Its unlikely to happen for a number of reasons. First of all is that Broadwell, Haswell and many of Intels other architectures show a reaction to the changing market, i.e. a massive spike in the demand for low power, ultra mobile chipsets for mobile devices such as phones, tablets and even household appliances such as all-in-one control centers. This field is dominated by ARM holdings at the moment but Intel is and has been trying to get an upper hand in this market. What they are attempting to do is create a single processor that works across all fields, mobile (laptop}, ultra mobile, desktop and server. However since each market has different requirements and picks processors for a variety of reasons its more likely that in the future Intel will move to not only having different processing "family's" (for lack of a better word) but even different iterations of its x86 instruction set for different areas. Admittedly though they wouldnt want this, but its inevitable since a single all encompassing processor for each market makes no sense and would lose Intel a lot of market share to a variety of competitors.
The biggest thing keeping me from thinking that this will be the case is the server space. Server customers of course take power and cooling into consideration when making a decision, but by far the biggest requirements are processing capabilities and expansion, neither of which would be handled well by an integrated solution. If their customers arent able to buy new motherboards and drop in more RAM and processors as needed then Intel would likely lose major business to competitors that would be more than happy to take their server and super computer customers away, competitors like IBM, Nvidia and AMD just to name a few. And seeing as the market for servers is going to grow exponentially withing the next few years (thanks in part by the push to cloud computing) Intel would lose a lot of money by ignoring this space.
I mention the server space for this reason, modern desktop processors are essentially re-binned server processors, case in point Sandy Bridge-E. It would be easy for Intel to charge a premium for a re-binned server cpu and sell them as desktop processors.
Of course the desktop market is shrinking, but its not going to go away for a long time, and this brings me to a point of another major part of their revenue, OEMs like dell, HP, lenovo, sony, etc. who make computers and sell them to the public. As this article points out AMDs market share is only about 20%, but if intel were to go to an integrated solution many of these companies would likely jump ship not only to AMD but likely Nvidia as well. Nvidia has long been trying to replace traditional processors with its GPU parts and something like this would allow them to begin making GPUs that can be dropped into a motherboard like processors.
Of course what makes this all possible is Microsoft finally moving away from supporting x86 exclusively and and supporting other instruction sets with Windows 8 and this is likely to continue. Nvidia has enough money and influence they could likely get Microsoft to consider adding CUDA support to Windows, and even if not Nvidia does have a license to use ARMs instruction set (or if they dont unlike x86 ARM does license it out to other companies). They could easily create one that runs on ARM.
What we have here, in my opinion, is intel trying to serve many markets with one chip, bit its just not going to work in the end. Intel finally has a lot of competition on the horizon (not just AMD anymore) and a move like this would hurt them in more places then it would help them.
