Apple is making a significant change to its custom silicon roadmap. According to a report from Bloomberg, the company will skip its high-end M6 chips and instead fast-track a new, AI-focused M7 generation aimed at powerful Macs arriving in 2027.
The shift represents one of the most notable strategic pivots in Apple Silicon's history and underscores how thoroughly the generative AI boom has reshaped the priorities of even the most vertically integrated hardware maker in the industry. For more context on this story, see our ongoing more AI stories.
What Apple Is Changing
Under the revised plan, Apple will still release a base M6 chip for entry-level Macs as soon as this year. An M5 Ultra chip could also arrive in 2026 for the top-tier Mac Studio and Mac Pro. But the company has canceled plans for higher-end M6 Pro and M6 Max chips, which would normally have powered the next MacBook Pro and high-end Macs.
Instead, Apple's next Pro and Max chips will belong to the M7 lineup, with the first M7 chips expected to launch in 2027 — with some reports pointing to the first half of that year.
In other words, rather than iterating its performance chips on the standard yearly cadence, Apple is deliberately pausing the high end to engineer a generation purpose-built for artificial intelligence.
Why Memory Bandwidth Matters for AI
The clearest signal of the AI focus is memory bandwidth, the spec that most directly determines how fast a chip can run large language models and other AI workloads locally.
The current base M5 chip delivers 153 GB/s of memory bandwidth. The upcoming base M6 is expected to reach around 200 GB/s, an improvement geared toward better graphics, faster AI processing, and smoother video editing. The base M7, meanwhile, could hit 240 GB/s — and according to one report, the M7's unified memory bandwidth could be roughly 56% higher than the M5 across the line.
That headroom matters because running modern AI models on-device is intensely memory-bound. The more bandwidth a chip has, the faster it can shuffle model weights in and out of memory, and the larger the models it can run smoothly without offloading to the cloud.
A Bet on On-Device AI
Apple's decision to reorganize its roadmap around AI is not happening in a vacuum. The company has lagged rivals like Google, Meta, and a wave of startups in shipping generative AI features, and its approach has consistently emphasized privacy-preserving, on-device processing over cloud-dependent services.
Building chips with substantially more memory bandwidth directly serves that strategy. It lets Apple run larger, smarter models entirely on a user's Mac — keeping data private and reducing reliance on remote servers — without the performance penalties that have so far constrained local AI.
The M7 generation, then, is less a routine upgrade and more an architectural statement: Apple wants its highest-end Macs to be credible platforms for serious AI development and inference, not just consumption devices. That ambition extends beyond memory bandwidth: a purpose-built AI chip can also dedicate more silicon to the neural engines and matrix math units that accelerate model inference, narrowing the gap between a laptop and the specialized hardware that trains these systems in the cloud.
For software developers, the implications are considerable. A Mac with substantially more memory bandwidth and larger unified memory pools becomes a far more practical machine for running open-weight models locally, prototyping AI features, and iterating without racking up cloud inference bills. If Apple delivers, the Mac could become a preferred development environment for the very AI engineers the company has been working hard to recruit.
What This Means for Buyers
The change carries practical implications for anyone in the market for a high-end Mac. Because Apple is skipping the M6 Pro and Max, the current M5 Pro, Max, and Ultra machines will have an unusually long life at the top of the lineup. Buyers who need maximum performance now will still be well served by those chips, but they may also face a longer wait for the next meaningful leap.
For entry-level buyers, the base M6 — with its ~200 GB/s bandwidth — should bring a solid step up in everyday AI and graphics performance when it arrives later in 2026.
The Bigger Picture
Apple's roadmap reset arrives as the entire chip industry is reorganizing around AI. Nvidia continues to dominate the data-center market for AI training, while competitors including Qualcomm, AMD, Amazon, Google, and a growing field of custom-silicon startups race to build alternatives. Even OpenAI and SpaceX are now reported to be designing their own chips.
What sets Apple apart is its control over the full stack — from the silicon to the operating system to the applications. By redirecting its most powerful custom chips toward AI, Apple is positioning the Mac as a distinct kind of AI device: one where intelligence runs privately on the hardware in front of you rather than in a distant data center.
The gamble is that by 2027, on-device AI will be capable and important enough to justify an extra year of waiting for the next great Mac chip. Given how quickly the AI landscape is moving, that may prove to be the most consequential hardware bet Apple has made since it first ditched Intel.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
