On the contrary, I'll say that it does matter --- and that's why Edison failed. It was x86, but not truly "IBM PC-compatible". Those who didn't care about PC-compatibility were unlikely to choose x86 over something like ARM, and for those who did, the Edison was useless.
If Intel had chosen to put an entire "real" PC on the SoC with, yes, plenty of legacy peripherals and such so that it could --- with suitable I/O interfaces attached --- basically act as a lower-powered desktop or laptop, I'm almost willing to bet it could've turned out very differently. They could've found applications in things like this now-dead product, for example: http://www.pcworld.com/article/2873118/mouse-box-wants-to-st... (discussed at https://news.ycombinator.com/item?id=8931999 )
Intel's strength is the immense backwards-compatibility of x86 and the PC architecture, but in trying to make a not-quite-PC platform, they basically threw away their competitive advantage.
"Intel" has such a product - called Minnowboard Turbot:
> https://www.minnowboard.org/
for which just a few weeks ago a new quadcore version was released. The reason why I put "Intel" in quotes is that formally the Minnowboard is developed/marketed by ADI Engineering/MinnowBoard.org Foundation, respectively and sold by Netgate. But it is well-known that the Minnowboard project/foundation is backed by Intel.
It's an embedded system, not a as low-power desktop.
If you are using it like an ordinary PC then you are probably using it wrong.
But at that point you're usually looking at proper Linux-based boards with lots of standard IO - even HDMI output - in the ARM space, and Edison provides absolutely nothing over those. It might've provided something if you could reasonably treat it as a bog-standard x86 computer with some extra functionality attached.
No, you can't define other peoples use cases.
For myself, as long as the device meets my requirements (peripherals, power, tools, size, price) it can use power8 or PDP/11 for all I care.
In which case, you'll likely be using a significantly cheaper ARM board which does all the same stuff, uses less power, probably uses a more standard distro, etc etc. Which brings us back to - nobody wants an x86 IoT device. They don't fit in anywhere ARM doesn't fit better.
First of all, you don't know my requirements yet you declare ARM winner. What if for my particular use case Tensilica is the best choice?
Furthermore, you generally cannot define what IoT means for other people. For one person it could mean a 8-bit garage door opener, for someone else it could be an octa-core 64-bit monster.
Finally, Intel has a simplified x86 design with very good power usage for use in IoT. This CPU is not used in Galileo and alike today, but it exists.
Actually, no, not when it came out. Other wifi devices were about the same price or even more expensive, and harder to use. The ESP8266 came out about the same time, but it was still a long time before hobbyists were able to use them, especially without another device to control it. I think they had a good market position when they started, they just didn't see the ESP coming.