If I Were the Head of Apple A.I.

Every new technology wave causes a veritable tsunami of disruption, and, as of mid 2026, Artificial Intelligence doesn’t appear to be any different. From the most ridiculous political slopaganda to the uncanniest of valleys through to A.I. psychosis, though, not all is well in the wonderful world of Artificial Intelligence.

It’s a time of irrational exuberance. Trillions of dollars¹ of investor money has been shoveled into these projects that are yet to prove any reliable path to profitability. Titans of the software industry – Oracle, Meta, Amazon, to name a few – have been writing some serious cheques here. Microsoft and Google have been spending every waking moment inventing new ways of tricking users into adopting whatever A.I.² tool they’ve decided to chuck against the wall. Maybe notepad copilot will be what sticks. Stranger things have happened.

There’s one company though that’s conspicuously absent from the fray. Apple has been remarkably conservative here; while Google is busy stepping on rakes, Apple’s copying its homework.³ And that’s not because Apple doesn’t have the cash to splash.

There’s a rationality to this approach. The tech world has a million-and-one examples of first-movers stumbling, falling, and fading relative to their up-and-coming rivals. It’s the reason that we’re not using Netscape on our Blackberries to browse Myspace on the MCI network. Technology emerges, the monkeys who discover fire manage to immolate themselves (and a few hundred billion along the way), and from the ashes more useful discoveries emerge.

So if we think of ChatGPT as a modern-day Pets.com, what’s next? Where might the future go?

Realistically, the thing to think about here is the scale of things. Everything to do with the current state of A.I. is ginormous. Big money, big models, data centres the size of small European countries. All of it is just big. The thing is, computers used to be huge, too. Mainframes the size of rooms have given way to microprocessors that offer more processing power at a fraction of the cost, size, and energy. Telecommunications went the same route. Mobile phones have gone from briefcases and businessmen to the pockets of the proletariat. Because it’s not size that drives value, it’s capability.

Even now, models don’t need to be huge. It’s not going to win coding benchmarks, no, but a raspberry pi with a small open-source model is perfectly capable of basic image generation and text inference.⁴

Centralised services are expensive. If your customers are using your processing power, it’s up to you to somehow cobble together the silicon needed to make it work. If you’re in the business of running 100B+ parameter models for immediate processing, you’d better start hunting for whatever capacity you can find. Fingers crossed that the market’s willing to pay your running costs. Gulp.

But if you’re able to get on-device A.I. working well, that’s a major cost gone (or at least borne by the consumer when they bought their iPhone). Sure, the distillation and model generation processes are energy intensive, but that’s a background process. A model that took a day to generate isn’t any different from one that took a week on older hardware. Once you’ve got your model, and distributed it to your customers’ devices, job done. You’re not paying every time someone uses an LLM to compose a heartfelt apology text.

And that’s just the monetary cost – OpenAI and Anthropic seem to be surprisingly certain that their customers will unquestioningly hand over their most personal and private data. In an age of social media, that’s been a reasonably safe assumption for now, but trust is a ming vase. When it’s broken, that’s it. No more. Can these companies come back from a hack that leaks raw SMS data and uploaded files? Maybe. I wouldn’t want to find out.

For that matter, it’s also a strategy that’s reliant on American firms being the only competitors in the market. China and its plethora of free models is rightly seen as the biggest competitor, but not only are they being unfairly (and short-sightedly) dismissed, they’re far from the only hefty, hungry tech market.⁵

Technology and the world at large are changing in ways that are unforeseeable. When the stakes are this big, let’s hope that the decision makers are making sensible, careful plans. Time will tell.