Self-hosted vs cloud team chat, honestly

The self-hosting conversation usually gets conducted by two people talking past each other. One is thinking about data sovereignty and vendor risk. The other is thinking about who gets woken up when the certificate expires. Both are right, and the argument only resolves once you separate the four things that actually differ.

1. Control, and what it's worth

This is the strongest argument for self-hosting and the one that survives every counter-argument, because there's no SaaS contract that fully substitutes for the data being on your disk.

• Data residency. The messages are in a jurisdiction you chose, on hardware you can point at.
• Network isolation. The tool works when the internet doesn't, which matters more in some industries than others.
• No eviction risk. Pricing changes, acquisitions, region shutdowns and deprecated features are all things that happen to somebody eventually. Self-hosted, the worst case is you keep running the version you have.
• Inspectability. With open source, you can verify the security claims rather than reading a compliance PDF about them.

The honest counter: for most companies, most of the time, none of this is load-bearing. If the reason you want control is a vague unease rather than a requirement, that's worth naming — vague unease is a bad reason to acquire a pager rotation.

2. Cost, including the part that isn't money

Infrastructure genuinely is cheaper, and it isn't close:

Illustrative figures. Now add the column nobody puts in the business case: staff time. A few hours of initial setup, roughly an hour a month of maintenance, and an unpredictable number of hours the one time something breaks badly.

At an internal cost of, say, $100 an hour, that's a few hundred dollars up front and a bit over a thousand a year. Which means:

• Under ~20 people: the time cost usually exceeds the savings. Self-host for control, not for money.
• 20–50 people: roughly break-even, decided by whether someone actively wants to own it.
• Over 50: self-hosting wins on cost clearly, and the gap widens every time you hire.

3. Reliability, and who is on the hook

The instinct is that a big cloud provider is more reliable than your VPS. Mostly true, with two caveats that matter.

• Cloud outages are correlated. When a large provider has a bad day, your chat goes down at the same moment as everyone else's — including, frequently, the tools you'd use to coordinate the response.
• Self-hosted outages are yours. A modest well-run VPS has excellent uptime, but when it doesn't, there is no status page to point at and no vendor to escalate to. There is you.

The real difference isn't uptime percentage — it's who is accountable. Some organisations strongly prefer “we can fix it ourselves.” Others strongly prefer “we can point at a vendor.” Both are legitimate, and it's a cultural question more than a technical one.

4. Security: differently secure, not more secure

Self-hosting is often presented as strictly more secure. It isn't — it's a different threat model, with a trade in each direction.

You remove a third party and take on operational security work. If you'll actually do that work, self-hosting is a net gain. If the box will run unpatched for eighteen months because the person who set it up moved teams, it is emphatically not.

A decision procedure

1. 1Is data residency or network isolation an actual requirement? Yes → self-host. Done, no further analysis needed.
2. 2Is there a named person who wants to own it? No → don't self-host. Infrastructure without an owner is a future incident with a date on it.
3. 3Are you over ~50 people on a per-seat plan? Yes → the cost case is strong enough to justify creating that owner.
4. 4Is a free hosted option available that meets your needs? Yes → take it, and revisit if control becomes a requirement later.
5. 5Otherwise → cloud. Not because self-hosting is bad, but because “nobody specifically wants this” predicts abandoned infrastructure with remarkable accuracy.

One more thing worth checking either way: can you leave? A hosted product with full data export and an open source, self-hostable core lets you defer this decision indefinitely, which is usually the smartest move available. Start hosted, move in-house if you ever need to, and don't spend a month deciding.