AI Glasses Stop Being a Demo: Shipments Up 263% and Alibaba Takes On Meta

Artificial-intelligence eyewear spent three years as a demo category. In 2026 it stopped being one. Counterpoint Research puts global AI-glasses shipments in the first half of 2026 up 263% year on year, with display-free models accounting for 96% of the volume and the display-equipped AR/AI segment growing 449% year on year off a much smaller base. That is no longer a rounding error in the consumer electronics market — and the component content per unit is where the interesting part begins.

Alibaba puts a flagship on the table

On 13 October 2026 Alibaba’s Qwen team launched two AI glasses — the N1 and N1 Pro — plus a clip-on AI earbud, and put all three on sale the same day. The headline specification is a 50-megapixel camera system with 4K video and multi-axis stabilisation, paired with bone-conduction microphones so the glasses can be spoken to without a boom arm or an earbud in place. Reported list prices sit at RMB 2,899–2,999 (roughly US$410–450) for the N1 and RMB 3,399 for the N1 Pro, with launch subsidies bringing the N1 to RMB 2,218. The Qwen Clip earbuds, developed with Bose, list at RMB 1,299 and were promoted from RMB 1,099.

The Pro adds three things worth noting for anyone reading a bill of materials: real-time vital monitoring covering heart rate, blood oxygen and skin temperature; eye-tracking interaction; and iris verification for payment authentication, developed with Ant Group and quoted at a false-accept rate below one in a million. Eye tracking and iris recognition are IR-sensing problems long before they are software problems, and they add emitter, filter and low-power image-processing content to a device that already has to run all day on a battery sitting in a spectacle arm.

The chip doing the always-on work

Himax disclosed on 14 October that its WiseEye HX6538 processor sits inside the N1 series, running an always-on visual-sensing path that is separate from the main camera and wakes the rest of the system only when it detects a person or a relevant event. Himax quotes about 5 mW for that mode. Whether or not the figure survives independent testing, the architecture is the point: the market has settled on a split design — a micro-power vision front end that watches continuously, and a higher-power application processor that only wakes for real work. Alibaba quotes roughly 25 hours of standby or 9 hours of music playback, with batteries in the arms that can be swapped during use.

The market behind one launch

• Meta still owns the category. Counterpoint put Meta at roughly 94% of AI-glasses shipments in the first half of 2026. Its Ray-Ban line was refreshed at Meta Connect on 23 September: Ray-Ban Meta Gen 3 from US$449, the display-equipped Ray-Ban Meta Display at US$799 with a 600 by 600 panel rated 5,000 nits at 90Hz, a US$349 audio-only model shipping 13 October, and a US$249 entry model following on 23 October.
• Volume is compounding. A tracker that aggregates Counterpoint, IDC and Wellsenn XR data models about 9.8 million smart-glasses units in the second half of 2026, up roughly 36% on the first half and enough to take the full year towards 17 million — more than double 2025.
• Display is the next leg. TrendForce expects AR glasses with displays to reach about 950,000 units in 2026, up 53%, with Meta’s optical-component orders revised sharply upward during the year.
• China is pricing the category down. RayNeo, Xiaomi and Alibaba are pushing sub-US$300 AI glasses into fourth-quarter promotions, and Chinese suppliers lead the display-equipped segment.

What this means if you build hardware

• Camera modules are the new spec war. A 50-megapixel still sensor plus 4K recording inside a frame the weight of a pair of glasses is an optical, thermal and flex-assembly problem. Module yield, not sensor cost, tends to set the price.
• Always-on sensing changes the power budget. Micro-power vision front ends, IR emitters and low-quiescent-current regulators move from optional to mandatory, because users judge the product on whether it survives a day.
• Rigid-flex and high-density interconnect get harder. Arms that hinge, batteries that swap and sensors that sit behind a lens all want thin, flexible, high-layer-count boards with tight impedance control.
• Substitution risk is concentrated. Waveguides, microdisplays, IR sensor dies and low-power vision SoCs each have few qualified sources. If your design depends on one of them, that is programme risk, not engineering risk.
• Cost targets are aggressive. With Chinese brands aiming at the sub-US$300 shelf, a five-point difference in scrap rate decides whether a design is viable.

Our position

We are not a smart-glasses brand, and we would not pretend otherwise. What we are is a Dongguan manufacturer that builds rigid, flex and rigid-flex boards, sources connectors, passives and sensor content, and assembles boards for customers bringing consumer devices to market. Wearable projects live or die on assembly tolerance and test coverage, and that is the part we do. If you are prototyping an AI wearable — or any small, high-density board — send us the stack-up and the placement density and we will tell you honestly whether it is a build we should quote. Our PCB and PCBA capability list is here, and our note on the MLCC supply cycle covers the passives side of the same bill of materials.