LibreOffice made EUR 2.18 million in 2025 and kept EUR 719,000 of it. Then it added a donation banner.

Quick answer (September 2026). The Document Foundation, the German charitable foundation behind LibreOffice, took in EUR 2,175,997.33 in 2025, after EUR 1,563,879.77 in 2024, EUR 1,398,933.01 in 2023, EUR 1,308,437.51 in 2022 and EUR 1,407,924.29 in 2021. Donations alone were EUR 1,976,825.25 in 2025, up 42.1% on 2024 and the largest single-year increase in cash terms in the foundation's history. It spent EUR 1,457,343.06 and kept EUR 718,654.27, a surplus worth 33.0% of revenue and 7.4 times the EUR 96,530.22 it kept in 2024. Seven months into 2026 it had already banked another EUR 448,647.47. Every figure here comes from the foundation's own monthly accounting ledgers.

Which makes the timing of what happened next genuinely strange.

In August 2026, LibreOffice 26.8 shipped with a donation banner in the Start Centre, the first time in the project's fifteen-year history that the software has asked its users for money from inside the application. The foundation's explanation, published two days later on 28 August, describes an organisation that is "significantly underfunded."

It had just closed the best year it has ever had.

Both of those things are true at once, and the ledgers explain why. They also contain two lines that do not match the summary the foundation published alongside them.

LibreOffice revenue, 2015 to 2025

The Document Foundation Stiftung, Berlin. All figures in euro, as first reported in each year's December income statement.

Source: the December income statement for each year, published as an ODS file on the foundation's public accounting ledgers page. "Total revenue" adds donations, commercial revenue and interest or exchange gains. The 2022 total is depressed by an exchange loss of EUR 9,131.

That table has not existed anywhere before, and the reason is slightly absurd: the foundation's thirteen annual reports publish the income and expense split only as percentage pie charts. The euro amounts are in the ledgers, which are published monthly, in ODS, translated from German, and read by approximately nobody.

Read the donation column and the shape is not a growth curve at all. It is a step, a plateau, and a breakout.

The step is 2020. Donations jumped 50.3% in a single year as people who suddenly needed a full office suite at home went and installed one. Then came the plateau: 2021, 2022, 2023 and 2024 land at 1.39m, 1.29m, 1.30m and 1.39m. Four years, and 2024 finished essentially level with 2021. In the middle of it, 2022 actually fell 7.3%.

Then 2025 added EUR 585,636 of donations in a single year, equal to more than three quarters of everything the foundation raised in the whole of 2015.

The surplus is the part nobody mentions

Revenue is the headline number. The surplus is the one that makes the banner interesting.

The foundation's spending in 2025 was EUR 1,457,343.06, the statement's own total expenditure line. The 2024 statement has no equivalent total, but summing its expense lines gives EUR 1,467,349.55. That is a decrease of 0.68%. Revenue rose 39.1% and spending did not move.

So the EUR 718,654.27 surplus is not the product of a windfall being carefully husbanded. It is what happens when income breaks out and the cost base stays exactly where it was. It is the largest surplus in the foundation's history in absolute terms, and at 33.0% of revenue it is a margin most commercial software companies would be pleased with.

2026 is running hotter. Through 31 July, seven months, the foundation reports EUR 1,448,795.52 of revenue against EUR 1,000,148.05 of expenditure, for a surplus of EUR 448,647.47. Donations alone at EUR 1,271,288.21 are already 64.3% of the full-year 2025 total in 58% of the year.

The foundation is not hiding any of this. Its own funding post says the surplus "is not left unused but is reinvested in the growth already underway," names three developers hired during 2026 with two more to come, and explains that recruitment takes months so income arriving before hiring completes shows up as a surplus that "represents planned work rather than unspent money." It also states a reserves target of at least one year of operating costs.

That is a reasonable position. It is also a different story from "underfunded," and both appear in the same document. The honest version is narrower and more interesting: the foundation has more money than it has ever had, and less money than the roadmap it has committed to for web, cloud and mobile requires. Those are compatible. They just are not the same sentence.

The app store line does not say what the foundation says it says

Here is the first place the ledger and the summary diverge.

LibreOffice is free. It is also sold, by the foundation itself, on the Mac App Store and the Microsoft Store, where the identical software costs money. The foundation's explanation for this is that the price "is set to cover the costs of maintaining the applications on those platforms."

The 2025 statement books that activity as a separate commercial sphere, and it looks like this:

That is an 84.6% margin, and it is the foundation's only taxable profit line for the year.

The fair reading matters here, so: this is sphere accounting, not concealed profiteering. The cost of actually developing LibreOffice sits in the charitable sphere, because the software is built for everyone regardless of where they get it, and none of it is allocated against the store revenue. Under German charitable accounting that is normal and probably correct. But "set to cover the costs of maintaining the applications on those platforms" describes a cost-recovery price, and the books record an 85% margin channel. Both cannot be a complete description.

