The M1 being $50 is a bit suspect.
5nm wafers don't cost what 14nm wafers cost. I suspect that also doesn't cover the fixed cost R&D that Apple had to put into creating and deploying the Chips. Those $50 numbers are usually the result of ( wafer costs $xxxx and Apple gets yyy usable die out of the wafer so SoC costs YYYY ) or it is Qualcomms is $50 base cost so Apple's is
the same.
There is an article that 5nm wafers were priced around $17k.
TSMC’s Estimated Wafer Prices Revealed: 300mm Wafer at 5nm Is Nearly $17,000
High performance and high transistor density come at a cost
If Apple got 500 usable dies out of the wafer then that's about $34 just for the raw wafer of dies. Uncut , Unpackaged , untested , un-binned. And as I mentioned above the amortized fixed costs from R&D ( design, compiler updates , software firmware changes , etc. etc. )
Apple has their own semi-custom RAM chips while Intel/AMD certifications against standard DIMMS/RAM is all baked into the price.
The cost aspects that are past the profit margins and "sales costs" that AMD and Intel have baked into their prices ... Apple probably has pay those also. Apple will save some but it is probably smaller than that number.
The M1 probably shares a die with the A14X. But the very likely larger variants of the line up for the higher performing Macs are likely going to have to amortize over a much smaller volume base and higher production costs per die. (because there will be fewer useful dies off of a wafer).
But yes... if apple is out to make an extremly good SoC for he iPad Pro they have almost pretty much paid for something that can prune off the bottom two (and highest selling) Macs also (the two lower range Laptops). Throw in the lowest end Mini and probably at volume far greater than what the iPad Pro was at ( so fatter margins there now too).
How does AMD compete.... they aren't really trying to primarily push here. The order in which new Zen release have come in is roughly GPU-less desktop Ryzen , EPYC , Threadripper , APU. (CPU+GPU combo). This particular subrange of CPUs that Mac lower range laptops is last on their priority list.
How can they complete? pretty sure the folks who like to put in their own so-DIMMs are happier. "raw" boot multiple operating systems folks are going to happier.
Apple's move to M1 chips will save $2.5B this year – IBM exec - 9to5Mac
An analysis by a senior IMB exec suggests that Apple's move to M1 chips will save the company around $2.5B this year, with far larger savings to come ...9to5mac.com
I'd guess that the savings for making their own chips at the high-end are much greater than at the low end.
Not necessarily. The Mac Pro run rate is probably closer to 100K ( or less). Once the volume drops one or two orders of magnitude and the R&D costs go up because deviating more and more from the iA-series baseline for the design the fixed costs are going up and the die yields off the wafer go down. At the extreme end of the Xeon W 3200 series where Apple is using 'M' class and paying a $3K tax for accessing memory over 1TB ..... Looking at the M1 I doubt Apple is going to make anything like that. [ And AMD is lower anyway. ]
If Apple keeps monolithic dies with integrated GPU cores as they go "biggest" and AMD keeps chiplets for higher stuff in the Mac range. The could close a bit in that zone.
