This analysis was first published in SvD Näringsliv, in Swedish, on April 9th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.
While competitors spend hundreds of billions on AI, Apple sits tight. It looked like a major mistake. But the next phase of AI development could be Apple’s revenge.
In a red leather armchair on a stage sits Steve Jobs. It is 2010, the year before he passes away, and Jobs is visibly physically affected by his illness.
His sharp tongue is still very much intact, however. When journalist Walt Mossberg asks about “search company Siri,” which Apple recently acquired, Jobs cuts him off immediately: “I don’t know if I would describe Siri as a search company,” he says. “They’re in the AI space.”
Somewhere around there — sixteen years ago — began Apple’s public journey toward AI. But the road has been anything but straight. After repeated disappointments with Siri, which neither understands what you say nor can do very much, the image of Apple as an AI company has faded.
Have they missed the wave they themselves were so early to identify? It is not quite that simple.
ChatGPT, Claude, and Google Gemini have admittedly sprinted past and created exactly what Siri was supposed to become. Companies like Meta, Google, and OpenAI have invested hundreds of billions of dollars in AI data centres.
In the short term, Apple is behind — but in the long term there is an opportunity for a strategy that competitors will find very difficult to replicate. At least at the same scale that Apple has.
If we begin with the shortcomings, they are obvious even internally at Apple. A licensing agreement with Google from January speaks clearly: Apple currently has no AI model of comparable quality to the competition. Siri is now to use Google Gemini going forward. Apple has had to reorganise its AI work after its previous head suddenly left prematurely.
Many Apple fans consider it a smart strategy to work with partnerships rather than participate in a very expensive race with other tech giants. What argues against it is all the work that has been ongoing with Siri over all these years. It is the inability to produce a good result that led to the partnership with Google — not a decision to not try. And the resources are there. Apple sits on around 1,350 billion kronor in cash and securities.
Looking further ahead, however, one can begin to glimpse something exciting — and entirely unique to Apple. Should AI development move in a certain direction, Apple could catch up despite its sluggishness.
In simplified terms, AI development can be divided into two parts — “training” and “inference.”
Training is the training of the AI model itself, a process that requires enormous computing power. When Anthropic and Meta release new models, the price tag for training them is counted in tens of billions of kronor. The enormous data centres filled with Nvidia’s GPU chips are used primarily for this type of training.
Inference is when the model is used to give you as a user an answer. It is drawing conclusions from the information in the model so that it can tailor responses to what you have asked. This does not require anywhere near the same type of computing power as training. And even more importantly in this context — in many cases it can be done directly on your phone, or your iPhone.
As AI models become more efficient, Apple’s own chips are emerging as a strong strategic advantage. In the over one billion iPhones in use today there is hardware that Apple does not sell to anyone else. With newer generations of AI models, one can imagine them being trained separately but then run on the user’s own device. Apple Silicon — their own chip — could then answer questions for the user without the information ever needing to leave the phone. It is private, fast — and everything you need is already in your phone.
Looking at Apple’s competitors, none of them has this capability. Meta has no mobile phones at all, and Google’s Pixel phones have a very low market share globally. OpenAI is working on various hardware products but nothing has been released yet. And unlike building data centres, it is next to impossible to roll out a new mobile phone that achieves the same coverage as the iPhone has. It does not solve everything when it comes to AI development, but it is an enormous advantage in distribution.
Apple’s journey toward AI seemed to begin with Siri in 2010. But the truly unique difference only started to become visible ten years later, with the launch of the Apple Silicon chips. Here Apple has something competitors find very difficult to match: a technology they developed themselves, and which is perfectly suited to AI.
It is a long game that Apple is playing. They will also need to get their own AI development in order over time. The need for AI training is not going to diminish. But the trump card is already in place.
You have it in your pocket — inside your iPhone.
