macOS 27 Golden Gate: The Ars Technica review

Apple’s latest macOS rollout, dubbed “Golden Gate,” lands with a thunderous splash of generative AI that can’t be turned off, while simultaneously offering the kind of under‑the‑hood polish longtime Mac fans have come to expect. Two years after the debut of Apple Intelligence, the feature is now baked into the operating system and, according to the Ars Technica review, defines the entire release. The result is a hybrid that feels part “Snow Leopard‑style” refinement and part hard‑to‑ignore AI overhaul, and it forces every Mac owner to confront a new reality: if you can’t run Apple Intelligence, you’re effectively left behind.

Apple Intelligence: From Optional Add‑On to Mandatory Core

The most striking change in Golden Gate is the removal of the toggle that let users disable Apple Intelligence and delete the gigabytes of AI models it downloaded. In prior releases, you could flip a switch, reclaim disk space, and keep your Mac free of on‑device AI processing. That option is gone. The review notes that “generative AI is unavoidable if you install macOS 27 Golden Gate,” underscoring Apple’s intent to make AI an inseparable part of the macOS experience.

Apple Intelligence now runs on an “AFM 3 Core” model built in collaboration with Google, while a more powerful “AFM 3 Core Advanced” variant powers the most demanding AI features, such as the new expressive voice capabilities in Siri and an improved voice‑dictation engine. The split between the two models introduces a hardware tiering that will shape how users experience the OS for years to come.

Hardware Requirements: The End of Intel Macs

Golden Gate is the first macOS release since the mid‑2000s that does not support any Intel‑based Macs. Apple has finally dropped the last handful of 2019‑ and 2020‑era Intel machines that could run macOS 26 Tahoe. From now on, the OS requires Apple Silicon—specifically an M1 chip or newer. The review’s testing hardware includes an M1 MacBook Air (16 GB RAM, 1 TB SSD), a 15‑inch M3 MacBook Air (24 GB RAM, 512 GB SSD), and a MacBook Neo (8 GB RAM, 512 GB SSD), providing a cross‑section of Apple’s entry‑level to mid‑range Silicon lineup.

While the OS runs “mostly happily” on any Apple Silicon Mac that isn’t falling apart, the advanced AI features demand higher‑end chips. Only M4, M5, and M6‑based Macs meet the RAM and performance thresholds for the AFM 3 Core Advanced model. Even an M1 or M2 with 16 GB of RAM is insufficient, and an M3 with just 8 GB of RAM also falls short. This creates a clear divide: users with older Silicon will still get Apple Intelligence, but they’ll miss out on the most expressive Siri voice and the best dictation experience.

What Happens to Intel Users?

For those clinging to Intel Macs, the review outlines a dwindling set of options. If your machine still supports macOS 15 Sequoia or macOS 26 Tahoe, you can continue on those versions until Apple ceases security and Safari updates—projected for the fall of 2027 for Sequoia and 2028 for Tahoe, assuming Apple maintains its historical update cadence. Staying on these older OSes means you’ll miss out on new features and may receive less comprehensive security patches, but you’ll remain officially supported.

Older Macs that stopped at macOS 14 Sonoma or earlier face a harsher reality. While browsers like Chrome and Firefox still run on legacy versions (Chrome back to macOS 13 Ventura, Firefox to macOS 10.15 Catalina), the security risk grows as Apple’s patching focus shifts to newer releases. Projects such as OpenCore Legacy Patcher once enabled newer macOS builds on vintage hardware, but the tool’s development has stalled, especially after Apple stripped Intel‑native code from recent releases. Consequently, running macOS 15 on mid‑2010s hardware is now a “hit‑or‑miss” endeavor, dependent on exact configurations.

Alternative Operating Systems: Windows and Linux

The review reminds us that Boot Camp still supports Windows 10 on many Macs, though Microsoft’s security updates are waning. Installing Windows 11 is possible but often requires using legacy Windows 10 drivers, leading to quirks and reduced stability. Given that most Mac buyers aren’t primarily interested in Windows, this path is rarely the best long‑term solution.

Linux remains a viable fallback for older machines, particularly those from 2017 or earlier. However, Macs equipped with the Apple T2 security chip present challenges, as the T2 handles critical I/O functions that many Linux distributions struggle to support. Specialized distros—such as CachyOS, which includes pre‑install asterisks for T2 compatibility—offer better out‑of‑the‑box support, yet the broader Linux‑on‑Mac community is increasingly focused on Apple Silicon, leaving Intel‑based Linux users with fewer resources.

Pricing Pressures and the Upgrade Decision

Beyond technical compatibility, the review flags a market reality that will influence many buyers: the “ongoing RAM and storage crisis” has driven Mac pricing higher across the board. For owners of older Macs, the cost of upgrading to a newer Apple Silicon model that can fully leverage Golden Gate’s AI features may be steep. The author recommends that anyone whose machine cannot run macOS 15 Sequoia—or who is already feeling the pinch of limited RAM and storage—should seriously consider a hard upgrade to a newer Silicon Mac, despite the price tag.

This recommendation is grounded in the fact that Apple Intelligence now runs on all Apple Silicon devices, while Intel Macs are left with a dwindling support window. As the AI features become more central to macOS workflows—think of integrated voice assistants, on‑device text generation, and AI‑enhanced productivity tools—being on a platform that can’t run them will increasingly feel like a handicap.

What Golden Gate Means for the Future of macOS

Golden Gate’s blend of incremental macOS polish and unavoidable AI integration signals a strategic shift for Apple. The OS addresses “the most irritating of macOS 26 Tahoe’s design sins,” delivering speed and reliability tweaks that longtime users will appreciate. At the same time, by making Apple Intelligence a non‑optional component, Apple is positioning generative AI as a core OS capability rather than a peripheral add‑on.

Looking ahead, the hardware tiering introduced with AFM 3 Core Advanced suggests Apple may continue to widen the gap between entry‑level Silicon and its flagship chips. As newer AI models demand more compute and memory, older Macs—especially those stuck on M1 or M2 silicon—could see a gradual erosion of feature parity. For enterprises and power users, the implication is clear: to stay on the cutting edge of AI‑driven productivity, investment in the latest Apple Silicon hardware will become increasingly essential.

This article was produced with AI-assisted research and editorial support. Reporting is based on the source material cited below. Sources: Ars Technica; arstechnica.com; Global1.News (16 September 2026).

By Jessica Ali, Staff Writer