Asymco

Asymmetric Intelligence

Apple Intelligence: Show me the money!

Research · 3 comments

by

AI sales are booming. AI FCF not so much. So what new cash flows does Apple Intelligence unlock?

Portrait of Charles Babbage (c. 1820)

Apple's release of a new framework for AI at WWDC26, centered mainly on new layers in the device/OS which process intent to either on-device or world models and couple it to experiences which are system-wide, including apps.

This is how the data and the inferences flow but what about the cash?

I am writing without a full picture in mind, and perhaps one does not yet exist, but my assumption for at least three years has been that the value of AI will split between the creation of (frontier) models and their distribution. Apple chose to focus on local models, license or partner for frontiers and stitch together the means for distributing both.

That is until on-device will be good enough to perhaps bring all models in-house.

But distribution is always where it's at. So how do the new announcements affect this?

Let's follow the money. The costs of inference will be partly borne by Apple for what they keep in private cloud but mainly by apps and hence subscriptions which will include marginally increasing compute costs. This will lead to increasing or new subscription pricing and the usual commissions for App Store distribution.

The fact that this is all implied rather than explicitly stated should be seen as a tacit acknowledgment that the Services business will not change much, if at all. It will perhaps benefit from an uplift if the new AI-powered apps are very popular.1 This new AI might create more demand or expand existing demand. That will all depend on the innovations developers come up with.

But the elephant in the monetization conference room is hardware. AI demands a huge step-function increase in computing resources--both compute and storage and integrated memory. The minimum requirements are already leaving two to three year old hardware behind.2

Again, if AI is compelling enough, users will need to upgrade. Not just their phones but the Macs, iPads, Vision Pros, Watches, probably AirPods too. Then there is Home which is seeing foreshadows of a significant AI enabled upgrade. This means HomePods but also probably new devices altogether. Rumors of spectacles coming are par for the course. Health will also be revolutionized.

This idea that AI is monetized through hardware should hardly be a surprise…

… Right now the greatest beneficiary is Nvidia which won the AI lottery by having the right chipsets at the right time. And with only a market of 5 customers and a few million processors it became worth $5 Trillion.

Apple has the processes and priorities in place to drive a consumer device AI experience with 1.5 billion customers and a few billion processors over the next decade. Perhaps even longer.

Apple, let's not forget, also has been monetizing software and services with hardware. Unlike the PC and Android world where software profits (and margins) accrue to software makers while hardware is left to starve, for decades Apple's integration meant that its business model afforded it great margins that could not be replicated.

  1. Services has a lot of headroom for growth as non-consumption is still pretty competitive
  2. But don't worry, adoption curves being what they are, and regional/political obstructionism will imply a decade or more before perhaps half the install base is using Apple Intelligence with some frequency.

Subscribe to Asymco One for more in-depth analysis and commentary like this.

Asymco One gives you full access to our research, analysis, and commentary: Nota Bene, Research, the weekly Bulletin, the Asymetric podcast, and monthly Office Hours. Subscribe at asymco.com/one.

Posted by:


3 comments

  1. deemery ·

    For what it's worth with respect to Apple Intelligence, I'm still at "show me the utility!" stage. I don't have any desire to talk to computers/phones, and the other features haven't generated any "ooh, I'll want to try that," either.

    But I AM hoping the next release of Mac OS fixes some of the bugs I've had for the last several years.

  2. farshad ·

    Part of the problem is that AI is enabling technology that can afford multiple jobs to be done. It’s better to look at each job and see how it might impact you:

    1. AI as a programming assistant ⇒ indirectly helping you because of the ‘help apple fix bugs’. I suspect these massive performance improvements may have been due to judicial application of AI. It’s not that AI figured them out, it is that AI opens the potential of trying a new protocol without bogging down the entire team for 2 months.

    2. AI as an app enabler ⇒ help developers build more adaptive apps. We didn’t know the utility of the front facing camera either. And look at its impact.

    3. AI as computational search ⇒ anyone who has cursed at photos app would use this.

    4. AI as general language fluent assistant ⇒ ask Siri: ‘where was that thing that lets me type dash greater than and get a real arrow’.

  3. deemery ·

    'programming assistant': I am cautiously optimistic about AI finding bugs. And maybe, aiding in the construction of proofs-of-correctness. But I'm not coding any more, so not much use for me.

    'More adaptive apps': Yes, please. I'm looking forward to those. Particularly photo analysis (see below).

    'Search engine': I've played with on-laptop cataloging of photos, mostly my collection of train pictures I've cribbed from the Internet (about 3200 of them....) The analysis side isn't sufficiently specific to be of much use to me, and the error rate was surprisingly large. I asked "show me tank cars", and about 50% of the photos returned by that simple query were not tank cars, but other kinds of railroad cars or a few other strange things.

    What this would require is for a way for me to BETTER TRAIN that system to provide more analysis of things I'm interested in. (And that takes me back to something I said a couple years ago. "Learn to train" will replace "Learn to code"! That's my $.02 on 'agentic AI', too :-) )

    'AI language assistant': Meh. See my comment about 'not talking to computers'. Maybe someday that will change. But for now, the only time I talk to my computer or phone is to swear at it, and I'm not expecting an answer.

    But for sure, "We live in interesting times!"

Leave a comment

Getting the form ready…