Engineering blog · community feedback · fixes shipped

What it costs to build an operating system without venture capital.

EuroOS has no investors, no telemetry to sell and, so far, no grant. Three months in, it boots to a desktop, installs to disk and passes 987 host tests. People ask how that is paid for. The honest answer is: mostly by not paying for it. This post puts numbers on that, and explains why our first funding goal is a deliberately small one.

The numbers, as of this week

Everything below comes straight out of the repository, the web server logs and the download mirror. No rounding up.

That last line is the reason for this post.

What "free" actually costs

An OS project has three kinds of cost, and only one of them is money.

1. Machines

EuroOS needs a build server that can compile a hundred thousand lines of Rust and run the test suite on every push, a host that can spawn QEMU virtual machines on demand for the live-try page, and a mirror that serves gigabytes of images. Today all of that runs on hardware we already own for other work, in Antwerp. That is why the bill is near zero, and also why it is fragile: it is borrowed capacity. At list prices, renting the same thing is in the range of 300 to 400 euro a month.

But the part that actually holds the project back is not the servers. It is physical test hardware. A kernel that only ever boots inside QEMU learns nothing about the real world: GPUs with their own firmware, Wi-Fi and Ethernet chips that behave differently from the virtio devices an emulator hands you, USB controllers, NVMe drives that lie about their queues, laptops with three different sleep states. Every one of those needs a real machine on a real bench, and every vendor family needs its own. Right now we test on the two laptops we happen to own. Broader hardware support for EuroOS, the thing people ask about most, is limited less by code than by the number of boards on the table.

So the first goal, 500 euro a month, is split in two: keep the borrowed infrastructure from disappearing, and buy, one at a time, the boards and cards needed to make GPU, networking and storage support real instead of emulated. Add the code-signing certificate a bootloader needs to be taken seriously by firmware, and the domains in 24 country variants, and 500 is what it comes to.

2. Time

520 commits in 12 weeks is roughly six a day, every day, weekends included. That is not a side project; it is a full-time job that nobody is paid for. We are not asking anyone to fund salaries yet. We are asking that the machines stop being borrowed.

3. Independence

This one is the point of the whole project. A European operating system that is financed by an American cloud credit programme, or by a VC that needs an exit in five years, is not sovereign in any sense that matters. The way to stay independent is boring: many small contributions from people who want the thing to exist, plus European public funding where it fits. We are applying for the second. This post is about the first.

Why the goal is only 500 euro a month

Because it is real. 500 euro covers the infrastructure and the test hardware described above, nothing else, and every board we buy with it will be listed here with its invoice. A goal of 5,000 would look more ambitious and sit at 2% for a year, which tells supporters their 10 euro does not matter. At 500, twenty-five people at 20 euro finish it. Your contribution moves the bar visibly the same day.

When the first goal is met, the next one will be equally concrete: one paid development day a week, so that the two of us can stop doing this only at night.

How to help

The live funding bar is on the home page. We will update these numbers here every quarter, including what the money was spent on.