I think the biggest reason this has such a huge performance advantage is that the 8/16GB RAM is built into the chip. Modern CPUs are mostly limited by memory bandwidth rather than compute performance, and this has way more memory bandwidth than competing CPUs because of the tight integration. This RAM might also be clocked much higher than what we're used to, because there is no long bus to the CPU anymore.

The downside of this SoC design, though, is that while you can fit 8GB or 16GB on a chip, it might be difficult to fit more. The 8/16GB limit might explain why this design is reserved to the smaller laptops for now, and they haven't replaced x86 in all of their lineup. If you want more RAM than that, then you would again be stuck with less memory bandwidth to your external RAM. You would maybe end up with a design where some applications are kept in the internal RAM and some in the external, or where your internal RAM acts as an L4 cache to the external RAM.

It's not that difficult IMO for Intel or AMD to replicate this with an x86 design. They might not have to modify the CPU core and its caches that much, mostly the memory controller. How much time they would need though, I'm not sure. There's some probability they were aware that this was coming and already had something in the works. Otherwise, maybe one or two years?