Apple’s new M6 chip gets more cores and more AI compute

What happened

Apple has announced the M6 chip for Macs, marking the company's first processor built on a 2nm process node. Alongside it, Apple revealed the M5 Ultra, which it describes as its most powerful chip ever, engineered for demanding workloads including 3D rendering and running frontier AI models.

This piece breaks down what Apple actually disclosed, what the architectural shift to 2nm means in practice, and what developers and power users should be paying attention to before making any purchasing or infrastructure decisions. It is aimed at anyone building for Apple hardware or evaluating Mac-class compute for AI and creative work.

Apple announced two chips: the M6 and the M5 Ultra. The M6 is the headline chip for the next generation of Mac hardware, and it represents Apple Silicon's move to a 2nm fabrication process — the first time Apple has shipped a Mac chip at that node. The M5 Ultra is a separate product positioned at the high end, aimed at professional and research workloads where raw throughput matters most, including 3D rendering pipelines and the kind of large-model inference workloads that have become a benchmark category for workstation-class machines.

How it works

Apple framed the M5 Ultra specifically as its most powerful chip ever. That claim sits alongside the M6 announcement rather than competing with it, because the two chips serve different segments. The M6 is the mainstream successor in the M-series line; the M5 Ultra is an extreme-configuration part assembled from multiple M5 dies and targeted at workflows that have previously required dedicated GPU servers or external accelerators.

The most concrete architectural change in the M6 is the jump to 2nm silicon. Moving from a larger process node to 2nm generally allows a manufacturer to pack more transistors into the same die area, run them at lower voltage, or both, which translates into better performance-per-watt ratios. Apple says the M6 brings improvements in both performance and power efficiency, which is consistent with what a 2nm transition typically enables, though Apple has not specified exact core counts or clock figures beyond the announcement text.

Why it matters

The AI compute angle is the other headline change. Apple has emphasized that the M6 carries expanded AI compute capability — more cores dedicated to neural-network workloads — which connects directly to the company's ongoing push to run on-device AI features across its product line. The M5 Ultra's positioning around frontier AI models suggests that the Ultra configuration, which typically joins two dies with Apple's die-to-die interconnect, significantly multiplies the available neural engine throughput compared to a single-die chip.

Creative professionals doing 3D rendering, video production, or other sustained GPU workloads have the most immediate reason to pay attention to the M5 Ultra. Apple's framing of it as the most powerful chip ever and its explicit mention of 3D rendering as a target use case signals that the Mac Pro or a similarly configured workstation will be the primary vehicle for that chip. If your studio currently farms rendering work out to cloud GPU instances, the M5 Ultra's specs, once fully disclosed, will be worth comparing against that operational cost.

Developers building AI features for Apple platforms should note the M6's expanded AI compute. More neural engine cores mean faster on-device inference, which matters for features that need to run at interactive speeds without a network round-trip. If you are optimizing Core ML models or working with the Apple MLX framework, the M6's updated hardware will be worth benchmarking directly, particularly for transformer-class models that stress memory bandwidth and matrix-multiply throughput simultaneously.

Who is affected

Apple has not specified release dates or pricing beyond the announcement. The M6 will arrive in Mac hardware, and the M5 Ultra will power the high-end desktop configuration, but the exact product names, configurations, and price tiers have not been confirmed in the available announcement details. The 2nm fabrication is being cited as a first for Apple's Mac line, so availability will depend on production ramp from Apple's semiconductor manufacturing partners.

To prepare before devices ship, developers can profile existing apps against current M-series hardware to identify neural-engine and GPU bottlenecks, which will remain the same categories of bottlenecks on the M6. Apple's developer tools — Instruments, the Core ML Performance Report, and MLX benchmarking scripts — will let you set a baseline now so that comparative testing on M6 hardware is straightforward once access opens through Apple's developer hardware programs.

What to watch next

The numbers that matter most have not yet been published. Core counts for the M6's CPU, GPU, and Neural Engine, memory bandwidth figures, and the unified memory ceiling for the M5 Ultra are all details Apple typically reveals at or after a product launch rather than at an initial announcement. Those figures will determine whether the performance and efficiency claims hold up against third-party benchmarks, which historically appear within days of hardware availability.

On the AI side, watch for Apple's own announcements around which on-device AI features ship enabled on M6 at launch versus which remain M4 or M5 minimum. That compatibility floor is often where the practical impact of new neural engine capacity becomes visible to end users and where developers face real decisions about minimum deployment targets.