Canon has a newer type of motor, called STM, which is introduced in lenses like 18-135 f/3.5-5.6 IS STM, 40 f/2.8 STM and so on.
STM also provides quick, silent and smooth focusing. Even the focusing is smoother than USM, and this will be better when recording video. But why are the new L-lenses still prefer USM over STM (it's like all of them are using USM, and none are using STM)?
Like many things when it comes to designing hardware for photography, there are always tradeoffs to be considered and made.
STM lenses sacrifice a little speed in order to be quieter and smoother (no jerky starts and stops). This is important when using Autofocus while recording video.
Lenses with USM focus designs are built for speed first and quiet operation second. Since they are optimized for still images, the jerky starts and stops that help them get there faster are of no consequence.
More advanced videographers tend to use manual focus, often with external focusing rigs that attach to the lens and allow much finer manual focusing control, and also enjoy the benefits of the superior optics in many "L" series lenses. That is, if they are not using the even more expensive Cinematic grade lenses that in addition to superior optical image quality include such features as parfocal zooming and corrections to eliminate focus breathing (but no AF).
USM is fast, STM is smooth. For still images (particularly with sports and wildlife) it is important that focus tracks quickly whereas in video the quick movements of a USM lens make the video feel jerky.
Focusing speed of USM is little faster than STM. Higher end lenses are used for some 'still' images, so USM is batter option.
