Apple skips M6 Pro/Max, fast-tracks AI-focused M7 to late 2027 as bandwidth race heats up

Apple will release a base M6 chip (codenamed Komodo) for entry-level Macs this year but skip Pro and Max variants, accelerating the high-end M7 family to late 2027, according to Bloomberg reporting and people familiar with the roadmap. This marks the first time since the M1 that a generation arrives without higher-end variants, pulling the M7 launch forward by as much as six months. The M7, codenamed Delos, is designed around on-device AI and heavier graphics workloads; Pro, Max, and Ultra variants (Andros codename) follow in late 2027, with an Ultra arriving in 2028.

The bandwidth numbers explain the rationale. The base M6 reaches approximately 200 GB/s memory bandwidth, up 31% from M5's 153 GB/s. Base M7 targets roughly 240 GB/s, a 56% bandwidth increase from M5. Memory bandwidth is the primary bottleneck for AI inference and on-device model execution. Current M3 Ultra already delivers up to 819 GB/s through die-fusion; high-end tiers, not base chips, carry the heaviest local-AI workloads. M6 pairs an upgraded Neural Engine with faster GPU (up to 12 cores vs. M5's 10) and CPU. Apple also plans an M5 Ultra (codenamed Sotra) shipping this year with 36 CPU cores, 80 GPU cores, and support for up to 768GB unified memory—a significant jump from M3 Ultra's 96GB cap, though memory availability remains constrained.

Apple pulled M4 Ultra altogether, skipping an entire generation at the high end, and has raised prices sharply across Macs and iPads. CEO Tim Cook blamed a "hundred-year flood" in demand for AI server capacity, which has diverted high-bandwidth memory allocation away from consumer hardware. Supply constraints are unlikely to ease in the near term. Apple has declined to comment on specs or roadmap timing.

For architects: the M6/M7 skip pattern signals Apple's shift toward on-device inference as a core competitive differentiator. Bandwidth-first design suggests expectation of larger, locally-run models on Apple silicon. The skip also reflects memory-supply scarcity—Apple is effectively throttling consumer generations to preserve capacity for servers and AI compute. Teams planning Mac infrastructure should model late-2027 availability for M7 Pro/Max and adjust capex timelines accordingly.