Everyone framed Apple as the company that “missed” AI. I think they’ll be the ones who pop the bubble.

For three years we’ve been sold doom, hype, and mythical models too dangerous to release. The result? Hundreds of billions spent, almost nobody paying, and corporate AI projects quietly shelved because the economics don’t work. That’s not a revolution. That’s a subsidy waiting to expire.

Here’s the uncomfortable truth the cloud crowd won’t say out loud: ninety percent of business AI isn’t frontier reasoning. It’s classification, summarisation, extraction, and lookup. You don’t need a model trained on the complete works of Shakespeare and a billion Reddit posts to tidy an invoice. You need something good enough — running locally, cheaply, privately.

A small model on an M-series chip isn’t just adequate for that work. It’s better. Faster, cheaper, and safer than shipping your data to a warehouse in Virginia. Six of the Magnificent Seven bet everything on renting you tokens. Apple bet you’d rather own the machine.

When a 128GB Mac Mini puts a server room on your desk for the price of a laptop, the cloud-only business model breaks. Not with a bang — with a spreadsheet. CFOs will simply do the maths.

It’ll feel like Y2K: years of fear, then a great big nothing burger. And Apple, quiet and patient, will have been right all along.