Good analysis. I honestly am a bit baffled by Intel's purchase of Altera (especially for the price they paid). I don't think this means there will suddenly be a scramble to buy up FPGA companies (Altera and Xilinx control something like 90% of that market). It's not clear that having an FPGA on your SOC will be a clear win in the datacenter - FPGA programming is hard partly because it's hardware development which is in many ways very different from software development, and partly because the tools suck. The FPGA dev tools suck because Xilinx and Altera have controlled 90% of the market and haven't been willing to spend much money to improve their tools. Maybe Intel will change this in the case of Altera's Quartus tools, but I tend to doubt it - Intel isn't known for their software either. I suspect nobody is in a hurry to acquire Xilinx because they'll wait and see how this plays out with Intel/Altera.
AMD would certainly be a poor match as a potential purchaser for the reasons you stated. Better choices might be Qualcom or Samsung, but I really doubt either of them will be in any hurry to buy Xilinx.