There have been rumors that "M3 Ultra" would be it's own dedicated SoC and not two M3 Max using the Ultrafusion connector. That the M3 Max does not have the Ultrafusion connector is likely the basis for this speculation, though it is not beyond the realm of possibility Apple could add the UFC to the M3 Max to create a "traditional" M3 Ultra.
All of the above being said, it makes little sense for Apple to have delayed M3 on the Studio to now if they have an M3 Ultra (either as a unique SoC or as a dual-Max) when M4 is now available and is a not-insignificant upgrade from M3.
Is M4 available? In the unique die case , there wouldn't be M4 Unique until the production of M4 Max was over the capacity that the M4 Max were soaking up. Likewise notice that the MBA didn't shift to M4 right away. Again need substantially more wafers for the volume that the M4 cranks out.
M3 Ultra right as M3 MBA wafer allotment drops to zero wouldn't necessarily be uncoupled.
Or if Apple Private Cloud Compute soaked up M3 Ultras as a higher priority than the Mac line up.
For N3 (and smaller) with the wafer costs higher Max'es for laptops with a completely useless die allocation is wasteful where the laptops run 'millions' and the others don't run multiple millions.
Nvidia and AMD big die GPUs are on TSMC N4 variants because N3 had bumps in the road to high volume production. If an even bigger single die , the Max 3 could have made it out the door while 'something bigger" still had cost issues.
If it is an 'even bigger die' then M4-series (and N3E ) has big die issues also. In short, there are a couple of timeline hiccups that can occur if the "ultras" are decoupled from the laptop Max's.
Gurman has a lot of sources in the supply chain and those sources only know what they see and have little to no context as to why such things exist.
errr? Does he? Gurman has lots of "I think" , "I suspect". And occasionally "Apple sources says". But a wide variety of other parties?????? Not really. Kuo did supply chain for a long time. Gurman in his heyday seems to have someone in marketing or product testing on speedial. Gurman's 'sources' don't seem all that broad. Also seems that some of the Apple ones got exposed at some point also.
When Gurman mentions "suppliers" doe that come before or after Kuo , Display silo , etc other supply chain info leaks? (jumps on echo chamber).
So I would not be surprised if Apple has worked on a dedicated "Ultra" SoC using the M3 as a base that just did not work out as a shippable product (as we have heard happened with the "extreme" version of the M1 and M2).
I'm would not be surprised if they skipped M3 Ultra. Not because "M5 is just around the corner". But that they were on either M-even Ultra or M-odd Ultra . An Ultra for every generation seems like major problem. Even with possibly throwing in a 100K for Private Cloud compute ... the run rate is relatively small for the size/complexity of the chip to toss them in the garbage can every 12 months.
While Apple did a M2 after the M1 , the roll out of the M2 Ultra was very elongated. And M2 to M2+? has been even longer.