Apple Skips M6 Pro and Max to Fast-Track an AI-First M7 Silicon Generation

Apple is making an unusual break in its silicon cadence. According to a Bloomberg report, the company will skip the high-end M6 Pro, Max, and Ultra variants entirely. Instead, Apple will ship a base M6 chip later this year, then jump directly to an M7 family designed around on-device AI. The M7 Pro, Max, and Ultra are slated for late 2027 and 2028.

This is the first time Apple has introduced a new M-series generation without accompanying Pro and Max chips since the transition from Intel. The move signals something deeper than a supply-chain pivot. Apple is restructuring its silicon roadmap around a single thesis: the next performance inflection point is AI, not raw CPU or GPU throughput.

What the roadmap actually says

The base M6, codenamed Komodo or H18G, will appear in a refreshed entry-level MacBook Pro later this year. It brings an updated memory architecture, an upgraded Neural Engine, and memory bandwidth of around 200 gigabytes per second, up from 153 GB/s on the M5. Apple has tested versions with up to 12 GPU cores, a modest increase from the M5’s maximum of 10.

Then the gap. No M6 Pro. No M6 Max. No M6 Ultra.

Instead, Apple will release one final M5 chip: the M5 Ultra, codenamed Sotra D, with around 36 CPU cores and 80 GPU cores, supporting up to 768 gigabytes of memory. That chip is destined for a refreshed Mac Studio, though component constraints have already delayed it once.

After that, the M7 family arrives. The base M7, codenamed Delos, targets memory bandwidth of around 240 GB/s. The higher-end M7 Pro and M7 Max, collectively dubbed Andros internally, are expected in late 2027. The M7 Ultra follows in 2028.

The report frames the shift as a response to “growing demand for on-device AI capabilities and increasingly graphics-intensive software.” Apple is also navigating an industrywide memory shortage that recently forced price increases across Mac and iPad lineups. But the decision to skip an entire tier of chips is not a cost-cutting measure. It is a strategic reallocation of engineering resources.

The AI-first argument

Apple’s Neural Engine has been a fixture since the A11 Bionic in 2017. But the company has historically treated on-device AI as a feature, not a product line. The M7 changes that. By skipping the M6 Pro and Max, Apple is effectively saying that the next generation of high-end Mac chips needs to be architected around AI workloads from the ground up, not retrofitted.

The numbers support the thesis. Memory bandwidth is the single most important spec for running large language models on-device. The M6 base chip’s 200 GB/s is a meaningful upgrade, but the M7’s 240 GB/s target puts it in a different class. For context, a 7-billion-parameter model quantized to 4 bits requires roughly 3.5 GB of memory and benefits enormously from bandwidth above 200 GB/s for interactive latency. Apple is signaling that the M7 generation will be the first where on-device AI is the primary design constraint, not an afterthought.

This is also a bet on the software stack. Apple Intelligence, the company’s suite of on-device AI features, has grown steadily since its introduction. But the real prize is running larger models locally, models that today require a cloud round-trip. An M7 with dedicated AI acceleration and high-bandwidth unified memory could make that possible for a meaningful fraction of Mac users.

What gets sacrificed

The move leaves a gap for professional users who need raw CPU and GPU performance for video editing, 3D rendering, scientific computing, and software compilation. The M5 Ultra, with its 36 CPU cores and 80 GPU cores, will have to carry that load through 2027. That chip is powerful, but it is also a dead end. Apple has never released an M4 Ultra, and the M5 Ultra appears to be a one-off.

For users who would have bought an M6 Max or M6 Ultra in 2027, the message is clear: wait. The M7 generation will arrive, but not until late 2027 at the earliest for the Pro and Max, and 2028 for the Ultra. That is a long wait in a market where AMD and Intel are shipping new workstation chips every year.

The risk is that Apple cedes the high-end creative professional market to Windows workstations for a generation. The reward is a silicon line that can run AI workloads that competitors cannot touch.

Johny Srouji’s bet

The roadmap shift falls under Johny Srouji, Apple’s Chief Hardware Officer, who recently took over all hardware engineering amid John Ternus’s transition to CEO. Srouji is also overseeing Apple’s 2-nanometer iPhone chips and custom silicon for the foldable iPhone and 20th-anniversary iPhone models. This is his first major architectural call as the top hardware executive.

Srouji is betting that the M7’s AI capabilities will create a new category of Mac use cases, not just improve existing ones. That is a harder bet to evaluate. On-device AI has been a promise for years, and most users still interact with AI through cloud services. But the M7’s hardware could enable applications that are not possible today: local agents that run continuously, real-time language models for accessibility, on-device video generation, and privacy-preserving AI that never sends data to a server.

What to watch

Two things. First, whether Apple ships the M7 on schedule. A jump from base M6 in late 2026 to base M7 in early 2027 is aggressive, especially given the memory shortage. Any slip pushes the Pro and Max into 2028, which would leave the high-end Mac lineup stale for two years.

Second, whether the software catches up. Apple Intelligence is improving, but it has not yet produced a killer app that requires the M7’s capabilities. If the hardware arrives before the software, the M7 generation will look like a solution in search of a problem.

Apple is making a structural bet that on-device AI will be the defining performance metric of the late 2020s. The M6 Pro and Max are the sacrifice. The M7 will be the test.