I’ve always considered this to be the primary factor in its widespread adoption into the mobile market.
I haven’t seen this mentioned and it’s a key part of it: a company designing a processing chip for a mobile device can license the ARM core, and plop that down onto as part of a chip that does all the other things that this application may need - you don’t have to have a separate chip just for the CPU. Reducing the count of the separate chips on a mobile device is very important. As far as I know, there was no way to license the design of an 8086 or a 6502 and integrate that within the larger chip that you’re building.
This is a critical point. Intel must realise this, and it isn’t clear they have any good counter to ARM on this. Atom will find it hard to compete with ARM when you seriously limited with creating an SoC. It isn’t just for current SoCs. Today’s multi chip device is tomorrow’s SoC, and no company is going to want to buy into an ecosystem that locks them out of many SoC solutions in a later product. Intel are looking at offering an SoC product, but it is going to be restricted to using only IP blocks that Intel owns, and an FPGA. This is pretty comprehensive, and Intel continue to buy other companies for the IP blocks as well. But, and it is a big but, it remains a clearly limited ecosystem, and one that is going to have a hard time against what is essentially the rest of the world as competition. It is not really conceivable that any of the other vendors would be happy with one of their IP blocks being delivered to Intel for integration into a bespoke SoC. Arm will clearly dominate the rest of the world, although I would suggest that no-one write off Mips just yet. Expect to see a lot of Chinese offerings in the future that use Mips.