VCUs seem like an excellent solution for this problem and I'm actually surprised they didn't opt to build custom hardware for this earlier on. I wonder if other video companies will hop onto this and start releasing their own VCUs. I imagine this kind of hardware could be very useful for any video streaming site as well as people who edit videos all day (Hollywood etc.)

The article didn't really go into technical details, so would anyone happen to know why a VCU happens to be faster than a traditional GPU? I thought the point of building GPU kernels was supposed to be able to speed up tasks like this (doing the same set of steps over millions of different inputs).

Lbry

TLDR: YouTube is creating custom video encoding hardware to lower costs.

This part is just a lie, however, given YouTube frequently practices censorship of political views and ideologies their employees disagree with:

> Our mission is “To give everyone a voice and show them the world.” Let anyone upload a video to show anyone else in the world, for free.

This seems like the same gaslighting we saw recently with YouTube CEO Susan Wojcicki receiving an award for free expression sponsored by YouTube itself: https://news.ycombinator.com/item?id=26880262