What baffles me is that Apple abandoned this completely sound theory for a risky (and entirely incorrect) bet on NPU hardware. They left OpenCL to bleed out, they simplified their GPU hardware to specialize functionality better, and ended up putting all their chips on the wrong bet. Now they pay OpenAI to run their models on Apple hardware, and apparently can't even do that without help from Nvidia too.
It'll be interesting to see what future generations make of Tim Cook's leadership. He started the decade so strong picking up right where Jobs left off, and ended the decade with several antitrust lawsuits, ballooning subscription services and the professedly failed launch of one Vision Pro headset. Perhaps Apple needs a decade to file a few of their own rough edges down again.
It wasn't a sound strategy. OpenCL was too little, too late, even though it was 15 years ago; CUDA was already dominant and OpenCL wasn't better in any way except being available on some hardware that wasn't as good as NVIDIA's, and OpenCL 2.0 a few years later was even more of a failure (NVIDIA basically refused to implement lots of new features and had enough leverage to force OpenCL 3.0 to make everything added after 1.2 optional). By contrast, Apple's NPU solved real problems for the iPhone in the domain of camera and computer vision features, operating within a reasonable power budget. Even today the NPU remains useful and superior to GPUs for some applications.
> they simplified their GPU hardware to specialize functionality better,
What features went missing, and when?
They haven't done that step for quite a while now. I feel like people keep repeating that story from years ago. But it doesn't hold anymore unfortunately...