The channel is also growing far faster than donations. App store revenue ran between EUR 11,000 and EUR 38,000 a year from 2015 to 2022, then went to EUR 94,292 in 2023, EUR 148,750 in 2024 and EUR 167,708 in 2025. That is a 5.3x increase in three years against a 54% increase in donations over the same period.

And the split inside it is the detail worth stealing. In 2025 the Mac App Store paid EUR 118,942.20 and the Microsoft Store paid EUR 35,392.91. Apple is 3.36 times Microsoft. Microsoft Store revenue actually fell 6.1% year on year while Apple rose 37.8%.

The foundation's own 2024 annual report measures the audience on the other side of that: 82% of visits to libreoffice.org came from Windows machines and 6% from macOS. Website visits are not users and not downloads, so treat it as directional rather than exact. Directionally, though, the smaller platform is paying several times more, and any operator who has ever argued about which desktop platform to build for first should sit with that for a moment.

Freelancers have out-cost employees for three years running

The second divergence is about who actually does the work.

The funding post says that development contracts used to be tendered out to companies, but "this strategy has changed, and it was decided to bring development expertise in-house." The account-level ledgers show the cash going the other way.

Freelancers have been the majority of the foundation's people spend for three consecutive years, and 2025 is a higher share than 2024, not a lower one. The employed payroll was essentially flat between 2024 and 2025, down 1.6%, while the freelance bill rose 8.8%.

One caution, because it nearly caught me. The 2025 statement changed format, moving to a Section 4(3) EStG sphere overview, and it splits wages, statutory social security and freelancers onto separate lines for the first time. Comparing the 2024 "employee expenses" line of EUR 508,359, which already includes social security, against the 2025 "wages and salaries" line of EUR 406,736, which does not, produces an apparent 20% payroll cut that did not happen. Adding the 2025 social security line of EUR 93,244 gives EUR 499,980 and the real answer, which is flat. The freelancer line was tracked separately in 2024 too, inside account group 2715 to 2722; it simply was not broken out on the summary statement.

"In-house" is presumably about direction of travel and about control over the roadmap rather than about this year's cash. But if you are reading the foundation's spending to work out who writes LibreOffice, the answer in 2025 is still: mostly people who do not work there.

The gap the foundation measures itself against

The comparison the funding post reaches for is Thunderbird, and it is worth checking because it is a claim about somebody else.

It holds up. Thunderbird's own State of the Bird 2024/25 reports that financial contributions reached USD 10.3 million in 2024, a 19% increase on the previous year. The foundation's EUR 1,391,189 in the same year converts to about USD 1.51 million at the European Central Bank's 2024 annual average of 1.0824 dollars to the euro. Thunderbird raised 6.8 times more.

Two donation-funded desktop applications, comparable user bases by the foundation's own account, and one raises nearly seven times what the other does. LibreOffice covers six applications, dozens of file formats and 120 languages. Thunderbird is an email client.

The difference is not product surface. Thunderbird asks. It has asked, in-product, for years, and it reported over 335,000 individual donors behind that 2024 figure. LibreOffice did not ask until August 2026.

That is the actual argument for the banner, and it is a much better one than "underfunded." The foundation is not short of money. It has been leaving money on the table for a decade because it never put the ask where the users are, and the organisation it benchmarks against has been demonstrating the size of that gap the entire time.

What the ledgers do not tell you

Three limits, stated plainly.

The prior-year column in each statement lets you check the previous year's figure twice. Five of the eight checkable years have a restated donation number, all of them tiny: the largest is a EUR 4,532 upward revision to 2020, or 0.33%. This article uses the as-first-reported figure for each year throughout. Nothing in the argument moves either way.

The ledgers are cash-basis and translated from German by the foundation's own process, and the 2025 format change means the expense categories before and after are not strictly comparable line by line. Totals are.

And the largest number is missing entirely. The foundation's EUR 1.46m of spending is not what LibreOffice costs to build. Most of the code has always come from ecosystem companies paying their own developers, which never touches these books at all. The foundation funds the foundation. Reading its income statement as the cost of LibreOffice would understate the real figure by a margin nobody has ever credibly quantified.

The takeaway

The interesting thing here is not that a non-profit is asking for donations. It is the sequence.

An organisation with 42% revenue growth, flat costs, a 33% surplus margin, a record cash year in progress and an explicit one-year reserve target decided that the correct response was to start asking users for money for the first time ever. Not because it was in trouble. Because a comparable project with a narrower product had been quietly raising nearly seven times more, for years, purely by asking.

Most companies would have read the same financials and concluded they were doing fine.

The uncomfortable version of that, if you run anything funded by voluntary payment, is that your current revenue tells you almost nothing about your ceiling. It tells you what people give you when you do not ask. LibreOffice ran that experiment for fifteen years and published the result monthly. Somebody should probably read it.