Emerging Interfaces

The Eye-Tracking Chip Making 'Always-On' AI Glasses Actually Feasible

Ganzin's Aurora IIE cuts eye-tracking power draw to roughly a quarter of conventional NPU-based systems while hitting 120Hz — the kind of power budget that decides whether eye tracking ships in consumer glasses at all.

Eye tracking has been a desirable feature for AI glasses for years and a genuinely hard one to ship, because conventional NPU-based eye-tracking systems draw enough power to make an all-day wearable impractical. Ganzin’s Aurora IIE, showcased at AWE 2026 alongside a broader ecosystem push, is built specifically to remove that trade-off: a proprietary Eye Processing Unit 2 (EPU2) ASIC that cuts power consumption to roughly a quarter of conventional solutions while still supporting up to 120Hz eye-tracking performance.

Why the power number is the actual headline

A minimalist architecture — a single processing chip, two eye sensors, two infrared illuminators — is what makes Aurora IIE compact enough for genuinely lightweight glasses rather than a bulkier reference headset. Eye tracking has typically meant picking between doing it well and doing it all day; a quarter of the conventional power draw is the kind of number that turns “eye tracking as a premium feature you turn on for specific tasks” into “eye tracking that’s simply always running in the background,” which is a qualitatively different design assumption for anyone building an interface on top of it.

The interface layer built on top

Alongside the chip itself, Ganzin introduced Gaze2AI 2.0, which combines voice input with real-time gaze attention — voice carries intent, gaze carries what the wearer is actually looking at — feeding both into the same AI context pipeline simultaneously. That pairing directly answers the design problem this publication has flagged before with combined gesture-and-object interaction research: a system that knows both what you’re saying and what you’re looking at can resolve ambiguous references (“what’s that,” “translate this”) in a way voice or gaze alone can’t.

Why the reference design matters as much as the chip

Ganzin and Jorjin Technologies also introduced the J9 AI Glasses reference design specifically so OEM and ODM partners can integrate eye tracking into their own products without re-solving the hardware integration problem from scratch, and ams OSRAM’s infrared illumination solution was formally folded into the platform for better eye-safety margins at lower power. Taken together, this isn’t one company shipping one product — it’s an attempt to make always-on eye tracking a commodity building block the whole AI-glasses supply chain can adopt, which is usually the more important signal than any single device announcement.