Er, they know they contradicted themselves, right? "This year there will be only these 3, because there will be other new chips next year." After saying it'd just be the one chip this year and the other 2 will follow later, next year.
Just the x800X3D chip being released this year seems most plausible. Definitely can see them announcing all to steal thunder from Intel's new chips. Will be surprised if many people can get even the one chip this year though.
Its also a shame that the x800X3D doesn't have a proper iGPU (even just the 16CU one of their current top APU would've likely been enough), as it would make it probably the most appealing chip across the board, as it should feature best overall gaming (iGPU and dGPU) performance for most, close to if not best efficiency across the board, while offering ability to scale up depending on power/cooling and performance across the board - 8c/16t would likely put it near the top of laptop chips in outright performance and would make it competitive on desktop as well). Such a chip I think would've nullified both Qualcomm and Intel's chips. It would've been an affordable halo chip.
I also wonder if such a chip would let them get away with using only the c cores whilst losing very little if any performance, so they could offer more cores, or could go with a larger GPU for the same die size. A 12c/24t Zen5c X3D chip with proper iGPU and then binned could've covered basically the same entire stack that Intel's ArrowLake-S will. If it let them up the CU count to easily take iGPU lead as well, definitely could've been worth it. Heck, it might would've been enough to not even bother with Strix Halo (which likely will need either 3D V Cache and/or embedded DRAM packaging to live up to its potential anyway).
A stack of 4 chips could've covered basically all of AMD's client market. Low end/compact chiplet that was Zen5c 6c/12t, smaller amount of vcache (32MB, but maybe even 24MB) and 12CU iGPU. Then a doubled up one. On desktop, do chiplets where you can do more balancing (i.e. when doing iGPU gaming, one chiplet is focused just on that so it gets the full cache for the GPU on that one while the other goes full CPU). Unless can get ok graphics/gaming scale up from chiplet (meaning, would have just the low base chiplet, then add chiplets up to 4 for 24c/48t), which would enable even more granular segmentation. I would guess X3D production wouldn't be able to meet the output (even at AMD's not even trying to compete with Intel's production level in consumer), but interesting thought as I think it could've really differentiated AMD this year and if they got economies of scale going to keep prices in check, would've transformed AMD's product stack from lackluster to much more interesting. Likely could've helped EPYC as well. It might would've alleviated Zen5's core latency, and boosted gaming performance (both iGPU and dGPU) while it shouldn't sacrifice what Zen5 does offer. But sounds like Zen 6 is where they get interesting with the packaging.
Would be really curious of comparison with embedded DRAM (will it make X3D mostly pointless by offering mostly what it does, like reduced latency but with more uniform memory addressing).