They are the leader in manufacturing consumer systems with sufficient high-bandwidth memory to enable decent-sized LLMs to be run locally with reasonable performance. If you want to run something that needs >=32GB of memory (which is frankly bottom-end for a somewhat capable LLM) they're your only widely-available choice (otherwise you've got the rare Strix Halo AI Max+ 395 chip, or you need multiple GPUs, or maybe a self-build based around a Threadripper.)
This might not be widely recognised, as the proportion of people wanting to run capable LLMs locally is likely a rounding error versus the people who use ChatGPT/Claude/Gemini regularly. It's also not something that Apple market on, as they can't monetize it. However, as time goes on and memory and compute power gradually decrease in price, and also maybe as local LLMs continue to increase in ability (?) it may become more and more relevant.
All current use cases, the ones that caught the public eye, just don't have a need for locally run LLMs. Apple has to come up with functionality that can work with on-device LLMs and that is hard to do. There aren't that many use cases for it as the input vectors all map to an app or camera. Even then a full fledged LLM is always better than a quantized, low precision one running locally. Yeah, increased compute is the way, but not a silver bullet as Vision and Audio bound LLMs require large amounts of memory
This might not be widely recognised, as the proportion of people wanting to run capable LLMs locally is likely a rounding error versus the people who use ChatGPT/Claude/Gemini regularly. It's also not something that Apple market on, as they can't monetize it. However, as time goes on and memory and compute power gradually decrease in price, and also maybe as local LLMs continue to increase in ability (?) it may become more and more relevant.