Every SaaS subscription is a lease on someone else’s backend. Leases are sometimes exactly right — but they have a shape: payments never end, they grow with your success, and the landlord can change the terms. This post lays out the rent-vs-own math honestly, including the cases where renting wins.
Key takeaways
- Per-seat and usage pricing compound with growth; owned infrastructure is a flat line.
- Rented tools carry leverage: pricing, retention and features can change under you.
- AI agents collapsed the old cost of owning — engineering time — to roughly a server bill.
- Rent the edges of your business; own the center.
The renting math: a worked example
Illustrative numbers, deliberately ordinary. A 40-person team on one per-seat tool at $8 per user per month:
| Year | Team size | Annual cost | Cumulative |
|---|---|---|---|
| 1 | 40 | $3,840 | $3,840 |
| 2 | 55 | $5,280 | $9,120 |
| 3 | 75 | $7,200 | $16,320 |
One tool. Most companies rent four or five (chat, project tracking, client portal, storage), so multiply accordingly. Usage-priced backends have the same shape with a twist: the bill is unpredictable — your best month of growth is your worst invoice.
And none of it buys the thing itself. Stop paying, and the product, the history and the workflows vanish.
The cost nobody itemizes: leverage
The subtler price of renting is that the vendor holds leverage over your business. Pricing changes, retention limits, feature gates, product sunsets — all decided by someone whose incentives aren’t yours. If a tool’s price doubled next year and you’d grudgingly pay, that tool owns a piece of your business, not the other way around.
The owning math
Owning used to be expensive in a different currency: engineering time. That’s the variable that changed. When your AI does the building on an open-source framework like jerrycan, owning costs roughly:
- A server — often tens of dollars a month for a compiled Rust backend serving a whole team, flat regardless of headcount.
- Occasional attention — updates happen by asking your agent, the same way the product was built.
Against that ledger sits an asset: the code and the data are your IP, portable to any host, immune to repricing and sunsets.
When renting still wins
An honest framework needs this section. Renting is the right call when:
- The tool is far from your core. Nobody should self-host their accounting or email deliverability.
- The ecosystem is the product. A managed platform’s integrations can be worth the meter — Supabase’s hosted Postgres is genuinely good.
- You need enterprise features today. SSO, compliance attestations and 24/7 support are real work someone must do.
The pattern: rent the edges, own the center — the products your business actually runs on.
A two-question decision framework
For every subscription you pay:
- If this price doubled next year, would I still pay? A grudging yes means the vendor has leverage — candidate for owning.
- Does this tool hold data I’d fight to keep? If yes, that data belongs in a database you control.
Two yeses and the math almost always favors owning — especially now that “building it” means describing it.
Frequently asked questions
Isn’t self-hosting more expensive once you count my time? It was — that’s why this post exists. When the build and the maintenance are conversations with an agent, the time cost collapses to reviewing what it did.
What about reliability? SaaS teams run ops for me. True, and for the edges of your business that’s worth paying for. A compiled binary with built-in backups on a reputable VPS is, in practice, boring — the good kind.
Where do I start? With the subscription that annoys you most. Team chat is the classic first win.
Run the two questions over your subscription list. The first “grudging yes” is one sentence away from being yours.