As far as I'm aware Apple M1 chips are not an order of magnitude (much less two) larger in area, layer count or resolution vs Intel/AMD or similar ARM chips. Thus it's highly unlikely for them to cost 100x more to produce. 2x or 3x more expensive I could believe, 100x... nope.

Defects in CPUs (not GPUs) for segmentation hasn't really been a thing for a while, maybe the Phenom X3, and that was already a bit of a stretch from what I recall as most X3s had a perfectly functional 4th core.

A cheap cellphone SoC is definitely not pennies, <10$ is a better estimate and it's usually without RAM included which is not always accounted for in comparisons.

I vastly agree with rmah and most estimates do too.

ARM is the Microsft of the semiconductor world. The company design and licenses CPU cores for several market segments, including: ARM cortex-M series cores intended for MCU - microcontrollers SOC (System-On-Chip); CPU-cores intended for embedded systems processors (MPU) such as ARM-Cortex-A series intended for embedded systems and mobile devices. And there are also ARM cores designed for servers, for instance A64FX Fujitsu chip that powers the Japanese supercomputer Fugaku, until now the fastest in the world. ARM cores highly customized with microarchitecture, such as Apple M1 and some Samsung ARM chips, are designed under architecture license.

A MCU is a self-contained chip containing up to 1MB of RAM, Flash memory up to 1 MB for the firmware; and several other peripherals such as digital IO, analog-to-digital converter, UART RS232 and so on. They are not comparable to Intel chips. MCUs are used in keyboards, mouses, car engine control unit, toys, home appliances and so on. The difference between the cores intended for MCUs and cores intended for MPUs is that cores for MPUs need external flash memory chips and external RAM memory. MPU chips and also they contains MMU - memory management unit for supporting virtual memory and operating systems that need this feature, such as iOS, Windows CE, Linux, BSD and etc.