Where is the monetization? Follow the compute
Office Hours Research #Q&A #WWDC · 3 comments
by Horace Dediu
An Office Hours question asked by Roger S., June 9, 2026.
Q: Beyond selling more capable hardware — phones, Macs, iPads — is there real monetization in all of this AI?
Absolutely. They gave us an architecture: it’s about intent, about orchestration, about routing a request through multiple models and input types, and then deciding whether it goes through an app or back to Siri. Where’s the money? My answer is: follow the compute.
If this works on-device, the user is happy buying the device, with the memory it needs, and Apple bundles all of that into the purchase price. Developers still sell their apps, but they’re not selling extra compute. But if a result requires heavy computation in the cloud, through an app, then trace it: who sends you the bill? That invoice comes from somewhere. If it’s a developer or a model supplier, it isn’t free — they’ll ask you to subscribe to get the answer, priced as a fixed fee or variable. Apple gets its usual commission, thirty percent or fifteen, somewhere in between.ssss
If Siri depends on services that don’t go through an app, then Siri charges back through iCloud for the compute. So all they’ve really done is take the exact same business model — Services with subscriptions — and either fold compute into existing subscriptions or add new ones. Apple gets the same distribution commission it always has. Nothing changes. In fact, the absence of any commentary from Apple about new economics is, to me, a tacit acknowledgment that nothing different is needed.
That means the App Store, and if anything economically interesting happens, you’ll see another kink in its trajectory — the way we saw a kink when subscriptions became dominant. To users, it’s invisible. Maybe the monthly services bill rises ten percent; maybe they add a couple of subscriptions and drop one. It all reflects in Services, at a 72%+ gross margin, with a hardware uptick alongside. I reckon it’s a ten-year cycle before half of the Apple base are AI active daily, paying something for it. Developers will just see more of the same, only bigger.
Editor’s Note: This was one of the questions asked by participants in Asymco’s June 2026 Office Hours live Q&A session, open to Asymco One subscribers.
Subscribe to Asymco One for more in-depth analysis and commentary like this.
3 comments
rbgayle ·
One of the hallmarks of this era is decentralization of power to the edges. Mainframes. Minicomputers. PCs. smartphones etc. The hyperscalers are all trying to centralize AI where everything goes through them. No matter what someone actually needs. And it allows them to charge a toll for anyone to use AI, and increasing that toll to increase profits. This is counter to historical trends.
Apple continues this historical trend by decentralizing AI usage to the device and the actual needs of the user. It gets money the more it makes AI useful, not by forcing people to use cloud AI no matter what.
deemery ·
That's a good point. But, the hyperscalers would respond, "You're substituting 3-5 big AI companies with a -single- gatekeeper. Where's your competition there?" And I'm sure they'll try lawsuits to enforce that point of view. (If you can't win in the markets, try the courts....)
FredHS ·
Yup.
Maybe that's why Jensen announced, just before WWDC, his RTX Sparc for PCs.
While they offer a great chip and CUDA, Apple literally offers the full stack up, down, and sideways (as in cross platform).
They will first arrive this fall, POR, for the high-end of the PC and Surface market. Questions remain about fragmenting the PC clone market, cross platform capabilities, and system stability.