At the Milan–Cortina Winter Olympics Chinese startup RayNeo unveiled a line of AI‑enabled AR smart glasses, using the global sporting stage to demonstrate hardware and software that backers say are finally ready for mass adoption. The appearance is more than a marketing stunt: brokers and securities analysts in China are calling 2026 the potential “take‑off” year for on‑device artificial intelligence, pointing to concurrent advances in model compression and edge compute capacity.
Analysts at Changjiang Securities argue that recent breakthroughs in making large models lighter, together with a steady rise in on‑device computing power, have removed many of the technical barriers to scaling AI at the end‑point. Improvements in model quantization, pruning and distillation, and the proliferation of specialized NPUs inside mobile SoCs now make it feasible to run useful inference tasks locally, lowering latency and easing privacy concerns compared with cloud‑only approaches.
That progress is already reshaping the commercial case for devices. On‑device AI creates a new monetisation route for model developers and a fresh consumer entry point for technology companies: smart glasses become not just peripherals but potential hubs for personalised, always‑on services. Equally important is the rapid iteration of product form factors and software — a dynamic that benefits component makers and assembly specialists focused on optics, imaging modules and precision manufacturing.
Investors are paying attention. Galaxy Securities singled out AI glasses as a key hardware category to watch, and firm filings show supply‑chain players are positioning themselves. Rongqi Technology has reportedly expanded precision assembly equipment for social‑platform partnerships, including work tied to Meta’s smart glasses, while Crystal Optoelectronics claims a decade‑long footprint in AR optics and camera modules, spanning display elements to 2D/3D imaging units.
The implications extend beyond consumer tech. On‑device AI reduces dependence on large datacentres for many applications, shifting some value upstream to silicon designers and component suppliers. That creates new battlegrounds over standards, interoperability and control of the data and user experience. At the same time, device makers still face stubborn constraints: battery life, heat dissipation, visible form factors and the need for compelling content and applications remain decisive for consumer adoption.
Geopolitics and regulation will shape the trajectory. China’s ability to supply a wide swathe of the AR glasses value chain — from optics to assembly — gives its firms an export opportunity even as Western competitors press on with their own platforms and chip designs. Regulators in multiple jurisdictions will also scrutinise privacy and safety issues arising from always‑on wearable cameras and localised inference, which could slow rollout or spur new compliance costs.
If 2026 becomes the “AI endpoint” inflection point the markets expect, the winners will be those who combine efficient on‑device models, affordable and comfortable hardware, and a thriving app ecosystem that leverages local inference. The Milan showcase is a visible signal that Chinese players have the components and integration capabilities to contest that race; converting technology demonstrations into mass adoption, however, will hinge on price, battery performance and the emergence of services that consumers care enough to pay for.