Around 80% of Nvidia's revenue comes from gaming. The other 20% is split between HPC, professional visualization, embedded, and automotive.
Actual HPC and compute market revenue is only around 10% of total revenue.
Nvidia is very much producing hips for gaming as their primary revenue source. There is still enough crossover between a gaming chip and compute applications that gaming can subsidize Nvidia's exploits in other market areas. At the current time, HPC and automotive is growing faster than the rest of nvidia is trying to aggressively push compute technology, but it is still firmly planted in developing a gaming chip that can be used for compute applications, not the inverse.
Even Volta GV100 at 800mm^2 still has things that are only useful for gaming like texture mapping units, ROPs, geometry processing, video decoder and a lot of transistors supporting DX features etc. Given GV100's size it would have been great to thro away all those useless transistors but the deisgn is too heavily based on a gaming product.
I expect in time Nvidia will continue to diverge the lines but for the time being, Nvidia develop gamign chips as its main focus. So does AMD.