AMD did what you may call a tick - tick strategy, as they changed the architecture and the manufacturing process, which is definitely a tricky thing to do. In addition, let's not forget that this APU has a gigantic GPU attached on the same die to AMD's 32nm CPU. This is one of the most complex chip that was ever made in computer industry and it is really a big one.
Good yields mean that company makes more money but we are hearing Llano chips in a wafer work in 50 to 60 percent cases. Furthermore, the chip is quite big and you cannot get too many of them from one 32nm wafer. Good yields would probably be if 80 to 90 percent of the chips worked, while 90+ yeild is a really good one.
AMD cannot make much money due to bad yields, but at least it proved that it knows how to marry AMD CPUs and graphics formerly known as ATI on the same chip. It's the graphics part that is the toughest to make, as this is the first time ever that Global foundries, previously AMD FAB, had to make a graphics chip on 32nm Silicon on insulator. This is anything but easy.
The next generation will be better, but we hear that Bulldozer has quite healthy yields. In Interlagos platform, Opteron 6100 has up to 16 cores and is definitely an enormous chip.
The CPU portion of the chip seems to be doing fine, but the graphics part is responsible for poor yields. Things are expected to get better in 2012 with the second version of Fusion in 32nm, Llano successor codename Trinity. This new APU has been demonstrated last week and this time AMD enlarged 28nm Radeon 7000 graphics core from 28 to 32nm and still made it work on SOI process.
That will be AMD's version of tock, the same manufacturing process and the second generation Fusion core design and the expectations for these chips are high. From a consumer point of view, Llano is definitely a good thing, as it's affordable and gives you great graphics for little cash, together with a decent CPU.