Apple and NVIDIA have been in each other's bad books since the 2012 Macbook Pro incident where discrete GPUs of those devices got hot beyond their thermal specs and melted themselves out of the soldered sockets, which NVIDIA refused to compensate Apple for. I know, because I had one of those. Apple took the hit to repair those Macbook Pros even while out of warranty, even if it was ultimately NVIDIA's poor engineering. They've not worked with each other since.

2012 Apple was pretty much the height of the quality apex.

Apple and NVIDIA have spent the last six months fighting over whose fault it is that NVIDIA doesn't have working drivers for the latest OSX.

https://create.pro/blog/nvidia-drivers-for-macos-mojave-time...

At least AMD is working hard to upstream their drivers into the Linux kernel, which I don't doubt Apple appreciate when they merge the code into theirs.

Keeping in mind that apple is primarily a consumer focused company,

What's the it-just-works consumer machine learning use case that requires GPGPU on end user hardware?

Apple has been pushing machine learning at virtually every WWDC and its integrations from developers into product launches since.

A developer can do the ugly hacking to get an NVIDIA eGPU working on macOS, but it's a hassle.