brings Bavaria back into a debate that the state has already lost once: Munich buried the LiMux project in 2017, Vienna rolled back "Wienux" from 2009. This time, the State Chancellery and Ministry of Digital Affairs must consistently incorporate the lessons learned from both defeats - and from today's successful projects in Schleswig-Holstein, France and Denmark - into the migration design.
The original plan of the Ministry of Finance under Albert Füracker (CSU) envisaged a framework agreement with Microsoft worth just under one billion euros. It would have covered the gradual rollout of Microsoft 365 to the entire Bavarian state administration. Digital Minister Fabian Mehring publicly rejected this contract and - after weeks of escalation in the state parliament and in the ministries - has now prevailed with a counter-model: an internal pilot operation in his own ministry, which is to operate around 40 out of 200 workstations on a sovereign stack by the end of March 2027. Heise describes the technical concept in detail, while PC-WELT confirms the financial dimension.
Mehring justifies the move with two sentences that can now be quoted for the procurement logic in every federal state: "Because digital infrastructure has long held the world together at its core, we need to make it crisis-proof" and "We no longer have time to discuss the importance of digital sovereignty in a cheap way". Translated, this means that sovereignty is no longer an academic simulation game, but a procurement task.
If a CSU-led federal state stops a billion-euro Microsoft contract at the last minute, then the business basis for all German IT procurement has shifted. The issue is no longer open source activism, but very practical risk management.
In 2003, the Munich City Council under Mayor Christian Ude (SPD) decided to migrate all of the city's workstations to Linux. The reason for this was Microsoft's announced end of support for Windows NT 4.0 - Ude later described the experience in Linux magazine: "As a customer with a five-figure number of devices, they simply gave us the choice: Eat or die." The LiMux project was a direct reaction to this lock-in.
By 2012, 12,600 of 15,500 municipal desktops had been migrated to LiMux. The city put the cumulative savings compared to the Microsoft path at around 11.7 million euros and emphasized the gain in strategic decision-making freedom. Internationally, Munich was regarded as the most visible example of how a large administration can productively and completely switch to open source.
In 2017, the city council under Lord Mayor Dieter Reiter (SPD) nevertheless decided to return to Windows and Microsoft Office by 2020, based on an Accenture report from 2016 that recommended the gradual dismantling. The critical review - by Heise and others - shows that the LiMux project did not fail due to technical limitations, but because of an antiquated IT operating model, a lack of political defense and a report prepared in camera by a consulting firm with significant Microsoft business.
In 2020, the new green-red Munich city government once again decided to switch to open source. The second attempt is now effectively five years old - but without the symbolic boost that the original LiMux project had for the whole of Europe.
Parallel to the Munich LiMux project, the City of Vienna developed its own Debian/KDE distribution for its 16,000 administrative workstations from 2004 onwards. The goal was a voluntary switch of up to 4,800 computers to Linux and 7,500 computers to OpenOffice. It was precisely this voluntary nature that was the first systemic design flaw.
In 2005, around 1,000 Wienux installations were running; by 2009, this had fallen to 280. The city migrated 720 PCs back to Windows Vista because central language support software only ran on Windows and had to be introduced under time pressure. In 2009, Vienna bought Windows licenses for one million euros - the end of the Wienux line in an operational sense. Anyone who cites the project as a success has not read a magistrate's report for ten years.
The lesson from Vienna is not "Linux doesn't work in administrations" - but rather: Those who volunteer migration make any specialist specialist application a killer argument. Sovereignty requires a political commitment that Vienna has never dared to make.
The lessons learned: Munich failed because the next political majority no longer wanted to defend the project. Vienna failed because it was never declared mandatory. Bavaria is now facing both failures - that is the opportunity.
Unlike in 2017, when LiMux was dropped, there are three ongoing migration programs in Europe in 2026 whose figures can no longer be ignored.
Schleswig-Holstein has brought the proportion of LibreOffice workstations in the state administration to around 80 percent by December 2025; Microsoft Office and Outlook are being actively uninstalled on these workstations. Sharepoint and Exchange will be replaced by Nextcloud, Open-Xchange and Thunderbird (with Univention-AD-Connector). According to its own press release, the state is already saving "more than 15 million euros" per year in licensing costs; a further nine million euros in one-off investment has been budgeted for migration and open source further development in 2026. Step two is the replacement of Windows with Linux.
France is driving forward the largest government migration in the western world. The Gendarmerie Nationale has been running its own Ubuntu derivative GendBuntu on over 100,000 workstations for almost twenty years; the annual license savings amount to around two million euros and the total cost of ownership per workstation has fallen by around 40 percent. In April 2026, the government decreed that each ministry must submit a migration plan for around 2.5 million administrative workstations by fall 2026 - with a clear goal: all workstations in the state administration to open source by 2029. Not only the office will be migrated, but eight areas at the same time: end devices, communication, antivirus, AI, databases, virtualization, cloud and network.
