Same thing happens on desktops. An M.2 slot relies on PCIe and SATA lanes from the chipset/PCH (Platform Controller Hub) which are limited in number. An example of a block diagram for one is here.
MB manufactures then have to (are allowed to) choose how they would like to distribute these lanes, some here some there, divided between SATA, T-bolt, USB3, Ethernet, and whatever else loves to have the speedy more direct communication lanes.
On many desktops the M.2 choice disables a couple of the SATA, for example, because they share lanes. The MB manufacture chooses where to distribute them, the user gets the choice of losing one for the other or visa versa. Some other manufacture might choose to knock out a few lanes for the other gpu slots.
Different CPU architextures (the cpu chip itself) have different quantities of lanes avialable. to further your question, the actual CPU that is in use would be interesting.
So that is the explaination you desired, the problem still is, there should be enough lanes available to cover these two items easily . Please provide all your specs, model cpu info in the question.
Other notes: Hubing the lanes out of the cpu is not without its own problems, there is minor slowdowns via the negotiation, and probably buffers used in the chips that set up for sharing of these direct lanes, which are of course no longer so direct.