>Therefore, they simply can't just put a GPU on every base station around the world.
I dont think that is what's happening here. Base station has been power limited for quite some time and part of the whole 5G / Cloud RAN promise was moving a lot of the processing off base station. Ignoring GPU there are a lot of the current stack fits Nvidia portfolio especially DPU. May be Nvidia have figure out a way to use CUDA and have it perform better than Ericsson and Huawei.
Nokia is also the smallest of the three and has been in decline for quite some time. Part of me also wish Nvidia just buy Nokia and start competing against Ericsson and Huawei.
You are right on your points. I agree on DPU but it's on the network stack. I think Nvidia wants to get into the PHY and MAC layer (CuPHY, etc.). That's where I find it unlikely due to the cost of latency. If Nvidia had wanted to buy Nokia, they could've already completed the deal. It's a possibility in the future but this $1B investment kind of showed that they are more interested in creating artificial demand for their GPUs rather than diversifying their product portfolio. I agree with you that Nvidia should just buy Nokia.