Indeed this is not entirely obvious. I would speculate this is as much about aquiring the talent inside of ARM and use that for the benefit of Nvidia. As well as setting the strategic direction of ARM itself to benefit Nvidia long term.

They probably want to compete with AMD which has greate talent both for desinging CPUs and GPUs. Aquiring ARM is a way of also getting a great CPU team.

There may also be a sort of Microsoft style embrace and extend strategy.

Shift the GPU IP used in ARM CPUs to match Nvidia, then make it so that anybody who wants the high end versions of this will have to pay up to Nvidia to get it. It could be about solidifying CUDA as an industry standard of sort.

I guess there are many potential angles to this.