if x86 only focus only server and high margin, it is going to be another IBM.
Leaving $5-$10 chips on the table is the most dangerous things for x86.
MIPS and PowerPC will not penetrate unless management drastically come up with some real value...
It will be quite an interesting battle between x86 and ARM. I could certainly see Intel eventually getting down to the tablet and smart phone levels of power consumption and cost - or at least close - but I can't see them going below that.
The cut off point between "microcontroller-based embedded systems and sensor nodes" and an IOT device will be up for a debate that will influence this market quite a bit. For example, if a device consists of a small 32-bit ARM and a WiFi module, with a total bill of material cost of $15.00, is it too small to be considered IOT.
What if that device is connected to your garage door so you can check status and open and close it remotely? I'd call that an IOT device even though it would quite likely be possible with 8 or 16 bit MCUs.
Intel has a long history in embedded controllers. And there is no reason to think that extreme low power is needed for the IOT. Last time I checked my cable box, TV, receiver, fridge and toaster were plugged in all the time. In fact, I am having a hard time coming up with anything running on battery power in my house that needs to communicate other than phones, tablets, laptops.
My point isn't that Intel will not be a success in the IOT or that IOT devices in general don't need a lot of power. There's a huge spectrum and I believe that x86 will take quite a large swath of that spectrum of IOT applications.
The point I was attempting to make is that if we take 8 and 16 bit devices out of the IOT classification, we are cutting out a very large portion of the IOT. It's down in that level that I don't think x86 will play a part.
It's important to note here that IDC is talking about Intelligent Systems which it defines as Net connected, secure systems that can run an OS.
This includes much of what is IoT but does not include the sub-32-bit devices and sensors others put in the IoT category. It does include whopping embedded comms infrastructure systems where Intel has a good leg up and ARM is beginning to find sockets.
I still don't understand what these 11 billion things are that need to be connected together. Adding internet connectivity to my fridge or toaster would add cost to them for no obvious benefit. Adding a sensor to my garage door is all very well, but for it to be able to open and close itself would take several hundred dollars worth or mechanical and electrical engineering.
The only things I can think of in my house that would obviously benefit from internet connectivity are my electricity and gas meters, but in order for that to happen, I'd need new meters fitting at a cost far greater than the IoT functionality.
What are all these things that need to be connected together?
Interesting. Not a huge market, though. I don't know what the total wordlwide market for new cars is, but it is just over 2 million here in the UK (for a population of 63 million), so about 1 per 30 people. I guess if I've got one in each car, one in my garage door, it might just add up.
Actually, one of my cars is already connected to the internet: it is leased and I can log onto the leasing company's website and see sorts of stuff, including where I've been. As can the leasing company, of course. I don't know what the physical layer is - I assume GSM/3G.
So the annual USA market for garage door openers is 4B employing 53K (per IBIS) people. Some may add IoT at current $15 retail to become aware of a door opening via an email to smartphone. If you had an IoT concentrator or used an 8bit with 6LoPan it could be $5 retail per node. When the cost gets low enough, this opens up more apps and volume.