If M3 Ultra exists and is a repeat of M2 Ultra (two M3 Max die connected together with an Ultra Fusion bridge), there is no more design work to be done. There may be post-silicon validation work remaining, however. Chips can't be production ramped as soon as first silicon comes back from the fab; before that happens a team of engineers has to extensively test all the chip's features to make sure no bugfix revisions are needed. (And then, when bugs are found, they have to work on figuring out whether there's viable workarounds or, absent any, how to make the most low impact change to the chip to fix the bug.)
The Ultra Fusion bridge adds a significant extra layer of complexity to post-silicon validation, which can easily explain why the products which require the Ultra chip (Mac Studio and Mac Pro) are delayed relative to the products which only require plain Max. Given that Apple already did M1 and M2 Ultra, they probably learned enough to get the M3 Ultra Fusion bridge working on the first try, but they still have to test.
Another factor which can play into delays like this: It's quite possible they're binning 'special' Max die to go into the desktop products. This could be Apple wanting to clock Mac Studio and Mac Pro higher than the laptops. But it could also just be Apple using desktop as a dumping ground for die which have worse leakage current and therefore too much power consumption for use in laptops. Either way, they might need to stockpile binned die over a period of time to support M3 Max/Ultra desktop launches.
