Schloss Dagstuhl Leibniz-Zentrum für Informatik publishes several open-access computer science conference proceedings, mostly through their LIPIcs imprint. LIPIcs charges a publication fee of 60 (sixty) EUR per published paper, all of which is used to defray Dagstuhl's direct costs. (Other Dagstuhl imprints charge similar publication fees.)
By far the most significant cost is the human labor of copy-editing and assembling complete proceedings. (Most of the actual typesetting is done by authors using Dagstuhl's latex templates; however, Dagstuhl staff do edit each paper to enforce their strict formatting guidelines. Similarly, the work of assembling proceedings is split between Dagstuhl and the conference organization.) Smaller costs include DOI acquisition, metadata extraction/maintenance, archiving, and server maintenance. Dagstuhl's publications are also permanently archived by the German National Library.
In reference to @Allure's rather dismissive answer, Dagstuhl does not publish anything on paper; they do significant marketing (compared to competing conference publishers like ACM, IEEE, or Springer); they "attend conferences" only in the sense that its scientific members are themselves active researchers; they use their own document management system (which was developed over many years with the support of European research grants); they do not do plagiarism checks (in part because program committees and reviewers already do that themselves); and they leave the actual content of proceedings (but not the formatting) entirely up to the conference organization. As usual in computer science, all substantive editorial work is done by volunteers.
Notably, LIPIcs does not charge their small publication fee to authors, but rather to conferences, where they are typically covered by registration fees. Conferences have to apply to publish their proceedings through LIPIcs, and their acceptance is decided by the volunteer editorial board, based primarily on the likely quality, impact, and longevity of the conference.
Dagstuhl's primary activity of hosting research workshops (again, largely organized by volunteers) makes significant/expensive marketing unnecessary. Dagstuhl doesn't have to advertise to the research community, because the research community already comes to them.
When LIPIcs was established, these publication costs were heavily subsidised by Dagstuhl's general fund, which primarily comes from the German Federal Ministry for Education and Research. In 2015, Dagstuhl was informed by their primary funding agency that their general fund could no longer be used to support publishing activities, because EU regulations forbid government interference in the private publishing market. So I believe the 60EUR fee is an accurate reflection of Dagstuhl's actual per-paper publication costs.