The bit on backwards compatibility concerns me a little.

Hopefully, they realize that they need to focus more on powerful hardware.

Another interesting point, they only made a deal for one (not two) APUs. I assume that will be the handheld and a dedicated home console is not even in early design yet.

I agree with a lot of what you've said. The only thing holding them back is that AMD don't have an ARM chip due until 2017, so we might not see NX until 2017/2018.

The Wii U system was under performing from the jump and much like the Wii, the extra care needed to make titles for it gave a number of developers and publishers the excuse to abandon it. The GC is in a similar boat, but its primary weakness was due to the medium size forcing either massive compression, cutting out material or paying for multiple discs which toward the end of the GC's life time hurt the ports.

Both system's had issues that made it harder for ports compared to the simple cut and paste you sometimes see with the PS and Xbox lines since the PS2 era. Nintendo has made decisions since the N64 that have made this much harder with their devices to be ported to, there isn't really a strong basis to say third party titles wouldn't sell on a Nintendo system because we have yet to see a decently powered device with a solid base try to sell them. The Wii U is the first Nintendo system that was strong enough to house most major 3rd party titles without constraints or major control issues, however the install base wasn't there due to Nintendo's failure to sell the system and relaying too heavily on the previous Wii's brand to try to sell it. Not to mention it basically pulled a Dreamcast, coming out first of the generation slightly more powerful then the last generation but showing nothing to display itself as 'next-generation' or a must own over just buying the other titles on cheaper similarly powered systems. It is why the Wii version of Just Dance I think still outsold all other version even when it went multiplatform.

I do agree, Nintendo can't really afford to try to cater to getting major third parties back since at this point there is probably a huge divide and no trust between all participants, which makes for a dangerous and stupid gamble in any business. However, they should not discount potential 3rd party support or build another system that will alienate it from getting anything down the road.

Which is the reason it won't be ARM. To delay a product one or two years waiting for semi-custom designs to be ready is unthinkable.

AMD has tapped out only Zen/Arctic Islands designs for 2016 on 16/14nm FinFets. It's a given the NX will release in 2016. So yeah.

If they were to take ARM it would be from someone else.

AMD is still planning to release an ARM core in late 2015, namely the AMD Opteron A1100 with ARM Cortex-A57 cores (no GPU). They sold dev kits back in 2014.

http://www.kitguru.net/components/cpu/anton-shilov/amd-delays-its-first-arm-based-opteron-seattle-cpu-to-second-half-of-2015/

http://www.kitguru.net/components/cpu/anton-shilov/amd-begins-to-sell-arm-based-opteron-seattle-kits-to-software-developers/

That is what most people have been saying since Nintendo merged their handheld and console hardware departments a few years ago.

That's a server chip though so the question remains, how useful that'd be in the design of a custom chip for the console space? Especially when it needs to be combined with a GPU.

That's an internet rumor/theory. Nothing more, nothing less.

They've got 3 new designs in the works. 

Two were announced last year, one being x86 and the other being ARM. One of these is expected to be gaming related and start being produced in the second half of 2016. One is expected to go beyond gaming but also production will start second half 2016. 

One announced this year. It is unknown when it is going to be produced. 

All 3 combined will be expected to bring up to $1.5 billion in revenue for AMD over the lifecycle of each.