Freescale isn't doing that bad, being the lead supplier of automotive chips for instance. Almost every modern car is running on a Freescale processor running linux.
Just like AMD, Freescale sold off their fabs, which was probably a wise decision. Reason: they can compete more on price, regardless of the technology of the products. The fabs themselves benefitted also from this: they are not owned by US companies any more, but by Arabian, Asian and European companies, that have a more long-term strategy instead of being terrorized by stock dividend races "pour la beauté du tableau", and that really boosted their potential, and see what happens now: even the most modern fab in the US in New York is built by GlobalFoundries, which is owned by ATI (the irony right? ATI here stands for Advanced Technology Investment), which is owned by the government and mostly citizens of Abu Dhabi and European investors.
This has given AMD time to breathe, and they have used their time very wisely. They have bought ATI (the former Canadian graphics card manufacturer) and SeaMicro, and they have a really good talent portfolio. The key to their strategy is the concept of the AMD APP, aka AMD "Fusion", aka "AMD Hybrid Architecture", which is the first next-gen PC platform of its kind. It focuses on next-gen operating systems (mostly linux), and is open sourced under AL2.0 license. I posted the link where you can download the AMD APP toolkit in another thread earlier today. It's a hugely successful concept that is embraced by the markets, because it brings HPC within reach of small businesses and consumers. AMD has been designing their graphics cards as GP-GPUs for a while, and came up with the CGN architecture, which in terms of gaming graphics performance on legacy platforms can still keep up with nVidia, and often even surpasses it, but at the same time delivers all the benefits of the better integration of the GP-GPUs in modern systems, and, AMD can even recycle some of the tools of the AMD APP toolkit on legacy systems, like Mantle, which is like a small part of the AMD APP linux toolkit that is used to accelerate legacy Windows/PS4/XBone consoles.
The selling of the fabs to non-US owners that are capable of long-term thinking, has enabled AMD to trade with non-US companies better (Nintendo, Sony, with a focus on straight deals), and still to trade with US companies like before (Microsoft, US companies focus more on "ersatz-compensation" deals than straight money deals), and the end result is that they shift a lot of chips to the market, one way or the other, and have Intel and nVidia peeing their pants in more way than one.
The AMD APP technology being there and being implemented in real-life existing systems that anyone can buy for a fraction of the cost of any other comparable performance solution, is a thorn in the side of Intel, that has been working on iGPU and Xeon Phi integration for a very long time, but can't seem to get it right. i-GP-GPU is the most simple thing ever with AMD APU's, but Intel can't even seen to get OpenCL working on its iGPUs, so that's a big problem for them, because the future is not making smaller litho chips that use less power, if at the same time, that does nothing for the total system performance. AMD isn't focusing on the CPU in delivering that performance, but on the CPU+GP-GPU or CPU+i-GP-GPU, aka APU. Those little SeaMicro boxes cost less than many gaming PC's, but they are true HPC wonders, and everyone can have one, use them, benefit from them.
AMD has been merging a lot of code into the linux kernel, and everybody knows that's the platform that matters for the future, even Apple is talking about open sourcing OSX to benefit from Linux and BSD evolution for their future products. AMD doesn't have to make a lot of noise about it, their APP is hugely popular with open source projects, indie game devs and open source game engines are being worked on to use the power of APP.
As a refernce line: with APP, the overall performance of a standard system with AMD CPU and AMD GP-GPU, whether as discrete components or as APU, rises on average 40%, that's without overclocking, and not a buck is spent on extra hardware. That means that this is a much more efficient route than making smaller and more process-specific optimized CPUs and GPUs, which is what Intel and nVidia have been doing. The Phenom II X4 for instance performs just the same as an Intel Core i5-750K, and outperforms when both are overclocked, due to better overclocking capacity due to the larger die and litho. Intel now has CPUs in the Core range that perform 15% better, maybe with cherry picked Haswell chips, even 20% better. AMD can improve the overall performance with 40% just like that, without users having to buy new hardware, without overclocking, and up to 500 % performance increase in compute applications. They have been silently following the route that IBM has been following for a long time, the route of "Many Core Processing", which is the next-gen computer technology. Another example: Intel and nVidia die-shrinks have made their chips more power efficient, but it has come at a price of less burst-performance improvement and less overclockability, and a more difficult binning process. AMD has been working on DPM and zero-core abilities, and although core-per-core their chips use more power, they can also shut cores off completely, and in the end, overall, there will not be more power consumption on a system level than with smaller lithos and dies, but there will be more performance when it counts.
So I don't think that AMD is going to die any time soon. They may be bought up in the long run by a non-US company as a strategic move, but they definitely have the edge of next-gen computing now in comparison to the others, and they are doing well, just like IBM.