I mean, having more CUs doesn't instantly mean it's "harder" to write code for. It's whether or not the developer wants to optimize their code for more parallel processing, when it comes to PC GPU's, the cards you see with more CUs are usually the ones designed with "pro" work in mind (software SPECIFICALLY designed with lots of parallelism in mind are what's really gonna take advantage of a card with like 60 CUs, games not so much
if they're not programmed that way). MS even referred to their CPU cores as "server-class CPU cores", Series X's APU is gonna be used for their cloud servers to run four Xbox One S sessions in
parallel, you can tell why they chose to go with more CUs from the beginning. Both companies have different priorities here.