Why Self-Hosting Matters (And When It Doesn't)
Self-hosting matters because it makes you, not a vendor, the one who decides whether your tools keep working and on what terms.
The cost is real: you become the person responsible for backups, updates, and the failure that happens at 2 a.m.
It is a clear win for files, photos, media, and community services, and a bad idea for anyone with no spare hours and no appetite for being the sysadmin.
Self-hosting matters because it is the only arrangement where the person using a piece of software is also the person who decides whether it keeps existing. When you run Nextcloud on a machine you control, nobody can change the terms, sunset the product, raise the price, or lock you out because an automated system flagged something. That is the whole argument, and it is a real one.
It is also not free. You trade a monthly bill for a standing obligation: backups, updates, certificates, disk failures, and the occasional Saturday you had other plans for. This page makes the case honestly, including the part most guides skip, which is the list of people who should read the argument, agree with it, and then not do it.
The argument in three paragraphs
Every hosted service you use is a lease. You rent a feature set from a company whose interests are only partly aligned with yours. That works fine right up until it does not: the tier you depend on is discontinued, the free plan gets a seat limit, the app is acquired and shut down, or your account is suspended by a classifier with no appeal worth the name.
Self-hosting changes who holds the lease. The software runs on hardware you own or rent directly, the data sits on a disk you can pick up, and the only things that can switch it off are you and the power company.
It is worth being clear about what this does not do. Nextcloud is not a better product than Google Drive. Immich has not beaten Google Photos on polish. They are comparable tools with a different owner, and the ownership is the entire point. If that idea is new to you, what data sovereignty means in practice covers the same ground from the community and legal angle.
What you actually get back
The case for self-hosting is built from specific things rather than grand ones:
- The software stops changing under you. You choose when to upgrade. A redesign you dislike arrives when you allow it, or never.
- Your data is in a format you can read. Files on disk, a database you can dump, photos in their original form. Export is not a feature you have to request.
- There is no second customer. No ad product means no reason to scan, profile, or retain anything past its usefulness.
- Costs level off. A hosted plan grows with your storage forever. A drive is a one-time purchase that gets cheaper per terabyte each time you replace it.
- It works when the internet does not. A server on your own network keeps serving files, music, and chat straight through an outage.
People underrate that last one until the first time it matters. Local services fail in a much smaller radius than cloud ones, which is why community networks lean on them so heavily. How mesh networks work explains why a local service and a local network belong together.
The counter-argument, stated fairly
Hyperscale hosting is not popular purely through inertia. It is genuinely good at several things that are hard to reproduce on one machine in a cupboard, and pretending otherwise is how people talk themselves into a bad decision.
Durability. Photos on a major cloud are replicated across data centres by people who do this full time. Photos on a single drive at home are one failure away from gone, unless you have built a real backup chain and tested a restore. Most self-hosters have never tested a restore.
Availability and patching. A vendor has staff on rotation and patches silently. You have you. If your server dies while you are away, it stays dead until you get back, and an unpatched service exposed to the open internet is a worse privacy outcome than a hosted account would have been.
The genuinely hard problems. Spam filtering, mail deliverability, abuse handling, mobile push notifications, and smooth video streaming to a phone on mobile data are all difficult engineering. Self-hosted equivalents exist and mostly work. The word mostly is doing real work in that sentence.
Who should not self-host
This is the section most guides leave out. If any of the following describes you, the honest advice is to stay on hosted services, or to pay someone else to host open-source software on your behalf.
- You have no time budget. Not a lack of interest, a lack of hours. A small stack needs an hour or two a month plus one bad afternoon a year. If those hours do not exist, the server rots.
- Nobody else can fix it. If you host the family photos and you are the only person who knows any of the passwords, you have built a single point of failure with a pulse.
- The data is regulated. Patient records, client financials, anything carrying a legal retention or breach-notification duty. Do not learn on that data.
- You are in a rough patch. Migrating your digital life during a house move, a new baby, or a hard month is how people end up without email for a week.
- You are doing it to save money. It can be cheaper at scale, eventually. It usually is not, once you count drives, electricity, and your own hours at any honest rate.
The obligations you inherit
Self-hosting makes you the operator. That job has four duties, and none of them go away.
Backups you have restored from. An untested backup is a belief, not a backup. The usual shape is three copies, two kinds of media, one off-site. Tools like restic or Borg pushing to a cheap remote bucket cover the mechanics; a quarterly reminder to actually pull a file back covers the rest.
Updates. Anything reachable from the open internet needs patching promptly. Anything reachable only over a private overlay such as Tailscale or Nebula can move slower. Working out which of your services are genuinely public is most of the security work.
