The Dutch government has a Linux desktop on trial. It can't say what it pays Microsoft.
Staff in a handful of Dutch municipalities are working on laptops that had already been written off. They now run DAWO, a Linux desktop that started inside the Dutch government, in small, closed pilots; in at least one municipality the elected council has joined the trial.
On 21 September, the state secretary responsible for digital policy wrote to parliament about the national digitalisation strategy: "At the same time I note that no additional funds are available".
Both are true. Over the past few weeks, most people heard only one of them. Much of the international coverage ran with the first and told it as the Netherlands leaving Windows. Dutch critics pointed to the second: no money, no prospect of money, no plan. Read the primary sources side by side and a third story appears, one that has little to do with Linux and a lot to do with bookkeeping.
What DAWO is
DAWO (Digitaal Autonome Werkomgeving Overheid, the government's digitally autonomous work environment) started as an Interior Ministry initiative, built behind the scenes by a small core team, and is now developed as an open community project. The work began in early 2025, after the United States imposed sanctions on the International Criminal Court in The Hague. As Victor Gevers, one of the core team, sums it up for Tweakers, the government realised that access to many essential services depends on the policy of the US government.
DAWO is a blueprint for a whole workplace, not just an operating system: the desktop, collaboration tools, cloud services, logins and access rights, AI, and the tools to manage every laptop from one place. Several of those parts, logins and cloud among them, are still placeholders.
The desktop runs on NixOS, a Linux distribution with Dutch roots that no company controls. According to Tweakers, the team tried openSUSE and then Fedora first, both tied to commercial owners, before settling on NixOS. The project still keeps a Fedora-based track as a fallback. NixOS has a practical advantage as well: machines are built from shared, written recipes, so a configuration that works in one office can be copied to the next. "Eighty to ninety percent of your Nix code is reusable", says core team member Rutger Putter.
Alongside the desktop runs MijnBureau, a collaboration suite that staff use in the browser and that the government runs on its own servers. It is developed at the Interior Ministry and built from open-source parts, many of them European, including components from France and Germany such as Nextcloud, a German file-sharing and collaboration platform. Since this spring, four municipalities ('s-Hertogenbosch, Zaanstad, Ede and Amsterdam) have been testing a Linux workplace built on DAWO, with MijnBureau, together with their association, VNG, in a trial that runs until the end of 2026. When the trial launched, VNG reported the first findings: the technical basis is there, but more cooperation and investment are needed.
In July, central government turned the idea into an assignment. A cross-ministry committee tasked the state's three large IT service providers, SSC-ICT, DICTU and DUO-ICT, with developing a standardised, more sovereign digital workplace for central government, building on existing initiatives such as DAWO and MijnBureau.
What DAWO does not have is a date. The developers told Tweakers there is no formal timeline and no implementation plan with deadlines. The desktop's latest release is 0.1.3, with 0.2.0 in testing. Version 1.0 is hoped for next year, and the community will decide by consensus when it is ready.
How "alternative" became "replace"
The Tweakers interview that much of the international coverage drew on is careful. Its headline calls DAWO a sovereign alternative to Windows and Office. The developers describe themselves as pragmatists. "We are not hardliners", says Putter, and according to Gevers the team built a graphical admin console so that ordinary Windows administrators can come along, rather than expecting them to learn the command line.
A few rewrites later, the headlines read "Dutch government to replace Microsoft Windows with NixOS-based operating system" and "Goodbye, Windows: The Netherlands Is Building Its Own Linux for Government". The team itself still has no timeline.
None of this makes DAWO less interesting. It does mean that a decision-maker who repeats "the Dutch are leaving Microsoft" in a board meeting is repeating a decision the Dutch government has not made.
No new money
The government's own documents are clear on the money. The 2027 budget moves digital policy-making from the Interior Ministry to Economic Affairs, and with it about 58 million euros a year, staff and running costs included. That is money moved, not money added. The new policy budget acknowledges that not all its ambitions show up in its figures, partly because most of the money for carrying them out stays with the Interior Ministry. Responsibility is split accordingly: Economic Affairs coordinates and sets the framework; the Interior Ministry carries it out.
The 21 September letter is plainer. The actions it lists will be fitted into existing budgets "or the ambitions will be adjusted". On cloud: without extra central funds, the government cannot put its own sovereign cloud into production; it can only prepare for it. The new Dutch Digital Service, set up partly to reduce the state's reliance on outside IT suppliers, runs largely on one-off money and staff moved over from the Interior Ministry, with no permanent budget planned.
To be fair, the budget does contain new money framed around European autonomy: a one-off 120 million euros for Dutch companies to join a German-led European AI project, and 18 million, added by parliament, for seabed surveys toward a Dutch landing of an Arctic subsea cable. Neither is aimed at making the government's own IT less dependent on a handful of suppliers.
The figure nobody had
On 30 September, the parliamentary committee for digital affairs debated the subject. One MP cited an estimate, reported by the Dutch newspaper AD, that all Dutch public bodies together spend two to three billion euros a year on Microsoft. Another pressed for the actual figure.
The state secretary responsible for central government IT could not say. The government does have insight into the large contracts, came the answer, but "because we are nowhere near being that one government", there are different registers, different dashboards and different reports, and no single budget holder for digital affairs. And: "We simply don't yet have the capacity to get that administration clearly mapped" (uncorrected transcript, my translation).
Days earlier, in writing, the government had repeated one combined figure for eight large vendors whose software central government buys jointly: 230 million euros in 2024. Asked for Microsoft spending per ministry, it declined to break the figure down, citing its negotiating position. That figure covers only central government's joint purchasing, so it cannot be compared with the estimate for the whole public sector.
The MP who pressed hardest put the consequence plainly: if there is no new money, you have to allocate smarter, and you cannot allocate if you don't know what you spend now.
What leaving costs, and what staying costs
The people building the exit can describe its costs. "Open source doesn't cost less, it costs differently", Putter told Tweakers. Licence payments go away, and the money moves to hosting, to salaries for in-house administrators and developers, and to supporting the open-source communities that fix the bugs. Some costs fall: DAWO runs on old laptops, so the pilots needed no new hardware.
Where a government has counted both sides, the comparison can be made. Schleswig-Holstein, the German state moving its administration to open source, said in December 2025 that it already saves more than 15 million euros in licence costs, against a one-off 9 million in 2026 to migrate and further develop its open-source tools.
In The Hague, the cost of staying has no total. It is spread across ministries and renewal cycles, and nobody adds it up. The two sides of the decision also arrive differently. Leaving shows up as new line items that someone has to request, defend and sign for. Staying hides inside budgets that renew themselves every year. A pilot that works can keep losing that comparison, because only one side of it is ever on the table.
Try it on your own books
Most organisations are closer to the Dutch government than they would like to think. To put a number on your largest software vendor, you usually need to gather:
- the central licence agreement, including the add-ons folded into each renewal;
- licences and subscriptions bought directly by departments, often on their own budgets;
- cloud consumption billed to projects rather than to IT;
- the partners and contractors whose work exists mainly to run, integrate or license that vendor's products.
If that takes a week of chasing finance, the gap is the finding. Without that number, any exit plan is arguing against a blank.
My read
Software is no longer the main obstacle. DAWO is early, and parts of it are still placeholders; one of them, the government's own sovereign cloud, is waiting on money. But in small pilots a handful of people have put together a government desktop from existing open-source parts, running on written-off hardware, and the municipalities testing it report that the technical basis is there. What decides the rest of Europe's sovereignty push is duller: whether anyone can put a number on the cost of staying, and who signs for the cost of leaving.
