Apple's New M6 Chip Is Its First Built on 2-Nanometer Tech — And the M5 Ultra Fuses Four Dies Into One

The M6 claims the world's fastest single-threaded performance while the M5 Ultra, Apple's first quad-die chip, delivers 4.5x the peak AI compute of the M3 Ultra.

Apple has introduced the M6 and M5 Ultra chips, both aimed squarely at AI workloads alongside general performance gains. The M6 is Apple's first chip built on 2-nanometer process technology and the first to combine three distinct CPU core types in one design: two "super cores," four performance cores, and six efficiency cores. Apple claims the world's fastest single-threaded performance for the chip, up to 1.2x faster multithreaded performance than the M5, and a dual 16-core Neural Engine delivering roughly double the peak AI compute of the previous generation. Memory bandwidth rises to 170GB/s. The M5 Ultra is the more striking engineering story: Apple fused two dual-die M5 Max chips together to create its first-ever quad-die chip, aimed at handling the heaviest AI and compute workloads Apple silicon has faced. It packs up to a 36-core CPU, up to an 80-core GPU, and 1.2 TB/s of unified memory bandwidth — 50% more than the outgoing M3 Ultra — with peak GPU AI compute rated 4.5 times higher than that chip. The pattern worth noting is how explicitly "AI compute" has become the headline metric for Apple silicon launches, ahead of the general performance numbers that used to lead these announcements. That's a real shift in what Apple thinks customers — and increasingly, on-device AI applications — actually need most from a new chip generation, and it puts Apple's custom silicon roadmap on a more direct collision course with the AI-accelerator conversation that's mostly been about Nvidia and cloud infrastructure until now.