Apple just dropped its biggest chip update in years, and honestly, this one’s worth paying attention to — the new M6 chip is faster, smarter about AI, and squeezed onto a manufacturing process so tiny it’s genuinely impressive engineering.

M6 is Apple’s first chip built on a 2-nanometer process, which basically means engineers packed way more transistors into the same tiny sliver of silicon. Think of it like fitting more rooms into an apartment without making the building any bigger — that’s the kind of density jump we’re talking about. The result is a chip that’s not just faster, but noticeably more power-efficient too.

The star of the show is a beefed-up 12-core CPU — up from 10 cores in the M5 — split across 2 super cores, 4 performance cores, and 6 efficiency cores. Apple’s claiming this setup delivers the fastest single-threaded performance of any chip it’s made, plus up to 1.2x faster multithreaded performance than M5, and a full 2.4x jump over the original M1. If you’re someone who compiles code, edits photos, or runs AI workloads regularly, that’s the kind of gain you’d actually feel day to day, not just see on a spec sheet.

Here’s where things get interesting for anyone tracking the AI hardware race. M6 packs a Dual 16-core Neural Engine, which doubles the peak AI compute compared to previous generations. Apple says its system software can tap both Neural Engines at once, so apps that lean on on-device AI — think local chatbots, image generation, or smart auto-complete — should feel noticeably snappier.

The GPU side got love too. M6’s 12-core GPU (again, two more cores than M5) now has a Neural Accelerator baked into every single core, which Apple says translates to nearly a 30 percent bump in AI compute over M5 and more than 8x compared to the original M1. In practical terms, that means faster prompt processing when you’re chatting with an on-device large language model — a feature that’s becoming a bigger deal now that more apps are running AI locally instead of pinging the cloud.

Memory bandwidth also ticked up to 170GB/s, a 10 percent gain over M5, and the chip now supports up to 32GB of unified memory — enough headroom to run local LLMs for private, on-device AI tasks without needing an internet connection.

Meet the big sibling: M5 Ultra

If M6 is Apple’s everyday workhorse, M5 Ultra is the chip built for people who need serious horsepower — video editors, 3D artists, researchers running massive AI models. Apple is calling it the most powerful chip it’s ever made, and the numbers back that up.

The clever bit here is architecture. M5 Ultra uses Apple’s UltraFusion technology to physically weld together two dual-die M5 Max chips, creating a quad-die design — a first for any Apple silicon chip. That interconnect moves data between the dies at over 4.4TB/s, fast enough that the four dies behave as one seamless processor rather than four separate chips awkwardly stitched together.

What that buys you: up to 36 CPU cores, an 80-core GPU, and a jaw-dropping 1.2TB/s of unified memory bandwidth — 50 percent more than the previous M3 Ultra. Apple’s also touting up to 512GB of unified memory, which is the kind of headroom that lets researchers or developers run massive AI models with hundreds of billions of parameters entirely on their desktop, no cloud required.

Both chips land first in updated hardware: M6 debuts in the new Mac mini, and M5 Ultra shows up in the refreshed Mac Studio. Apple’s also leaning hard on its developer story here — frameworks like Core AI, Core ML, and Metal are built to tap directly into these Neural Accelerators, meaning third-party apps (not just Apple’s own software) should benefit from these AI gains fairly quickly.

Zoom out a bit, and this release fits a pattern we’ve seen across the industry — the AI PC/Mac trend has shifted from marketing buzzword to actual hardware differentiation. With rivals racing to add dedicated AI acceleration to their chips, Apple’s move to double down on on-device Neural Engine performance and massive unified memory feels like a direct response to the growing appetite for running big AI models locally, rather than shipping every query off to a data center. For a company that’s made privacy a core pitch, that on-device AI muscle isn’t just a spec bump — it’s a strategic bet.