Does switching video formats affect streaming energy consumption? We measured it.

Disney+ is reportedly dropping HEVC in favour of VP9, Google's royalty-free video format. On paper, it's a licensing decision: HEVC comes with patent fees, VP9 doesn't. But every format switch also changes how much electricity gets used to encode and play back video — and nobody had actually measured it.

So our Watt Lab did.

The first version was wrong

GoS President and Watt Lab co-chair Ben Schwarz posted an early measurement from OWL, our energy test bench. Two well-known figures in video compression, Thierry Fautier and Jan Ozer, pointed out that the comparison wasn't fair — the formats hadn't been tested on equal terms.

They were right. So the team reran it overnight, this time properly: every format encoded the same way, on the same server, then played back on the same five living-room devices, with electricity metered at the wall.

We're leading with the correction because it's the more useful story. A single flawed number would have told Disney+ and everyone watching the wrong thing.

What the second run found

For viewers, switching to VP9 doesn't cost more energy — as long as their device can decode it in hardware. On devices with a built-in decoder, all four formats tested cost the same to play, within a tenth of a watt. On older devices without hardware support, software decoding is needed instead, and there VP9 was actually the cheapest of the four, HEVC the most expensive. Since a title is encoded once but watched millions of times, this is the part that matters most for streaming's overall footprint — and it doesn't argue against the switch.

For encoding, the settings matter more than the format's name. Pushing any modern codec to its slowest, most thorough setting costs several times more electricity than using its defaults. Changing how hard you compress the video changes your energy bill far more than which format you chose.

The content itself is a bigger factor than the codec. A fast-moving cartoon used twice the power of a quiet drama scene, regardless of format. And whether your audience's devices have hardware support for a given format is worth checking before switching — it's the difference between decoding that's nearly free and decoding that isn't.

What we still don't know

We didn't measure how many bits each format needs for the same picture quality — and therefore what a switch does to network and delivery energy. That's a different question, and we're not going to guess at the answer. It's the subject of a joint article coming up with Thierry Fautier.

The bottom line

A royalty-driven move to VP9 doesn't have to cost energy at the viewer's end. What decides the outcome is whether people's devices can decode it in hardware, and how the encoding settings are configured — not which logo is on the codec.

Full numbers and technical caveats are in Ben's LinkedIn post, with a link to the complete write-up in the comments. Thanks to Thierry Fautier, Jan Ozer, and Tania Pouli for making the second run the right one.