Boredom. The realistic failure is not a dramatic hack. It is a full disk in month nine, a certificate that quietly stopped renewing, and a container that has not started since a reboot in March. Uptime Kuma and a push alert to your phone catch nearly all of that.
Succession. Someone other than you needs to be able to get in.
Where self-hosting clearly wins
Some workloads are close to a free win. Take those first, and leave the hard ones until you have a restore you trust.
- Files and sync. Nextcloud, or plain Syncthing if you only need your devices to agree with each other.
- Photos. Immich handles phone auto-upload and face grouping and has closed most of the gap with the hosted photo apps.
- Media. Jellyfin for video, Navidrome for music, Audiobookshelf for audiobooks. None of it depends on a company staying in business.
- Home automation. Home Assistant is the case where self-hosting is plainly the better product, because your lights keep working when broadband is down.
- Documents and reading. Paperless-ngx for scanned mail, FreshRSS for feeds, Forgejo for code.
Community-scale services are the other strong case. A neighbourhood group running its own chat, file share, and calendar keeps functioning when the wider internet does not. Folkmoot applies the same reasoning to group decisions and voting, where handing the tally to a third party is exactly the thing you want to avoid. For a build order, see how to self-host your own cloud.
Email is the exception, and it is not close
Self-hosted email deserves its own warning. Running Postfix and Dovecot is not the hard part; that is a weekend and a decent guide. Deliverability is the hard part. A small number of very large providers decide whether your mail is seen, and a fresh IP address with no sending history starts life under suspicion.
You need correct SPF, DKIM, and DMARC records, reverse DNS that matches, an address block with no history of abuse, and the patience to file delisting requests. Even then, some of your mail will land in spam folders you cannot see. That is worse than an outage, because it fails silently and nobody tells you.
The sensible compromise is to own the part that matters and rent the part that does not. Buy a domain, use it for all your mail, and point it at whichever provider you like. If that provider disappoints you, change the MX records and keep your address. Most of the sovereignty lives in owning the domain, not in running the mail server. The same logic runs through the step-by-step de-Googling guide.
The middle path most people should take
Self-hosting is not binary. There are at least four positions on the line, and the two extremes are the least sensible for most households.
| Approach | You control | You maintain |
|---|---|---|
| Hosted proprietary service | Nothing | Nothing |
| Managed open-source hosting | Your data and your exit | Nothing |
| A VPS you administer | Data, software, configuration | The whole stack |
| Hardware in your own building | All of it, including the physical layer | Everything, plus the power bill |
The second row is badly underrated. You pay someone to run Nextcloud or a Mastodon instance, your data stays in an open format, and you can leave with a database dump. You give up the physical layer and keep nearly all of the exit rights, which for most people is the right trade.
Owning the physical layer earns its keep when the network itself is the thing you are trying to own: a building, a block, an off-grid site. That is where local hardware and mesh radio start to make sense together, and where kit from Blackout Labs fits. More guides are in the learn index.
Frequently asked questions
Is self-hosting actually more private than using Google or Apple?
Usually, but not automatically. Nothing is scanned for advertising and no company holds a copy it can hand over on request. But a service you expose and never patch, or an unencrypted backup sitting on a rented box, can leak more than a hosted account ever would. Privacy comes from how you run it, not from where it runs.
How much time does self-hosting really take each month?
For a small, stable stack of two or three services, roughly one to two hours a month once it is running: applying updates, glancing at disk space, confirming backups ran. Add one longer session a year for when something breaks properly. Setup is the expensive part, usually about a weekend per service the first time.
Is self-hosting cheaper than paying for cloud subscriptions?
Sometimes, at larger storage volumes and across several years. A used mini PC and a couple of drives can eventually beat a family cloud plan. But drives fail, electricity is not free, off-site backup still costs money, and your own hours have value. Self-host for control, and treat any savings as a bonus rather than the reason.
What should I self-host first?
Something you could survive losing. A media server such as Jellyfin, or a read-only file share, teaches you containers, reverse proxies, and backups without risking anything irreplaceable. Move photos, documents, and passwords across only after you have successfully restored a test file from your backups at least once.
Should I self-host my own email?
Only if you enjoy the problem for its own sake. Getting mail to reliably reach large providers takes correct SPF, DKIM, and DMARC records, clean IP reputation, and ongoing attention, and the failures are silent. Owning your domain gives you most of the benefit anyway: you can change mail providers without changing your address.
Can I self-host if my connection is slow or my IP address changes?
Yes. Dynamic DNS handles a changing address, and a small VPS or an overlay network such as Tailscale can reach a server stuck behind a carrier-grade NAT that blocks inbound connections. Upload speed is the real constraint: it caps how fast remote devices sync and how well video streams when you are away from home.