ZDNET’s key takeaways
- Macs went from neglected pro machines to AI workhorses.
- Local AI makes Apple Silicon Macs surprisingly compelling.
- Apple’s processors may matter more than Apple Intelligence.
Back in 2018, I had all but given up on Macs ever being updated. I had Windows and Linux infrastructure for some projects, but my main day-to-day desktop workloads were Mac-based. And I was starting to worry that I would have to move everything off the Mac, a migration process that was bound to be anything but pleasant.
The Mac ecosystem just didn’t offer enough power and headroom for heavy power users like me. You’d never have guessed back then that Mac minis would soon become the most highly sought-after machines for high-performance AI workloads.
Everything changed when Apple unveiled the M1 Apple Silicon processor in 2020. That’s because Apple Silicon is a system-on-a-chip (SoC), which means that all basic system components, such as the CPU, GPU, and RAM, are produced on a single silicon wafer. This solves many of the problems that plague PCs — from heat to modularity bottlenecks to the high cost of components.
But this change took some getting used to for power users who liked being able to add RAM or swap out GPUs. However, putting everything on one chip is what makes the Mac an absolute fire-breathing monster of an AI platform.
The AI advantage of Apple Silicon
Apple Silicon chips have internal processors. The CPU handles all the basic computing tasks. Apple’s CPUs also have extensions that can do high-precision math calculations.
Also: Apple Mac Mini M5 Pro review: Serious power, for the right users
The GPU can process data-heavy workloads. Most graphics-intensive calculations require a ton of array processing. And you know what also needs a ton of array processing? AI. That’s why GPUs are in such demand.
Then you have Apple’s neural engine, a chip designed to run specialized AI tasks. Somewhat simplified, the GPU handles huge amounts of data, while the neural engine can handle specific AI tasks with very little power, such as speech recognition or image classification.
Also: The iPad Pro M5 has the best battery life in our lab testing, and it’s on sale
You can do many AI workloads on the original 2020 Apple Silicon M1-series Macs. My four-year-old M1 Mac Studio is running major AI workloads in my home lab right now, or you can pick up a hot new 32GB M6 Mac mini for about $1,700.
As these devices get more powerful, running local AI loads has another benefit: they’re local. You don’t have to worry about data sovereignty or token fees. Slam a near-frontier LLM on the thing, and go to town.
But not Apple Intelligence
Sixteen months ago, I replaced my 64GB M1 Max Mac Studio with a 128GB M4 Max Mac Studio, which is my daily driver.
Just last month, Apple upgraded both the Mac mini and Mac Studio to M5 Max and M5 Ultra. But my older Mac Studio is doing so well that I don’t need to upgrade.
Had you told me, back in 2018, that Apple would have small, compact Macs with so much power that I wouldn’t need to upgrade as soon as a new model came out, I’d have laughed. This is the company that went four years between Mac mini upgrades and a whopping six years between Mac Pro upgrades.
This year? Apple is updating its headless Macs almost as fast as it does iPhones. Of course, much of that’s because Macs and iPhones share Apple’s in-house processor technology. Apple has far more control of its own architecture, making upgrades more practical and possible.
Also: My 5 favorite Siri AI ‘onscreen awareness’ tricks – and why
Plus, there’s the AI demand. I don’t think Apple could have possibly planned for the AI boom way back in 2020. But the company created a capable processor design for their customers’ workloads, most notably graphics-related ones. It just so happens that AI shares many of the architectural requirements of graphics, uniquely positioning this new Apple-owned architecture far ahead of the Intel world it previously depended upon.
It’s actually quite ironic. Apple Intelligence has turned out to be something of a flop. But Apple’s processors, designed for an entirely different class of work, have turned out to be almost perfectly optimized for AI.
The company’s hardware is taking the AI world by storm, while the company’s AI software is barely making a dent in anyone’s plans — but I’m not complaining.
You can follow my day-to-day project updates on social media. Be sure to subscribe to my weekly update newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.