Denmark is taking the municipal and ministerial lead. Since mid-2025, the Danish Ministry of Digital Affairs has been running a complete Microsoft exit - with Linux and LibreOffice as the standard. The two largest cities, Copenhagen and Aarhus, have announced the same path; Copenhagen's Microsoft license costs increased by 72 percent between 2018 and 2023 (from 313 to 538 million kroner). Aarhus is already running Nextcloud productively and reports around 70 percent lower operating costs in the converted areas.
What these three programs have in common is what Munich and Vienna lacked: a binding end-state date, a political mandate across legislatures, a layered migration from the office to the operating system to the collaboration platform - and the knowledge that license savings must only be the result and not the main motive.
Schleswig-Holstein shows that a state administration with serious political backing can move from full Microsoft dependency to 80 percent LibreOffice in five years - without administrative downtime and with a positive license balance. This is now the benchmark for Bavaria.
As soon as the Bavarian decision is cited in any IT discussion, the counter-question comes. Niklas Hanitsch formulates it as a representative in a widely shared LinkedIn Pulse article: "Hyperscaler exit - everyone talks about it, no one is taking it." Hanitsch quotes Forrester: "No European enterprise will shift entirely from US hyperscalers in 2026." He cites four arguments that Bavaria, Kiel, Paris and Copenhagen do not dispel: firstly, the public sector has hardly placed any budget with European providers to date - AWS, Microsoft and Google continue to share over 70 percent of the market share in Europe, with European providers accounting for around 15 percent. Secondly, Nvidia dominates the AI GPU market with around 85 percent; without an AI stack, there is no exit. Thirdly, egress fees and proprietary APIs make migrations technically expensive. Fourthly, true sovereignty requires 15 to 20 years of consistent policy over several legislative periods - it took Airbus 33 years to overtake Boeing.
This analysis is correct as a snapshot - which is precisely why it is the wrong strategic basis for 2026. Three findings shift the picture:
Gartner itself expects a trend reversal: sovereign cloud IaaS spending is expected to rise to 80 billion dollars worldwide in 2026, an increase of 35.6 percent compared to 2025. By 2030, Gartner predicts that over 75 percent of companies in Europe and the Middle East will geopatriate their virtual workloads - up from under five percent in 2025. That's not a hyperscaler confirmation, that's a shift forecast in the double-digit quarterly range.
According to the latest surveys, 86% of CIOs are planning at least partial cloud repatriation - the highest figure ever measured. 40 percent of companies want to operate hybrid architectures in business-critical workflows by the end of 2026 (compared to eight percent previously). The movement is underway - but not as a "full exit", but as a controlled workload shift.
Bavaria's billion-dollar freeze, France's 2.5 million-seat order, Schleswig-Holstein's 80 percent quota and Denmark's ministerial exit are not isolated cases, but the result of Trump 2.0, the CLOUD Act, executive orders and a Palantir manifesto that officially defines US software as "hard power". These strategic triggers simply did not exist in the 2024 Forrester data.
Why Gartner and Forrester nevertheless arrive at their "status quo stable" reading can be explained methodically: both companies primarily measure ongoing contract volumes and CIO surveys with a twelve-month horizon. The installed Microsoft 365 base continues to run in precisely this window - even in the four flagship migrations mentioned. What the models systematically underestimate are, firstly, procurement decisions that are made today and will take effect in 24 to 36 months (Bavaria, France), and secondly, the regulatory levers (EU AI Act, BSI C3A, NIS-2, DORA), which are not yet fully effective in the reporting period.
Hanitsch is right on one point that is crucial for Bavaria: the hyperscaler exit will only become a reality when the answer to "Where to?" ceases to be theoretical. This is precisely the mission of the Bavarian pilot project - not to complete the full exit, but to make the sovereign target image operational and procurable. If you don't want to be among the 75 percent predicted by Gartner as geopatriated in 2030, start with the roadmap in 2026. This is exactly what Bavaria has just decided.
The strength of Bavaria's move lies in the fact that the political decision was made before the contract was signed - not afterwards. To prevent the pilot from becoming a second LiMux, eight concrete levers are needed for further implementation:
SecureCloud operates its platform 100% in Germany on its own hardware at noris network AG, BSI C5:2020-certified, with key management in Europe, Bring Your Own Key on request and exclusively open source core components. We thus deliver exactly what Bavaria has defined for its sovereign basic workplace: productive cloud, mail and collaboration modules without CLOUD Act exposure, without FISA-702 access and without vendor lock-in.
For us, the decision against the Microsoft framework agreement is not a political signal, but a long overdue standardization of procurement. If the largest federal state stops a contract worth billions for reasons of sovereignty, the arguments of "no alternative" and "extreme position" lose their basis in any private-sector cloud discussion - from now on, they will only follow the procurement line of a CSU-led state.
Bavaria has stopped the most expensive Microsoft contract in German administrative history and at the same time taken on the most difficult task: to build a migration in such a way that it survives LiMux and Wienux. This cannot be achieved through pilots alone, but only through a binding end-state date, a non-partisan majority in the state parliament, a stack-wide migration, a real TCO calculation and a dedicated change management budget. The templates are on the table in Kiel, Paris and Copenhagen - Bavaria just needs to copy them consistently. Sovereignty begins with a decision. This time, it was made before it was signed. This is precisely where the opportunity lies.