The vision pro is not 'huge'. Lots of substantially costly parts there are unique to Vision Pro. M2 is used elsewhere but the R1 is absolutely not. Screens? not. Headbands ? not. battery pack not. gesture control with hand ... not.
It may not be selling in iPhone numbers, but it's not a cottage industry: the rumoured "disappointing" sales figures are still in the hundreds of thousands - it's not like they're just making a few hundred prototypes. Plus, it is likely that many of the companies that manufacture those parts
also make iPhone parts (or would like to) so Apple is in a strong bargaining position.
The question is still
how anybody can reliably know the cost prices of those custom components.
Similar to how folks thought Apple could sell 9-10x less SoC processors than Intel and still think the SoCs were going to come out much chearper than Intel CPUs for laptops/desktops.
Intel's sales may be 10x larger in total, but they're spread over
scores of different SKUs with different core counts, cache, TDP, GPU options (and that's just the i-series, not counting the Xeons and budget celerons/whatever) ... and Apple weren't always buying the most popular models.
Apple are only making 3-4 fundamentally different SoCs per generation - regular, Pro, Max and Ultra - with a few variants obtained by disabling cores or fitting different RAM chiplets to the package - and who knows what their financial arrangements with TSMC are?
Apple may be lagging way behind HP, Dell and Lenovo int terms of PC sales - but have you
seen the mess of different models and ranges on the Big 3's websites - all with different CPUs, GPUs, RAM types, screens keyboards...? It makes you nostalgic for the Mac Performa! Apple's Mac sales are spread over a much smaller range of actual models, so they should be getting pretty impressive economies of scale on the parts for each individual model. Plus, I bet they do it with 72.5% less committee meetings than Intel...
Apple moved from Intel to M-series chips without significantly increasing prices - so it doesn't seem very likely that M-series are costing them significantly more than Intel.
The M-series reuse some cluster function units , but the on chip network, die masks , level of verification difficulty
Each generation is using the same CPU, GPU core designs, media engine, TB/USB/Display drivers etc. some of which has also been developed for the iPhone, and while, no, you don't just copy and paste those to make a new chip it does mean that large chunks of R&D have already been done. Plus - although this has changed with the M3 - the M1/M2 Pro, Max and Ultra shared exactly the same die design. True, they had to be manufactured separately, but if there weren't significant time & cost savings from re-using designs like that
someone wasn't doing their job.