Apple Was Always 1 Step Behind in AI. In Chipmaking, It’s Now 1 Step Ahead.

Apple (AAPL) just made a strong move to improve its AI reputation. The company spent two years being seen as a step behind in AI, yet it has quietly become the first to introduce a 2nm chip. Its new M6, built on TSMC's (TSM) most advanced 2nm process, is the first of its kind in a consumer computer. The chip debuts in a new Mac mini, alongside a more powerful M5 Ultra in the refreshed Mac Studio.

Apple says the M6 delivers nearly 30% more peak GPU compute for AI than the M5 and over eight times that of the M1. This lets it run AI models on the device faster. With up to 32GB of memory, it can run large language models (LLMs) locally without sending anything to the cloud. Preorders are now open, and the Mac Mini starts at $899, shipping Sept. 22.

Look past the specs, and the M6 is really about something else. Amazon (AMZN), Microsoft (MSFT), Meta (META), and Alphabet's (GOOG) (GOOGL) Google spent a combined figure of over $650 billion this year on AI data centers. Apple's AI budget is a small fraction of that. The company is betting the future of AI lives on the device in your hand rather than some remote data center. Apple already has roughly 2.5 billion devices in use. Every M6 it ships turns another Mac into a machine that can run AI on its own, privately. Rivals are spending fortunes to run AI in the cloud. Apple's edge is that it can put AI directly onto the massive base of devices people already own, without that kind of spending.

There is one issue worth noting, and it is something I wrote about recently. Apple's weak spot has never been the hardware. It is the software meant to run on it. The revamped Siri slipped by more than a year and now leans on Google's Gemini, and the camera AirPods just moved to 2027 for similar reasons. The chip is ready. Whether Apple can deliver the models to use it fully is still in doubt.

One choice stands out. Apple put its first 2nm chip in the Mac, not the iPhone. That is because the Mac is where people run the heaviest AI locally, and it shows Apple expects serious on-device AI to take off on the computer first.