Definition
Plain language
Apple's own line of computer chips, built into recent Macs and iPhones instead of the Intel-style chips used in most other machines.
As stated in the literature
Apple's family of ARM-based system-on-chip processors (M-series and A-series), notable for unified memory shared between CPU and GPU; a deployment target where GPU-centric LLM serving stacks like vLLM don't map cleanly, motivating bespoke runtimes in VibeServe-style work.
Why it matters: Because these chips share memory differently from typical graphics cards, standard AI-serving software doesn't run well on them, so getting good performance on Macs requires purpose-built tools.
For example, running a large language model directly on a MacBook uses the chip's shared pool of memory so the processor and graphics unit draw from the same store instead of copying data back and forth.
Heard on the show
“New hardware — Apple Silicon's unified memory, custom accelerators.”Episode 027 — When AI Agents Build the Serving Stack: A Bet on Bespoke Infrastructure