I don't buy it. The transistor counts of modern GPUs far exceeds that of even multi-core CPUs. It's not rational to say "oh, well, that's all just trivial engineering, it'll go away". Especially as GPU performance continues to be taxed to a far higher degree than the CPU typically is. Modern popular AAA PC games like Battlefield 3 or Starcraft 2 will eat all of the GPU you throw at it and then some.
Intel control the whole platform now that they have PCH, which enable them to put the GPU in the CPU. With Ivy Bridge you even have OpenCL support. There is nothing stopping Intel from embedding a proper GPU now that they don't have to fit it in the northbridge. The only reason NVIDIA is still in business is because embedded graphics sucked too much.
Hmmm, you are right... maybe I shouldn't have said "maxed out" but "mostly acceptable". I think that "maxed out" will be a niche market, specially since the computers able to really max out games are already niche, and PCs in general are loosing market share. I'm not implying its a trivial engineering problem, just a problem Intel will be able to solve in one/two years and deliver processors for a small box you plug into your TV and happily game without investing money and real estate space into a PC.
I should have mentioned, that chart is for sales in 2011. Some of that is low graphics gaming, but people aren't spending $18.6 billion on bejeweled (maybe on Farmville though). If you take just the segment of the market that is "high-end graphics gaming" then it's still a multi-billion dollar a year industry. Why would a self-sustaining industry of that size just spontaneously go away?
And then you see what game devs are working on for the future such as this: http://www.youtube.com/watch?v=Gf26ZhHz6uM or this: http://www.youtube.com/watch?v=Q4MCqM6Jq_0 or this: http://www.youtube.com/watch?v=EhwZ7Sb0PHA and the notion of "good enough" quickly flies out the window.