Apple Intelligence, PCC, and M4 Macs
In Apple’s announcement of its AI service, the company proclaimed most of the computing will be done locally, as in on the device level. Unlike its competitors, Apple Intelligence is supposed to only communicate with the servers when it needs to, and the processing will be done privately on PCC (Privacy Cloud Compute), a data center Apple is building out of Apple Silicon chips. It is no surprise Apple decided to use its in-house chips to build a data center, but one cannot simply stack ARM chips to power an LLM.
This is where the interesting rumors and speculations start to pour in —Apple might be in building the data centers using ordinary Mac desktop lines. Mac Pro is offered in server rack size since 2019, and Mac Mini and Studio lines are known to be powerful enough to replace minor server tasks. Considering the size of the operation Apple is undertaking, it is rather unlikely there is a server farm stacking Mac Minis. However, it hints there will be a demand for Mac Pro, hence more the reason for the company to invest into ‘serious’ desktop lines.
On more personal note, I would be interested in hearing more about Apple’s workstation and server lines of products once again. Apple discontinued OS X Server quite awhile ago, and the server app a user could install on the machine after the fact was also discontinued. Existing Mac Pro line has little to no advantage over Studio line aside from PCIe slots. And of those PCIe slots, Apple has discontinued support for dGPU. Apple’s presence in pure computing or cloud computing is no more. If Apple decides to take the ARM server torch and scale it up to the point where it would be competitive, —Apple’s proprietary hardware didn’t always succeed— we would definitely see a new ecosystem of desktop computers settle in. Until then, PCC would remain as Apple’s Siri server farm powered by Macs.