Bloomberg’s Mark Gurman, who has stellar track record at reporting on internal projects at Apple, believes the company is in talks with Google to licenses Gemini and OpenAI to license GPT. That news comes a day after Apple released a paper that it is working on its own multimodal AI model, similar to Gemini and OpenAI.

The bottom line is Apple knows it needs to move fast and is exploring all options when it comes to adding generative AI into their products. In the end they may do both, come out with their own model to power some features across their product line and license a model to power other features. We should hear more on the topic in a month during the March quarter earnings call.

If Apple was to license a model, I bet they would pay several billion dollars a year. There are three key topics related to the cost dynamic:

1. It would probably be usage-based vs a flat guarantee like the search deal (Google pays Apple around $15B a year for default position in Safari). The typical pricing for AI models is based on tokens used in API calls.
2. How much Apple pays Google depends on what they use it for. If it powers the majority of the AI features on the phone, then it would probably be in the billions. If it’s just an offering for limited use cases, then it would be in the hundreds of millions. My sense is it would power the majority of the highest value AI features on the phone.
3. Apple will negotiate a discount from the stock rate card because Apple holds the high cards with proprietary access to about 1.4B of the most attractive customers worldwide.

In the end, Apple is still in the pole position to lead in personalized AI. While the potential of licensing a foundation model adds some complexity around who has the rights to your personal data (Apple or Google for example), I believe in the end Apple will find a way to get users comfortable with opting into personalized AI.

A likely form that personalized AI would take is an AI subscription model, like the existing iCloud, Apple TV+, News, Music, etc. My guess is it would be priced similarly at about $10/month to have access to personalized AI.

Apple has 2.2B active devices and I estimate there to be 1.4Billion users. If 15% of its users subscribed to Apple’s personalized AI, it would add 6% to their overall revenue.

Apple has been designing an electric vehicle since 2014, about three years after Tim Cook took over as CEO. The project, known as Titan, never had stable leadership which meant its go to market differentiator (will it be full self-driving or advanced driver assistance) changed three times during its 10-year run.

From the beginning Apple did not want to follow Tesla and develop multiple models at different price points. They wanted to go big and aim for a superior, fully autonomous vehicle without a steering wheel or pedals. Over time, fully autonomous became semi-autonomous as technical challenges were faced with automation.

As painful as it is to see the car get canceled (given the car market is about 4x bigger than the smartphone market), it’s the right decision for Apple given it frees some investment dollars, and more importantly shifts the company’s growth focus to AI.

Over the past seven years, Project Titan was always a long shot, and on the edge of being canceled. The straw that broke the camel’s back was Apple could not realize AGI’s potential unless resources and the growth focus was aligned.

The first benefit to ending Titan was I believe it frees up between $4-$5B annually, or about 11% of the company's overall R&D spend, to AGI.

In FY23, Apple’s reported net income margin was 25.3%. Excluding investment in Titan, I estimate the company would have reported 26.3%.

As I see it, directing investment dollars to building AGI was half of the benefit of shutting down Titan. The company has plenty of money to invest a few billion dollars a year into building a car.  For example, FY23 the company had operating income of $114B.

The second benefit to ending Titan was it clearly focuses the company's growth North Star on AGI.

I’ve written about the potential of AGI over the past eight years, much of that has been over the past year. While there will be ebbs and flows in the growth of AI, just like we saw over the 15 years it took to build out the internet, I believe the substance of AGI will surpass today’s high hype bar.

If that’s the case, the DNA at Apple needs to be rewired from a device, software, and services company to a device, AI, software, and services company, in that order.

Apple’s margins have defied the laws of gravity for a decade. Despite total revenue being down 2.8% in FY23, net income margins were 25.3%, essentially in line with FY21 record 25.9% margin. That compares to Apple’s 21.7% margin in FY13.

As I look forward to how margins will trend over the next couple of years, it may be worthwhile to read between the lines. Given all that Apple has done over the past month, ending Titian, announcing their own multimodal model, and being in talks with Google and OpenAI to licenses their models, it’s clear the company is in investment mode. When it comes to AI, Big Tech doesn’t sip AI investment dollars, they chug it. Just look at NVDA’s share price.

Apple's AGI ambitions go beyond building or licensing a model. They'll likely include data center build outs, an area that Apple has lagged Google and Meta over the past five years. Closing that gap will likely have a slight negative impact on margins.

My guess is over the next couple years, starting in the back half of this year, we will see net income slowly decline, losing 130 bps, to closer to 24.0%, down from 25.3% in FY23. This implies an annual step up in AI investment of about a $5B. Some of that will be capex and will find its way to the net income line over deprecation in the years ahead.

In the end, embarking on an AI investment journey is the right decision for the company to see even higher margins in the long-term.

Disclaimer