The Apple that sold you spatial computing as the next paradigm just deleted it from the roadmap and handed Siri's brain to Google — in the same breath, both moves approved by the same incoming CEO.
Read that sentence twice. It scans as two panicked retreats. The argument of this piece is that it's one disciplined bet.
Last week, in the Ternus Era thread, we flagged the shape of the transition and the irony hanging over it. This week the blank got filled in — with concrete numbers, a canceled product line, and a rival's name stamped on the inside of Apple Intelligence.
Start with the roadmap, because it's the cleanest signal of intent. Per analyst Ming-Chi Kuo, Apple's XR and glasses lineup now contains exactly two products: display-less AI glasses in the Ray-Ban Meta mold, reportedly targeted for 2027 mass production, and optical-waveguide AR glasses pushed all the way out to 2029.
The Vision Pro line — the thing Apple spent years and billions positioning as the future of computing — has been removed entirely. Not iterated. Not shrunk into a lighter model. Removed.
Kuo is explicit that John Ternus approved this, and that Apple had reportedly quietly started executing it already. He also calls it the right call. That's the tell.
Because this isn't a hardware failure in the usual sense. Vision Pro worked. It was a marvel. What it never had was a consumer case — a reason a normal person needed a face computer badly enough to buy one and keep using it.
Ternus is, by trade and temperament, a hardware and design guy. The reporting around his ascension, per Bloomberg, frames him as the executive pulling Apple back toward product engineering and design primacy after years of finance-led priorities. A design-led leader looks at Vision Pro and sees a beautiful answer to a question customers weren't asking.
So he triages it. Killing Vision Pro frees engineering, money, and organizational attention to chase a category that is actually winnable: lightweight, all-day glasses that lean on AI instead of on displays.
That's the first half of the reset. The second half is the Gemini deal, and it's the same move wearing different clothes.
At WWDC 2026 — Tim Cook's last as CEO — Apple rebuilt Siri from the ground up and rebranded it Siri AI, complete with its own standalone app, bubble-style continuous conversation, history, iCloud sync across devices, and stronger understanding of what's on your screen.
Underneath all of that sits Google's Gemini. Apple reportedly made Gemini the underlying architecture of Apple Intelligence, running both device-side and in the cloud. The most important software brain in Apple's ecosystem is now licensed from a competitor.
Sit with how strange that is for this specific company. Apple's entire identity is control of the whole stack — silicon, OS, and the intelligence on top. It just conceded the intelligence layer to the one rival that already builds the best models and ships its own hardware to run them on.
But strange isn't the same as incoherent. Read the two moves together and a single thesis appears: a design-led Apple has quietly admitted it cannot win the model race, so it's betting everything on being the best-integrated wrapper around someone else's AI.
A design-led Apple has quietly admitted it cannot win the model race, so it's betting everything on being the best-integrated wrapper around someone else's AI.
Kill the hardware you can't justify. License the model you can't build. Redirect everything toward the product where Apple's actual advantage — integration and taste — is the whole point.
This is defensible on paper. Even a sympathetic reviewer of Siri AI concedes Apple isn't trying to out-muscle OpenAI or Google on raw capability; it's making AI feel native to the ecosystem, where the iPhone's camera and Siri's context understanding work together in a way rivals struggle to replicate.
That's a real moat. Integration and taste are things Apple genuinely does better than anyone. If your differentiation has to be something, that's a good something to have.
Here's the trap.
Integration and taste are exactly — precisely — what a display-less pair of 2027 glasses has to prove. Glasses with no screen aren't a spec-sheet product. They live or die on how well the software feels, how seamlessly the assistant anticipates you, how invisible the seams are.
Which means Apple's entire glasses bet rests on the quality of an experience whose brain it now rents from Google.
Line the timeline up and the exposure gets sharp. Gemini powers Siri today. Meta reportedly already ships Ray-Bans, in volume, in the category Apple won't reach until 2027. The waveguide glasses — the ones that would actually showcase proprietary Apple display technology — don't arrive until 2029.
So for roughly the next two years, Apple is design-led on the outside and rival-dependent on the inside, with no glasses in market to justify the arrangement.
That's the structural risk. Dependence isn't fatal in year one. It becomes fatal when it hardens — when Gemini stops being a bridge and becomes load-bearing infrastructure, when Google's roadmap starts dictating Apple's, when the leverage quietly inverts.
The 2027 glasses are the deadline for proving the wrapper strategy works before that hardening sets in. Ship a device where Apple's integration makes borrowed intelligence feel unmistakably Apple, and the Gemini deal reads as pragmatism. Slip it, and Apple spends the back half of the decade as a beautifully designed front end for a competitor's model.
There's one more piece, and it's the counterweight. On August 13, Cook and Commerce Secretary Gina Raimondo cut the ribbon on a new Houston facility that is reportedly already shipping Apple's first advanced AI servers — hundreds of millions invested, with Mac Mini production coming later this year.
Read that as the hedge. Apple may be renting the model, but it's building the physical layer the model runs on — its own AI server assembly, on domestic soil, under its own roof. It's the part of the AI stack Apple still fully owns, and it's quietly insurance against total dependence.
That's the shape of the Ternus reset. Not two retreats. One bet, with a physical hedge underneath it.
And it only pays off on a clock. The glasses have to ship in 2027, they have to feel like Apple, and they have to do it before renting Google's brain stops being a shortcut and becomes the thing Apple can't operate without.
Coherent, yes. Comfortable, no. The reset is clarity. Whether it's the right clarity, the 2027 glasses will decide.