Finn Brooks: Rough week for anyone trying to argue Apple doesn't own wearables — did you see what Counterpoint dropped?
Clara Bennett: The Q1 Global Smartwatch Shipments Tracker, yeah. The ninety percent figure.
Finn Brooks: Apple Watch, ninety percent of all edge AI smartwatch shipments worldwide. Which sounds like a knockout until you go one line down in the same report and it says Apple has twenty-three percent of the total smartwatch market. And I just — wait, that doesn't — those two numbers can't both be describing dominance in the same way.
Clara Bennett: Right — they describe dominance in a very specific sub-category. The question is whether that sub-category reflects something real, or whether Counterpoint built a fence around Apple's existing hardware and called it a new market.
Finn Brooks: Which is exactly where I landed. Because the definition — edge AI means a dedicated NPU running at least one primary health or safety feature locally, on-device — that definition fits the Apple Neural Engine in the S-series chips perfectly. Samsung Galaxy Watch is rolling out Gemini right now. Google Pixel Watch, also Gemini. Cloud inference. Still excluded from the count.
Clara Bennett: Now, is cloud inference actually worse for the user, or is that just a hardware spec that most people never feel?
Finn Brooks: That's — okay that's the real question, because the edge AI segment grew seventy percent year-over-year versus four percent for the broader market. If users couldn't feel the difference, why is that sub-segment exploding?
Clara Bennett: The growth could reflect Apple's installed base absorbing a new label, not new buyers choosing edge AI over cloud. Those are very different things.
Finn Brooks: Okay but — actually, wait, that framing cuts both ways. If it's just relabeling, why did Apple's total shipments rise twenty-one percent in Q1 while the whole market crept up four? That's not a reclassification artifact, that's people buying.
Clara Bennett: That's fair — and that's exactly where I want to pump the brakes on the category-is-fake argument. The definition isn't arbitrary. Think of it this way: edge AI is the calculator sitting on your wrist. Cloud AI is faxing the math problem to a server somewhere and waiting for the answer back.
Finn Brooks: The fax machine thing — yeah, that lands.
Clara Bennett: So fall detection, arrhythmia monitoring, sleep inference — those are running locally on the Neural Engine right now, no round trip. The latency difference is real, the privacy difference is real, and the battery tradeoff is real. Those aren't marketing bullets.
Finn Brooks: Right — but here's what I keep tripping on. Samsung and Google didn't miss the NPU memo by accident. They're dependent on third-party chip vendors for Wear OS. So is Apple's lead actually about being smarter, or is it just... Apple controls the silicon, controls watchOS, controls the health platform — they could do this because no one else has that whole stack?
Clara Bennett: That is precisely the point. Vertical integration isn't a side note — it's the structural cause. Apple didn't get lucky with the Neural Engine. They designed the S-series chip specifically for watchOS constraints, which is why Mohit Agrawal at Counterpoint framed this as edge AI moving from premium differentiator to mainstream expectation. Apple could make that move. Wear OS couldn't, yet.
Finn Brooks: Wait — so Gemini on Galaxy Watch and Pixel Watch, that rollout happening right now, it's still cloud-hybrid because the underlying chip architecture can't support full on-device inference at that level?
Clara Bennett: In practice, yes. You can ship Gemini features — and Samsung Galaxy Watch and Google Pixel Watch are doing that — but shipping Gemini through a cloud round-trip and running inference locally on a dedicated NPU are architecturally different things. One of them Counterpoint counts. The twenty-one percent shipment growth Apple posted isn't a Siri marketing story. It's the fingerprint of owning the whole platform.
Finn Brooks: And that's — okay, I'll grant that. The cloud-hybrid thing is real, not just a spec on a sheet. Picture a runner, mid-5K, checking her wrist — Apple Watch fires stride analysis and breathing alerts right there, no ping, no wait. Galaxy Watch is doing Gemini, but that query is making a round trip to Google's servers. That latency delta, that privacy gap — that IS what the ninety percent actually measures.
Clara Bennett: That's the kernel. Yes.
