Apple has just done something that was unthinkable until recently: publicly admit that you don’t have the best AI. That after fifteen years of trying to make Siri work, with the advantage of hitting first, he gives up. That the brains of Apple Intelligence, including the new Siri, Google will put it. And yet, it has just gained momentum to preserve its dominant position for the next decade. A technological paradox.
This isn’t a move Apple should be very proud of, but it has a nicer side: in the age of AI, being the best may not be so important. What matters is being the door.
For half a century, the value in technology has been in innovation. IBM, Microsoft, Google, Facebook… they were all winning by creating something that no one else had. The reading with this step by Apple is that that era may be over: if AI models are updated every quarter and the difference between the best and the second is indistinguishable for 95% of users, what sense does it make to spend 50,000 kilos on research to go behind?
It sounds sexier, especially to investors, to be the one who charges a toll for each interaction. And for that you don’t need the best model, you need the device that people have in their pockets.
That’s the bet: Siri will continue to work, being owned by Apple and running on Apple hardware, but the piece that changes is the intelligence, the LLM. The most expensive piece to develop and the one that possibly provides the least differentiation when you have a billion iPhones.
Apple does not give up something that matters to it at all, but rather outsources the part in which it cannot compete. Bittersweet for the company, bitter for its devotees, reasonable for its investors.
The real deal is not in what Apple pays, but in what it gets. Google pays 20 billion a year for being the default search engine in Safari, and now sells (or delivers, the terms of the agreement have not been made public) the Apple Intelligence feed.
But Apple not only charges, it also receive data on how 1 billion users interact with AI in mobile context:
- You know what they’re asking.
- When.
- How they formulate queries.
- What do they reject?
- What do they repeat?
Google gets better distribution, and Apple gets tremendously valuable training.
If having the best AI is no longer a competitive advantage, what is?
- OpenAI has the best product.
- Anthropic has the best technology.
- Google has the best infrastructure.
- But Apple has the iPhone.
And in a world where AI is gone commoditizingin which one model is valid until the next one arrives three months later, the only moat What holds is the device. There is not so much need to innovate if you control access. You just need what comes through your door to be good enough. AND Gemini is fantastic.
Therein lies the problem. In the age of AI, whoever controls the device can live off income by letting others innovate.
What incentive does Apple have to really improve AI? As long as Gemini works well on iPhones, Apple won’t care if there are models that are 12% better. Their business is collecting the toll, not pushing the border. Innovation still exists and Google / OpenAI / Anthropic / xAI will continue to compete, but Now it is made by companies that do not capture all of its value while it is exploited by those who do not create it..
Welcome to digital rentism. Where the one who controls the door decides how much those who pass through it should improve. AND “Sufficient” always beats “exceptional” when the decider does not pay for the difference. Apple did the rationally right thing. And that, precisely, should scare us.
Featured image | Rubaitul Azad, Dennis Brendel
GIPHY App Key not set. Please check settings