"Free" software with a salary attached, versus a subscription with none. This page does the full maths — including the costs self-hosting fans hide and the costs cloud fans hide.
Self-hosting advocates say "it is free". Cloud advocates say "your time is not free". Both are telling half the truth, and the halves are different sizes for different offices. Here is the complete cost list for both sides, so you can put your own numbers in.
Illustrative example only — prices change, so check current rates. The shape is the lesson: rent grows with headcount, owning does not.
The cloud side looks simpler: a per-user monthly fee. Put your current numbers in — the exact rates change often enough that you should check today’s price list rather than trust any guide, this one included. The structural facts matter more than the rate: the bill grows with every hire, it is charged in plans with tiers that push growing offices up a level, and it never ends. A 25 person office commonly discovers its three-year cloud total would have bought the server several times over — but that subscription also bought an entire global engineering team’s uptime, which is a real thing your server does not have.
Self-hosting hides: downtime risk (when your server is down, your office is down, and the fix is your problem at your speed), training time, and the cost of the day the one admin is away. Subscriptions hide: price rises over time, the migration cost of ever leaving, data sitting under another company’s rules, and the fact that the "cheap" tier quietly lacks the storage or features you end up needing. An honest comparison writes all of these down — the decision usually stays the same, but it stops being a surprise later.
Rough guide from real offices: below about five users, a subscription is usually cheaper in total — the admin hours dominate. From roughly ten users upward, self-hosting usually wins on money within the first year or two, and the win grows with headcount. In between, the decision is not really about money: it is about data control, internet reliability and whether your office has the named owner the system needs. Run your own three-year totals — hardware plus admin hours versus seats times 36 months — and let that sheet, not a slogan, decide.
Whichever side you choose, price the leaving as well as the joining. Leaving a cloud means exporting everything and retraining everyone — this series has a whole migration guide because it is real work. Leaving self-hosting means the reverse. The cheapest system is often the one you can stay on for five years, so the fit questions — owner, profile, growth — are worth more than this quarter’s arithmetic.
Watch the system in the video above, then do your own three-year sheet. The numbers for your office will make the decision obvious.
Watch Full VideoYes. Nextcloud and the OnlyOffice Docs community edition are open source and free for business use at this scale. You may optionally buy support subscriptions, and larger enterprises often do, but nothing in the software demands payment or counts your users.
Once settled, roughly two to four hours a month for a small office — updates, a backup check, the odd user change. Budget more in the first month and in migration month. If that time is a paid consultant’s, put their real hourly rate in the maths; it is still usually cheaper than 25 seats.
Backups — not the drive, which is cheap, but the routine: off-site copies, quarterly restore tests, and a second trained person. Offices skip costing it because it is time, not money. It is also the cost whose absence is discovered exactly once, at the worst possible moment.