Per-minute transcoding pricing looks reasonable until it is multiplied by ladder depth, and then again by every re-encode. That is how catalogues end up stuck on ladders nobody would choose today.
Encoding is the tier where the cost model quietly changes what you are willing to build.
A managed encoder bills per minute, per rendition. So the unit is not "an hour of content" — it is an hour times the ladder depth, times every time you touch it.
| Decision | Cost on a meter | Cost on owned hardware |
|---|---|---|
| Add a low-bitrate rendition | Re-encode the catalogue, billed in full | Queue time |
| Adopt a newer codec | Re-encode the catalogue, billed in full | Queue time |
| Fix a bad ladder choice | Re-encode, billed in full | Queue time |
| Nightly re-encode of the archive | Not seriously considered | Electricity |
The second column is what makes an improvement a budget conversation. The third makes it a scheduling one. That difference decides how good the ladder gets over five years far more than the initial encoder choice does.
That is close to the ideal case. Compare it with sign-in traffic, which is spiky, latency-sensitive and unpredictable — and which is why the hybrid pattern leaves that tier rented.
Live encoding runs on the same GPUs with a latency-sensitive profile and priority over batch work. The scheduling rule is unglamorous and important: an overnight archive re-encode must never be able to starve a match.
Catalogue hours, ladder depth, and how often the ladder changes. Those three usually make the case on their own, in either direction — for a small catalogue that never changes, a meter is genuinely cheaper, and that is a legitimate answer.
If the pattern in this write-up describes your setup, the next step is reading your actual numbers against it. Send twelve months of billing and your concurrency curve.