Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

MacRumors

macrumors bot
Original poster


Apple's Mac mini and Mac Studio have become the machines of choice for running AI agents, according to Doug Brooks, Apple's senior product manager of Apple silicon.

apple-silicon-feature-joeblue.jpg

Brooks made the claim while discussing Apple's chip strategy in a newly published interview with The Deep View conducted just prior to WWDC 2026 in June.

Brooks says that the company has seen "incredible demand" for the two desktop Macs. When it comes to agentic workloads, "people often want a system that's under their control, isolated from their primary machine, and capable of running 24 hours a day, seven days a week," said Brooks.

"A Mac mini is an amazing system for that," he added.

Many AI tools are also Mac-first or Mac-only, which Brooks says has helped cement the Mac's standing among developers, including those at frontier AI labs where Macs are said to be a common sight.

The Apple executive also conceives of agentic AI as a whole-chip problem rather than a GPU one. "It's not just about the GPU crunching on an LLM anymore," he said. "It's about the whole chip contributing to different parts of the task, tool-calling, and the things that are happening around those workflows. It really plays to the strengths of Apple silicon."

Brooks links Apple's position of strength in modern AI back to chip decisions made long before LLMs like ChatGPT arrived. He points to the Neural Engine, which is built for power-efficient matrix math, along with lesser-known neural accelerators inside the CPU that handle time-sensitive tasks like speech.

Apple more recently added neural accelerators to the GPU, which has extended AI performance across the board from iPhone-class parts up to the Mac's largest silicon. Brooks ties that progress to Apple's design method, where a chip is built for a specific machine, and the hardware and software are developed in tandem.

He also described a shift toward running AI locally rather than in the cloud – a move motivated by privacy, security, and the rising cost of inference as agents consume more tokens. However, Brooks envisions a hybrid future in which agents decide what runs on-device and what gets sent to the cloud.

He also singled out what he calls "transparent AI" on iPhone and iPad, referring to features scattered throughout the operating system and third-party apps that work quietly without announcing themselves as AI.

Some of the examples he cited include Draw Things, an image generator that runs across iPhone, iPad, and Mac, and SwingVision, which analyzes tennis and pickleball gameplay in real time using the iPhone's cameras.

"The speed of AI development right now is just crazy," Brooks said. "I can't imagine where we're going to be a year from now, three months from now, or even a month from now," he added.

You can read the full interview over on The Deep View website.

Article Link: Apple Silicon Exec Explains Mac Mini AI Demand and On-Device Future
 
How much memory are people generally getting for AI on those Mac minis? Are they mostly M4 Pros with 48 GB RAM or are many M4s with 24 GB or less? Just curious.

To run a local AI for a specific task, you need 48GB or more to make it run comfortably. The models used are often around 32GB and when loaded into the unified memory can run very fast. It can be done via SSD as well, but that will slow the AI processing extremely, not to mention the wear it will cause on the SSD. There are smaller AI models for Mac's with less RAM, but at some point they are not as "smart", fast and useful as the larger ones. Of course it depends on the task at hand.

If more users do move away from AI in the cloud as the current trend shows, I'm sure that the demand for more memory in Mac mini's or Mac Studio's will increase big time and the worldwide shortage will continue for the coming years. Even when datacenters stop buying everything.
 
Are you really believing this guy's claims? Sure, people have gotten some toy models to work at a snail's pace, but nobody is paying Apple's RAM prices for real LLM models to run on.

I run a DGX Spark here (128gb RAM) and an M5 Max (128gb RAM) and the Mac is faster, but the CUDA stack has better software support. There are some large, quality models that run well in MLX mode on the Mac, and I'm basically just waiting until I can buy whatever M5 Ultra is released, whenever it's released, specifically for local AI.
 
The way the guy tells it you'd think Apple got on this bandwagon through proactive forward thinking designs decisions rather than sheer luck, problem is they are stuck with one fab for their chips, can't get enough capacity to satisfy demand and have a severe memory shortage problem that's not going away, how's that for forward proactive thinking.
 


Apple's Mac mini and Mac Studio have become the machines of choice for running AI agents, according to Doug Brooks, Apple's senior product manager of Apple silicon.

apple-silicon-feature-joeblue.jpg

Brooks made the claim while discussing Apple's chip strategy in a newly published interview with The Deep View conducted just prior to WWDC 2026 in June.

Brooks says that the company has seen "incredible demand" for the two desktop Macs. When it comes to agentic workloads, "people often want a system that's under their control, isolated from their primary machine, and capable of running 24 hours a day, seven days a week," said Brooks.

"A Mac mini is an amazing system for that," he added.

Many AI tools are also Mac-first or Mac-only, which Brooks says has helped cement the Mac's standing among developers, including those at frontier AI labs where Macs are said to be a common sight.

The Apple executive also conceives of agentic AI as a whole-chip problem rather than a GPU one. "It's not just about the GPU crunching on an LLM anymore," he said. "It's about the whole chip contributing to different parts of the task, tool-calling, and the things that are happening around those workflows. It really plays to the strengths of Apple silicon."

