Apple Skips M6 Pro and Max to Fast-Track an AI-Focused M7 Chip for 2027
Mark Gurman reports Apple is rewriting its chip roadmap around on-device AI - skipping an entire high-end generation to get there faster.
Apple plans to skip the M6 Pro, M6 Max, and M6 Ultra entirely, jumping straight to an Apple M7 chip lineup for high-end Macs in 2027. Mark Gurman at Bloomberg broke the story June 25. Entry-level Macs still get a base M6 chip as early as late 2026, but developers on M5 Pro or M5 Max machines face a longer upgrade cycle than Apple's typical two-year chip cadence would suggest.
Apple Ships Base M6 This Year, Skips the Pro and Max Variants
Apple did not scrap the M6 entirely. A base variant still goes to entry-level Macs - think MacBook Air and low-end Mac mini - arriving as early as late 2026. What Apple cut is the M6 Pro, M6 Max, and M6 Ultra: the three high-tier variants that normally follow the base chip within 12 months and power the MacBook Pro, Mac Studio, and Mac Pro. Skipping all three in one move has no recent precedent in Apple's silicon history.
Apple M7 Chip Targets 240 GB/s Bandwidth for On-Device AI
On-device AI drives the decision. Apple M7 chip's architecture centers on memory bandwidth and Neural Engine throughput, with Gurman reporting a target of around 240 GB/s - roughly double what the M4 delivers at about 120 GB/s. For developers running local large language models or diffusion pipelines, that bandwidth directly determines how large a model fits into memory and at what speed.
Running a 7B parameter model locally on an M5 MacBook Pro already works. Speed is the bottleneck. Higher memory bandwidth means models load faster and generate tokens quicker - getting closer to response times developers expect from cloud APIs, but without per-token billing. For teams building agentic AI workflows that chain multiple model calls together, local inference at 240 GB/s starts to look viable for production-adjacent testing.
Qualcomm's Snapdragon X Elite laptops have already claimed strong on-device AI performance on Windows, creating competitive pressure Apple cannot ignore. Apple silicon's unified memory architecture - where CPU, GPU, and Neural Engine share the same memory pool - means bandwidth gains benefit all three simultaneously, which is a structural advantage Qualcomm does not replicate with the same efficiency.
M5 Pro and M5 Max Owners Are Looking at a Two-Year Wait
M5 Pro and M5 Max buyers face an unusual position. Apple's standard cadence puts a new Pro/Max chip every 18 to 24 months - but with M6 Pro scrapped, the next Pro-tier chip is M7 Pro, which Gurman places in 2027. For developers who bought a MacBook Pro in late 2025 or early 2026, the smartest move is probably to hold - a better upgrade is coming, just not on the usual schedule.
M7 Ultra, the chip that typically powers the Mac Pro and top-end Mac Studio, could arrive as late as 2028 by Gurman's estimate. Apple has not confirmed any of these timelines publicly.
Apple Intelligence Features Hinge on What M7 Can Handle Locally
Apple Intelligence - Apple's on-device AI suite across iOS and macOS - runs on the Neural Engine inside every M-series chip. Local matters more than ever. Apple rebuilt Siri with Google Gemini at WWDC 2026 partly because on-device processing for advanced language tasks pushed past what current Apple silicon could handle cleanly. M7's AI-first architecture suggests Apple plans to route more of its intelligence features back toward local processing rather than through external servers.
M7 Pro and M7 Max arrive in 2027. Apple has not announced which Mac models get M7 first - that detail typically surfaces at a product event, not a chip roadmap report.