Apple hardware is one of the best platforms for low scale AI workloads. A simple M4 Pro MacBook Pro with 64GB RAM can comfortably run the best “at-home” model currently available (Qwen3-Coder-Next) around 30-40tkps.
A comparably specced AMD Ryzen AI 395+ can do about half of that.
If Apple delivered that in server-scale, easily scalable packages, there would definitely be buyers.
I didn’t even want to open that can of worms…
No you wouldn’t need millions…
The two things I’ve listed above were done with self hosted LLMs and a $20 Claude subscription orchestrating them.
That’s the thing I’m talking about - these capabilities are literally given out for chump change. That’s what makes it dangerous.
Except it wasn’t just “throwing everything and the kitchen sink at a server”. Modern LLMs have been trained well enough to not just create a list of potential attacks and execute them, but to:
of course this can be used for good too. I did this to pentest my own homelab stack. I used the very same flow to hack into a pair of smart glasses I own (not the creep glasses, mind you, but the “wearable monitor” kind of AR glasses - turns out the pair I own runs on a pretty decent base, a somewhat recent Linux kernel and minimal userspace, with some exploitable interfaces) within about a day.
What’s truly dangerous is that these tools have the ability to turn a relatively simple “hack into the NASA servers” instruction into a detailed, executable plan of actually breaching the servers in question. That a person with more than one and less than three brain cells to rub together - someone who’d think this scene is super cool and completely legit looking - can, with minimal guardrails-bypassing, increase their own chance of hacking into ANY server, from less than 0.000000000000000001% to 20-30%. Now that’s scary.
Honestly, this could be a great equaliser.
Home grown AI models have been shown to be extremely good at market analysis and predictions, some managing quite nice gains by daytrading small amounts and micro transactions - gains that previously were reserved to the coke-moustached shark investors and savvy finance geeks.
If hosted models can bring results even just half as good as those, to the average people, that will be a make and break point for the finance fuckers. Because why would you go to an investment firm who offers maybe 40-50% gains YoY while taking 15-20% of the profits, when you can pay £20 for Claude or ChatGPT and have 300-500% returns on your base investment every month?
Oh really? Simplex would block someone from accessing your phone and thus Simplex’ data?
Does your compose file in itself give you a nice, easily readable status display, analytics, metrics, logging, etc. interface?
Of course the pricing will be different. My point here was that if Apple can jam that much compute into something as small as a Mac Mini, imagine what they can deliver at 4-5U scale.
Apple and Nvidia hardware combined might just be the solution to more scalable AI compute that uses less power.