Over the last 20 years AMD and Intel have traded blows but for the last 5 years, Intel delivered a series of killer blows that really put AMD on the back foot. The acuisition of ATi also made problems for AMD as an organisation, and whilst important for AMD's long term roadmap (namely, developing APU's) it put the company behind. Arguably, AMD also shot themselves in the foot by focusing on hexa and octacore solutions, most applications can't use that many cores simultaneously and is therefore superfluous at the moment.
I believe that the decision not to pursue higher end chip development over the last few years has been tactical. Not only would it have required a massive injection of money which AMD doesn't really have access to at the moment, but it would also detract from the new server and mobile system focus that AMD is trying to shift to. The idea is to steal some areas of chip manufacture that Intel doesn't even compete well in yet, expanding its potential for revenue. Its proven relatively successful in its low power designs so far but Intel is rapidly catching up.