Finn Brooks: But here's where I get stuck — Samsung and Google launching Gemini right now, that's the starting gun, not the finish line. Counterpoint is projecting edge AI penetration hits thirty-two percent of all smartwatch shipments by end of 2026. It's already at twenty-five in Q1. That window is closing fast. So does the ninety percent crumble once the hardware catches up, or — wait, actually — is the hardware even the real moat here?
Clara Bennett: It's not. The hardware is the entry ticket. The moat is years of health data, watchOS muscle memory, the iPhone lock-in. Gemini doesn't dissolve any of that.
Finn Brooks: No, and that's — I mean, that's actually the part that's hard to overstate. Absolute unit gap between Apple and competitors is still widening even if competitors grow share, because Apple is also growing inside the fast-growing slice.
Clara Bennett: Right — so Samsung Galaxy Watch and Google Pixel Watch can add Gemini features today, but they cannot add three years of a specific user's arrhythmia history and sleep baseline. That data doesn't transfer.
Finn Brooks: Which is a switching cost that has nothing to do with chip specs.
Clara Bennett: Exactly. And that's why Mohit Agrawal's framing — edge AI going from premium differentiator to mainstream expectation — is true AND slightly misleading. Mainstream expectation just means competitors have to show up with the hardware. It doesn't mean they inherit the installed-base gravity.
Finn Brooks: And honestly, there's a version of this whole conversation where the ninety percent turns out to be beside the point — because Meta is already monetizing AI wearables on a completely different form factor, smart glasses, and that track may end up defining where this goes. We'll get there.
Clara Bennett: And that Meta thread is actually where the calibrated version of this whole story lives — because ninety percent of a category is only as valuable as that category's future. Meta is already charging nineteen ninety-nine a month for 'Conversation Focus' on its AI glasses. That market is monetizing on its own timeline, completely outside the wrist.
Finn Brooks: Wait — nineteen ninety-nine a month? That's a subscription. That's not a feature, that's a revenue model.
Clara Bennett: Right — and that matters because it signals the smart glasses market is past the hobbyist phase. Meta isn't testing whether people will pay. They already know. Now, the uncomfortable question that raises: if ambient AI migrates to your face or your ears in the next eighteen months, Apple's Neural Engine advantage is wrist-specific. It doesn't transfer.
Finn Brooks: Which means — wait, actually that reframes the whole Mohit Agrawal claim. He says edge AI is becoming mainstream expectation. Mainstream on what form factor, though? Because if the answer stops being 'the wrist,' Apple's vertical integration story has to be rebuilt from scratch for a completely different device category.
Clara Bennett: Exactly that. The watchOS ecosystem, the Neural Engine in the S-series, the health data moat — those are watch-specific structural advantages. None of them automatically extend to smart glasses or earbuds.
Finn Brooks: So the defensible claim is... the ninety percent is real, the structural advantage behind it is real, but it's a dominant position in a category that may not define the wearable AI winner.
Clara Bennett: That's — in practice, yes. Apple Watch owns the wrist. That's not nothing. But Meta is already monetizing a parallel track, and the form factor question is open in a way that the chip architecture question isn't.
Finn Brooks: The hot take was right that the definition favors Apple. The hot take missed that the definition might stop mattering.
Clara Bennett: Which — that's actually where we started, isn't it. Rough week for anyone arguing Apple doesn't own wearables. Except now the answer is: they own a wrist. A very profitable, ninety-percent-of-the-edge-AI-category wrist. That's real. But Counterpoint's own projection — thirty-two percent penetration by end of 2026 — that's the clock ticking on how long the wrist stays the definition.
Finn Brooks: Yeah, and I mean — when Gemini actually closes the inference gap on Galaxy Watch and Pixel Watch, like genuinely closes it, the test isn't whether anyone switches. It's whether Apple's health-data moat and watchOS lock-in hold at feature parity. And if they do... that stops being a product story. That's an antitrust conversation.
Clara Bennett: That's probably where this lands. Not with a winner — with a regulator.
