As correctly assumed by the skeptics here, the two-year number is sleight of hand. As of one year ago, the collaboration between nVidia and Sony was apparently not finalized, and rather was being explored as one of two or more PS3 GPU possibilities. Sony attempted but apparently failed to develop a graphics solution in-house, and as I understand it was approached by nVidia with the suggestion that a competitive solution would require shader and other technologies Sony was ill-equipped to deliver on par with ATI's Xbox2/NR chips.
From what was said by an nVidia insider at the time, Sony's in-house solution was set to be little more than a dramatic polygon count evolution of the PS2's graphics chip, scaled in rough approximation to the PS1-PS2 jump, with relatively few new hardware effects (and perhaps lacking support for emergent display technologies). The theory behind nVidia's discussions with Sony was that the Japanese hardware and software development communities had generally missed the boat on the various next-generation shadowing, texturing, shading, and geometry technologies incorporated into ATI and nVidia chipsets, and consequently had fallen behind Western developers on the software development curve for such features.
Now that all of the major console hardware players have partnered with Western graphics chipmakers, the expectation is that the next generation (or two) of software will begin to show more dramatic visual gains than the last one did. EA's recent "screenshots", for example, aren't too far off from reality except in the specifics.