That's a big part of why the release of Max/Ultra on desktops would be a reasonable move to get some more revenue flowing in during Q3 as well as building anticipation for M4 through the remainder of the year going into a wider autumn release of M4 laptops. If the desktops don't sell as many units and have a higher profit margin compared to laptops, putting out the desktops while chip yields are still being ramped up allows the technology to be making Apple money earlier in the cycle as they prepare for the much wider release of the chip on laptops several months later.
"If" the desktops doesn't sell as many units? That really isn't really an 'if'. When Apple last stopped reporting laptop/desktop split it was close to 70% laptops. Things have only gotten worse since then. Not just in Mac space but in Windows space also.
The very upper range desktops that cost so much are in likely the 1-2% of unit sales range. The MBP is likely is the 15-20% range. That is an order of magnitude difference desktops would have to 'make up' in margin percentage.
The flaw here is that the MBP 16" has high margins also. The CPU/GPU die likely isn't the biggest margin grabber. Apple's mark up on RAM and SSD capacity is substantially high. As long as they can get substantively more laptop buyer to bulk up on RAM and internal SSD space the margins are going up, but the costs are mainly staying level with the capacity increase. Versus an Ultra Package where the costs are gong up for extra fancy 3D packaging and matching twin large dies. (and extra PCI-e stuff Apple needs to toss in to keep MP somewhat differentiated. ) .
The MP actually helps users avoid Apple's high SSD capacity markup ( going to 3rd party internal SSDs in a standard PCI-e M.2 carrier card. )
If the M4 wasn't already shipping (and generating revenue) then possibly a large die could be a 'pipe cleaner' for the fab process , but the iPad Pro is already shipping. So the 'pipe cleaning' is already happening. ( And none of the rest of the M-series line up does any 'pipe cleaning' for the Ultra's packaging pipe cleaning since they don't use it at all. ). The 3D packaging probably has a defect rate (yield growth) factor also.
With the current bubble of mega AI packages , 3D packaging is also a chokepoint. TSMC is scrabbling to keep up with demand there. Nvida, AMD, others are throwing giant wads of cash at it. ( Apple isn't going block folks by buying it all out). TSMC N3E is a bit quirky in that the "Max" die could get even bigger where Apple could bust the recticle limit for InFO-LSI ( and "extra" CoWoS-Lsi capacity has completely evaporated. )
Periodically, Apple could possibly do a 'bigger first' roll out. This cycle isn't one. If the rest of the line up is trying to digest M5 generation then perhaps the top end could adopt M6 'early' (having skipped a gen). If N3B hadn't had hiccups and Apple could have pulled the M3 Max up to Spring 23 then perhaps. It is far more better aligned for one ot the TSMC 'incremental refinement' generation ( N3P or N2P ... 2nd or third iteration after a larger node shrink. N3E regresses ( on density) almost as much as it moves 'forward' (on power). To label it 2nd gen is slightly misguided. It is really the start of a library design compatible progression 3E -> 3P -> 3... N3B is pragmaticvally forked off by itself. There is a '3' in there, but it is substantially different. )
I'm sure you know more about what you mean here than I do, but my assumption would be that Apple's hardware engineers take all easily foreseeable (and probably a lot of not very obvious) internal capacities and complexities into account during the earliest design phases, in order to avoid missteps wherever possible that might cause basic incompatibilities or slowdowns. It would seem to be an expensive mistake to get this far down the line in development of M4 lineup without having solved things like making sure all the variations scale to the rest of the hardware used.
It isn't about "incompatibilities and slowdowns". Apple Silicon needs to delivery at some target cost. If 'too expensive' SoC raises too high then substantially not going to be same product target market anymore either. It is about the product segmentation that their customers (the Apple Mac product managers ) are asking for. The laptop folks would like to grab more users out of the 'legacy desktop' userbase. For example someone who bought a MP 2008 could now buy a MBP 16" + XDR/Studio Display and have enough to get work done. The Mac Studio covers what an Intel MP 2013 did. The Mini Pro or MBP 15" covers the computational workload the 27" iMac segmented out of the user base.
Generally, Apple is trying to put more performance in a smaller container. But the smaller container comes with self imposed constraints. ( Apple likes magical cooling with no clearly visible vents. ). Since laptops is most of what they sell , they are driving scale that maxes out in the laptops.
The RAM and SSD capacitiy has to do with more die edge space which only comes with bigger dies. The plain Mn needing to fit in an iPad Pro (thinnest Apple device ever) will cause issues. Although Apple likes to talk about how Apple Silicon allows them to sell the products they want to, the higher development expense of these SoCS also entangles the SoC into multiple products to keep the overall costs down. Apple's "poor man's HBM" via relatively very wide LPDDRs aggregate memory width drives that. And it a semi-custom memory packages so have to sell more memory package units to keep their costs lower. (similar with the 'nobody else has' resolution laptop screens. )
