How much computer does a home server actually need?
Self-hosting forums hand first-timers a six-core parts list for what is really a storage job. Serving a household's photos and documents is mostly the disk and the network waiting, not the CPU. Size the always-on box for the work it runs every minute, then spend the saved money on the one thing that actually fails you: the off-site copy.
Ask a self-hosting forum what to buy for your first home server and you will get a parts list. A six-core CPU, 32 gigs of memory, a motherboard with enough lanes for a future expansion you have not planned yet. The replies are kind and detailed and mostly aimed at the machine the replier wishes they had bought. Almost none of them start by asking the only question that matters: what is the box actually going to do?
For most people typing "my first home server" into a search bar, the honest answer is small. They want their family's photos and documents off the consumer cloud and onto a drive they control, reachable from their phone, with a copy somewhere safe. That is a storage job. It is also one of the lightest workloads you can hand a computer, and sizing it like a workstation is the first and most expensive mistake.
The workload is mostly waiting
A file server spends almost all of its time idle. When someone opens a photo, the machine reads bytes off a disk and pushes them over the network. The slow parts of that sentence are the disk and the network. The CPU is a clerk handing packets along, and a modest one keeps up with a household without breaking a sweat.
The numbers that bound this are not the CPU's. A single USB or SATA drive reads at a few hundred megabytes per second at best, and your home upload speed, the thing that decides how fast you reach those files from outside the house, is usually a small fraction of that. Both ceilings sit far below what a cheap, low-power chip can saturate. Buying a faster processor to move files is like buying a faster engine to sit in the same traffic. The jam is somewhere else.
"Future-proofing" pushes you the wrong way
The usual counterargument is headroom. Buy big now so you can add Jellyfin, a few virtual machines, maybe a Proxmox cluster later. It sounds prudent. It quietly inverts the design, because it sizes the always-on machine for the workload it might run someday instead of the one it runs every minute.
Headroom on an always-on box is not free. A machine that lives in a closet running 24 hours a day pays for its idle power continuously. A low-power ARM board idles in the single-digit watts; a mini PC or a repurposed desktop typically idles several times higher, and that gap is heat and electricity you spend every hour for a CPU that is doing nothing. The thing that is always on should be sized for the thing that is always running. Heavy, occasional work, like transcoding a movie, belongs on a machine you wake up when you need it, not on the one holding your files.
Restraint is the skill the parts lists skip. The right box for personal storage is the boring one: small, silent, sips a few watts, disappears onto a shelf. The current Harbor Dock prototype runs on a NanoPi Neo3, an ARM board you could lose behind a router, precisely because the storage job does not need more and the always-on tax is real. The point is not the specific board. It is that the answer to "how much computer" trends toward less, not more, the moment you size for the actual work.
The one place not to economize
None of this means cheap out everywhere. It means move the money to where the risk actually is, and the risk is not the processor. It is that a single small box, chosen perfectly, is still one dead drive away from losing everything on it. A server is not a backup. A mirror in the same box is not a backup either; it survives a drive failure but not a theft, a fire, or a careless delete that copies to both halves.
The household version of the old 3-2-1 rule is small and worth the whole budget: keep your files on the box, keep a second copy somewhere the house fire cannot reach, on a schedule that runs without you remembering, and restore from it once so you know it works. That last clause is the one people skip. An untested backup is a guess. The cheap off-site copy, not the fast CPU, is what turns owning your storage from a hobby into something you can rely on.
When you really do need more
The small box is not a universal answer, and pretending otherwise would be its own kind of dishonesty. If you run several people editing video off the same share over SMB, or a media server transcoding 4K to a few phones at once, or a dozen containers that actually earn their keep, you have a real compute workload and you should buy for it. That is a legitimate build. It is just not what most first servers are, and it is worth being clear-eyed about which one you are building before the parts list talks you into the other.
For the common case, the family that wants its own photos and documents back under its own roof, the machine is smaller and quieter than the forums admit. Size it for storage, spend the saved money on the second copy, and add compute later only if a real need shows up. Most of the time it never does, and the boring little box on the shelf keeps doing the one job you actually bought it for.