Apple Intelligence rejects the assumptions of Big AI.
It is a turning point for the industry.
While analysts are all focused on Apple letting OpenAI’s ChatGPT answer questions for it, the headline should be how Apple managed to become an AI player based solely on data users already bring to its party. Siri will now be an interface for everything the iPhone, Watch, and Vision Pro do. In time it will become the primary interface.
Apple could have built an LLM. It could have packed its data centers full of Nvidia hardware to do Generative AI. It could have written, or bought, Generative AI software. That it chose to essentially subcontract this out is important. It shows that Apple understands something I’ve been writing here for a while now.
Waiting for the Bubble to Pop
Despite Huang’s Law the cost of LLM computing, both in energy used and energy wasted in the form of heat, is too high to scale. You can’t charge people enough, and deliver enough value, to be worth the cost, in the consumer market.
This means the industry is running toward a crash like that of 2000. At some point, the maximum value of AI will be reached. Everything will end in tears.
This might even end early. What’s the churn rate for Microsoft’s Co-Pilot? ChatGPT is already losing popularity. I know the initial take-up rate for Co-Pilot was high, but how many companies are dropping it, and how fast are others taking it up?
The Bubble could have popped already, if someone had read recent earnings releases from Salesforce.com and ServiceNow. Along with Adobe, they were hit hard recently. But no one seemed to take note, even though these are the folks who deliver AI value to business customers right now. They have the data, customer plans, and the operational bottlenecks. They can apply AI software to those bottlenecks, boost productivity and deliver a return on the investment.
Yet they’re not growing fast enough? Hmmm…
The problem is compounded by the fact that the AI business today is filled with hucksters. Elon Musk and Sam Altman are salesmen. They’re not technologists. They’re barely businessmen. Musk may have once slept at his factory, but did he really get his hands dirty? Or did he just berate the people under him until they figured out how to do what Henry Ford did a century earlier? Altman, meanwhile, stole a non-profit, got someone to write a demo, turned it into a company for himself, and he’s now collecting billions from investors who don’t know any better.
Apple’s Not Stupid
Apple knows that the cost-reward calculation for even Little AI requires hardware at edge data centers. Over time this will become Apple hardware, but there’s a cost. To pay that cost, Apple is limiting new features to the iPhone16 and later models, setting up an upgrade cycle.
But if there were big profits to be had deploying Big AI, at scale, right now, don’t you think Apple would be doing that? I do. Instead, they’re handing cash flow back to shareholders, in the form of dividends and buybacks.
That is what I call a Clue.