Brooks links Apple's position of strength in modern AI back to chip decisions made long before LLMs like ChatGPT arrived. He points to the Neural Engine, which is built for power-efficient matrix math, along with lesser-known neural accelerators inside the CPU that handle time-sensitive tasks like speech.

Apple more recently added neural accelerators to the GPU, which has extended AI performance across the board from iPhone-class parts up to the Mac's largest silicon. Brooks ties that progress to Apple's design method, where a chip is built for a specific machine, and the hardware and software are developed in tandem.

He also described a shift toward running AI locally rather than in the cloud – a move motivated by privacy, security, and the rising cost of inference as agents consume more tokens. However, Brooks envisions a hybrid future in which agents decide what runs on-device and what gets sent to the cloud.

He also singled out what he calls "transparent AI" on iPhone and iPad, referring to features scattered throughout the operating system and third-party apps that work quietly without announcing themselves as AI.

Some of the examples he cited include Draw Things, an image generator that runs across iPhone, iPad, and Mac, and SwingVision, which analyzes tennis and pickleball gameplay in real time using the iPhone's cameras.

"The speed of AI development right now is just crazy," Brooks said. "I can't imagine where we're going to be a year from now, three months from now, or even a month from now," he added.

You can read the full interview over on The Deep View website.

Article Link: Apple Silicon Exec Explains Mac Mini AI Demand and On-Device Future
I think there was a typo in Brooks quote. I fixed it:

“"I can't imagine where we're prices are going to be a year from now, three months from now, or even a month from now," he added.”
 
The way the guy tells it you'd think Apple got on this bandwagon through proactive forward thinking designs decisions rather than sheer luck, problem is they are stuck with one fab for their chips, can't get enough capacity to satisfy demand and have a severe memory shortage problem that's not going away, how's that for forward proactive thinking.
TSMC is pretty much the only option for cutting edge nodes on the fab side, and do you think the memory shortage is an apple specific problem?
 
How about a build option for the Mac mini for up to 128 GB of RAM? 😛

After being spoiled with a studio max, I can never go back to lesser ram.
 
  • Like
Reactions: centauratlas
Are you really believing this guy's claims? Sure, people have gotten some toy models to work at a snail's pace, but nobody is paying Apple's RAM prices for real LLM models to run on.
Apple Silicon is well supported and has a very active community surrounding it.

Running local LLMs is basically a choice between CUDA (DGX Spark or a rig with multiple Nvidia GPUs) or Metal (Apple Silicon Macs). The others (AMD and Intel) have very immature software ecosystems and aren't worth the hassle.

It's definitely not just about running toy models. I run DeepSeek V4 Flash on my 128GB M4 Max with an inference engine called DwarfStar, speed is good and output quality is practically indistinguishable from something like Sonnet 4.6 (I have uncapped access to Anthropic models at work and I'm required to use them so I have a good feel for that).
 
If more users do move away from AI in the cloud as the current trend shows, I'm sure that the demand for more memory in Mac mini's or Mac Studio's will increase big time and the worldwide shortage will continue for the coming years. Even when datacenters stop buying everything.
I don’t use AI much. But I’m just hoping I can afford a new Mac Studio (M5, M7…) with 64GB of RAM and a 2TB SSD. As things stand right now, it’s not looking good.
 
Apple Silicon is well supported and has a very active community surrounding it.

Running local LLMs is basically a choice between CUDA (DGX Spark or a rig with multiple Nvidia GPUs) or Metal (Apple Silicon Macs). The others (AMD and Intel) have very immature software ecosystems and aren't worth the hassle.

It's definitely not just about running toy models. I run DeepSeek V4 Flash on my 128GB M4 Max with an inference engine called DwarfStar, speed is good and output quality is practically indistinguishable from something like Sonnet 4.6 (I have uncapped access to Anthropic models at work and I'm required to use them so I have a good feel for that).
I didn't know this. I'm still learning about AI tools, but my impression so far was that Intel was the preferred platform; you could run these models on Mac, but not as well.

For something like Stable Diffusion, what would you say is better, an M-series Mac, or an Intel PC?
 
I didn't know this. I'm still learning about AI tools, but my impression so far was that Intel was the preferred platform; you could run these models on Mac, but not as well.

For something like Stable Diffusion, what would you say is better, an M-series Mac, or an Intel PC?
I don't have much experience with image/video generation, not sure how it stacks up tbh (I'm just focused on coding)

I tried Comfy UI a few weeks back out of curiosity (can't remember which model). It generated a few images that looked pretty good, no major issues (e.g. correct number of fingers lol). No idea how it compares to other platforms 🙂
 
TSMC is pretty much the only option for cutting edge nodes on the fab side, and do you think the memory shortage is an apple specific problem?
People forget that Intel has been woefully inadequate at producing efficient performant chips for quite awhile and they still are. I’m still wondering which year will be the one where everything Intel sells is more efficient/performant than Apple’s M1.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.