The transcoding cost that multiplies

Encoding is the tier where the cost model quietly changes what you are willing to build.

The multiplication

A managed encoder charges per minute, per rendition. The unit is not "an hour of content". It is an hour times the number of quality levels, times every time you touch it.

The second column is what makes an improvement a budget conversation. The third makes it a scheduling one. That difference decides how good the quality levels get over five years, far more than the initial encoder choice does.

Why the workload suits owned hardware unusually well

• Predictable. Catalogue hours are known in advance.
• Batchable. Nothing about a VOD encode is latency-sensitive.
• Restartable. A failed job is requeued, not lost.
• Bounded by a resource you can count. Throughput is GPUs, not a vendor's concurrency quota.

That is close to the ideal case. Compare it with sign-in traffic, which is spiky, latency-sensitive and unpredictable, so the hybrid pattern leaves that tier in the cloud.

Live shares the hardware, not the priority

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.

The number to bring

Catalogue hours, how many quality levels you encode, and how often they change. Those three usually make the case on their own, in either direction. For a small catalogue that never changes, usage pricing is cheaper, and that is a legitimate answer.