What if it lacks DIMM slots but supports this via PCIE slots:
www.anandtech.com
Apple has not joined the CXL consortium at all. To be of practical use CXL requires PCI-e v5 . Apple is still 'stuck' on PCI-e v4 . Apple largely used PCI-e v4 in the M1 series to reduce the number of PCI-e lanes provisioned out of the SoC (more edge I/O die space to be allocated to the memory controller paths). They treated x1 PCI-e v4 == x2 PCI-e v3 as a 'feature'. CXL support has to be built into the PCI-e controller and into the internal IOMMU subsystems.
CXL was initially driven by Intel. It has broad base support now. AMD even delayed their Zen 4 Epyc solution to pick it up.
But 2-3 years ago it wasn't as crystal clear that it was going to be as broadly supported as it is now. If Apple was not paying attention over the last couple of years and latched onto a "not invented here" attitude ( Intel stuff? Pftt. ) . CXL is not a path to better Perf/Watt. And if that is all that matters to Apple there is a pretty good chance they blew it off (even if they did some PCI-e v5 work. It isn't part of the PCI-e standard. It is another protocol the reuses the lower foundation for transport. Similar to how Apple has a "ssd-storage-PCI-e" variant they run between their SSD controller and their NAND modules. Not the same protocol but foundation circuits can just re-use relatively cheaply.).
The other issue is that CXL memory isn't really transparent. It is cache coherent , but latency sensitive apps will need to account for the NUMA impacts.
Nor is CXL memory going to be cheap. ( a significant fraction of the gotta have DIMMs slot grumbling is motivated by folks who want to buy more mature DIMMs at lower prices in the future as those who absolutely require more memory sooner. )
There is a decent chance that Apple actually views CXL as a threat to their "Unified Memory" value add. If so, they aren't really going to be highly motivated to help it mature faster and erase their advantage. If so they are gong to drag their feet as long as they can before market dynamics makes them put it in. (even though it would be helpful in getting around their RAM cap problem. ). PCI-e v5 has very extremely limited utility across the rest of their SoC line up. Probably not in a hurry to go there either. More likely to put more work into "Thunderbolt next" than that.
Apple hasn't done much with PCI-e rebar to do "more Unified Memory" stuff ( Intel , AMD , Nvidia , Microsoft have been doing more "smart memory" like things to varying degrees of success). If Apple looked at CXL and said 'we already solved the unified memory problem with our design and UltraFusion" then adoption is going to a bumpy road.