Apple AI – Desperate Days

Desperation makes dangerous bedfellows.

• The idea of Apple using Google to power its on-device AI chatbot is an indication of a worsening position, as using Google to power its AI agent Siri could destroy its differentiation, leading to falling margins and/or market share loss to Android.   
• Having held discussions with Anthropic and OpenAI earlier in 2025, Apple is now speaking to Google about the possibility of using a customised version of Gemini to power Siri.
• It is no secret that Apple is having real problems with its AI as its closed, secretive and privacy-focused approach does not create an environment that is optimal for the creation of a high-quality AI language model that is capable of conversing with and delighting users.
• This is because LLMs are trained using massive amounts of data, which Apple’s privacy policies mean that it has no access to, and even if it did, to use it would be a violation of the promises it has made to its users.
• Data security and privacy are elements of its brand and market position, but in this instance, they are undermining its ability to build a high-quality language model.
• Furthermore, the fact that the company is really struggling to find revenue growth means that its stock option program is not a very attractive proposition when it comes to hiring new talent.
• Fortunately, Apple has some time, as AI agents on smartphones have no influence on the purchase decision by a user, as Google’s puny market share in Pixel devices clearly demonstrates.
• However, going outside of its hallowed walls for a critical piece of technology is potentially dangerous even if it is customised to Apple’s exacting standards.
• RFM Research is far from convinced that AI agents will take over the user experience of the smartphone, but if they do, then Apple in its current form, is in real trouble.
• This is because any user wanting to use Siri to control his or her phone and complete tasks on his or her behalf would rapidly run into the worst user experience on pretty much any smartphone anywhere.
• This means that if users begin to use AI agents on smartphones to carry out tasks on their behalf, then the Apple user experience as it stands today would go from the best in the industry to the worst.
• Left unresolved, this would likely lead to a loss of high-end users from iOS to Android, as well as an erosion of gross margins and a fall in the valuation of the company.
• This is a Nokia-like scenario and is one that Apple will be keen to avoid at all costs.
• However, going to Google to power Siri could easily have a similar, albeit not nearly as severe, effect.
• This is because a customised version of Gemini will always be somewhat behind the state of the art, given the time that will be required to take the latest flagship model and customise it.
• In this scenario, Apple will have a far better solution than if it sticks with Siri as it is today, but crucially, it will never be a better AI agent than Gemini.
• This means that relative to Android, Apple’s differentiation as the best user experience will begin to erode, and the more AI agents become important, the more its differentiation will vanish.
• Going with Anthropic or OpenAI is less risky than going with Google, but both of these companies have ludicrous valuations to defend and taking some of Apple’s brand value in smartphones would be a good way of propping themselves up when the AI bubble finally bursts.
• The net result is that Apple is in a real bind unless it can find a way of bringing its in-house AI up to an acceptable standard.
• One way of doing this would be to acquire one of the other players, leave it as an independent company, but with only one customer.
• This is a viable proposition as Apple has $132bn of cash and near-cash instruments on its balance sheet and is generating around $80bn of cash from operations every 90 days.
• Hence, I think that this is the most likely outcome as it is the one that will give the best result without threatening Apple’s brand or its privacy-strong reputation.
• Apple has time as AI agents remain quite far from controlling smartphones and have little or no influence on the purchase decision, but to be ready, it needs to make a definitive move quite soon.
• Apple is trading on 28.6x 2026 PER, which is meaningfully more expensive than Google and given that Google is in a much better position going forward, owning Google over Apple would be a no-brainer for me.