Apple plans to truncate its M6 generation to just a base chip, skipping the usual Pro, Max, and Ultra variants entirely. According to Mark Gurman's latest Bloomberg newsletter, the company will release the base M6 this fall, then follow with the AI-first M7 base chip a mere six months later in early 2027. The M7 Pro and M7 Max would arrive by late 2027, with the M7 Ultra targeting 2028.

The base M6 will push memory bandwidth to 200 gigabytes per second — up from 153 GB/s on the M5 — via a new memory architecture, a boosted Neural Engine, and 12 GPU cores instead of ten. But the M7 generation represents the real leap: base M7 bandwidth climbs to roughly 240 gigabytes per second, with architecture built from the ground up for AI workloads. Gurman notes the M7 Ultra could approach Nvidia Blackwell accelerator performance and support up to 1.5 terabytes of unified memory.

That server-grade capability is no accident. Apple is reportedly developing a dedicated AI server chip based on the M7 Ultra, targeting a 2029 debut. The current M5 Ultra servers arriving soon would serve as an interim step. By collapsing the M6 tier, Apple redirects engineering resources toward the silicon it believes will define its next decade of Mac and server products.

The compressed timeline signals Apple's urgency to match the industry's AI acceleration pace. Consumers buying M6 Macs this fall face an unusually short relevance window, while pros waiting for M7 Pro/Max hardware gain a clearer roadmap — if the schedule holds.