Renting software was a phase

For twenty years the answer to every software question was the same: don't run it yourself, rent it. It was good advice. Servers were expensive, ops teams were scarce, and the software you could buy outright was worse than the software you could subscribe to. So the industry moved everything to someone else's computer and called it the cloud.

If you own a network, that trade has quietly gone bad.

The asset mismatch

Fiber in the ground is a 25-year asset. Some of it will outlive the company that installed it. The software that designs and manages that asset, though, is rented month to month — priced per seat, hosted by a vendor, subject to their roadmap, their pricing changes, and their continued existence.

That mismatch has consequences. Your as-built records, splice data, and service maps — the operational memory of the network — live in a system you don't control, in formats chosen for the vendor's convenience. Exporting is possible in theory and painful in practice. Every year the data gets heavier and the exit gets harder. That's not an accident. Lock-in is the business model.

Meanwhile the bill scales with your headcount, not your network. Add a designer, pay more. The network didn't change; the rent did.

What changed

Two things made renting the default, and both have weakened.

The first was the cost of running software. It used to take a team to keep a server patched, backed up, and monitored. It doesn't anymore. Modern tooling — and lately, AI agents that handle the routine work — has made operating a single well-built service something a small operator can do without hiring for it. The ops argument for SaaS was real in 2010. It is mostly gone.

The second was software quality. Rented software won because it shipped faster and broke less than the boxed software it replaced. But the gap has closed from the other direction: it's now feasible to deliver owned software that updates as easily as SaaS does — pulled onto your machine, on your schedule, with your say-so.

What owning looks like now

Owning your software doesn't mean going back to DVDs and license dongles. It means a few plain things:

It runs on your machines. One operator, one instance. Your outage is never someone else's incident queue.
Your data stays in open formats. GeoJSON, PostGIS, files you can read with tools you already have. Leaving should be a copy command, not a negotiation.
The source ships with it. If the vendor disappears tomorrow, your network management doesn't. You can read it, fix it, extend it, or hire anyone you like to do so.
You pay for the software, not for permission to keep using it.

None of this is nostalgia. It's the same judgment that made you build the network instead of leasing capacity: infrastructure you depend on for decades should be infrastructure you control.

The pendulum

The industry is starting to say this out loud. Costs of the subscription stack keep compounding; the companies that moved everything to rented services are doing the arithmetic on what it would cost to own again. For most kinds of software, the answer will stay "keep renting." But for the systems that hold your core asset's data — the ones you can least afford to lose access to — the answer is changing.

You own the network. The software should come with it.