Renting a platform against owning one, as a category
The argument on this page is about the model rather than about any vendor. It applies to every hosted streaming platform including the good ones, and it works in both directions — there is a volume below which renting is simply cheaper, and this page says where it is rather than pretending it does not exist.
Two cost curves, and the shape is the whole argument
A rented platform is a variable cost. It starts near zero, which is its great virtue, and it rises with usage — bytes delivered, minutes encoded, sessions measured. The curve has no ceiling, and its slope is set by somebody else's rate card.
An owned platform is a fixed cost with a step. It starts high, because capacity is bought before it is needed, and then it is flat: inside the commitment the marginal viewer costs nothing. It steps when the commitment is renegotiated, and between steps it does not care how successful you are.
Everything else people argue about downstream of this — control, residency, exit risk, customisation — follows from where the machines are. The curve is the reason the machines are there.
Where the curves cross, and why the answer is regional
The crossover is not a universal number, but the inputs are always the same five: delivered volume per month, catalogue size and growth, peak concurrency, encode hours including re-encodes, and the amortisation period your finance side will accept. Given those, both curves resolve and can be compared honestly.
Geography moves the crossover more than most operators expect, because the rented curve is not the same everywhere. On the North American card the per-gigabyte rate falls 4.25× across the volume tiers, which is what lets a large service rent delivery and still have a margin. On the South Asian card it falls 1.51×. Tripling delivered volume there multiplies the bill 2.89× and improves the blended rate by 3.8% — so scale, which is the standard answer to a rising rented bill, is not available.
That is why an argument that reads as ideological in one market is arithmetic in another. A service in Dhaka and a service in Denver making the same decision are not making the same decision.
Below roughly a terabyte a month, renting wins and it is not close. The commitment sits idle, the hardware depreciates against traffic that is not there, and the meter is genuinely the cheaper instrument. Any vendor telling you otherwise at that volume is selling.
What each model actually costs you, beyond the invoice
The columns most comparisons leave out, because neither is a line item.
| Rented | Owned | |
|---|---|---|
| Cost at launch | Near zero — the model's real advantage | Capacity bought before the audience arrives |
| Cost at scale | Rises with every viewer, with no ceiling | Flat inside the commitment; steps when it is renegotiated |
| Who carries hardware failure | The vendor. Genuinely worth something | You, or whoever you pay to be on call |
| Utilisation risk | None — you pay for what you use | Yours. Below plan is money already spent |
| Exit | A migration project on their timetable | Nothing moves; support stops and the estate keeps running |
| Data residency | Wherever they host it | Wherever you put the machines |
| Speed to launch | Days to weeks | Weeks to months, gated by procurement and metadata |
| Who you depend on | A vendor's roadmap and their continued existence | Your own operations capability, and the runbooks you hold |
The one question that settles it
How many terabytes did you deliver last month, and where were the viewers? Everything else on this page is downstream of that pair, and most operators evaluating platforms cannot answer it from CDN logs — they answer it from an estimate, which is how a decision this size ends up being made on a feeling.
Get the number first. Then run both curves against it, including ours, and let the arithmetic be the argument. If it says rent, rent — that outcome is on this page because it is a real one, and a vendor whose comparison page cannot produce it is not comparing anything.
Run it against your own numbers
The calculator on the pricing page uses published rates and shows its arithmetic. It is allowed to come out against us, and below about a terabyte a month it does.