By |
Edited by | Wen Shuqi
In the tech world, the debate over the form factor of AI hardware is far from over. From AI cards to AI glasses to AI earbuds, companies around the globe are experimenting with different product shapes, all trying to figure out how to move from smartphones to the next AI-powered computing platform.
Guangfan Tech is one of those startups jumping into the ring, hoping to snag the chance to define what AI-era hardware should look like.
Founded in October 2024, Guangfan Tech has already pulled off several funding rounds in just over a year. Its backers include big names like Shokz, Goertek, Lenovo, XPeng, CATG (BoRui Capital), GigaDevice, plus financial investors such as CDH Investments, Alpha Startup Fund, and InnoAngel Fund.
Recently, the company officially launched its first product—the “Guangfan AI Full-Sense Wearable” set that includes a pair of AI earbuds and an AI watch. Price tag? The bundle goes for 2,099 RMB, while the earbuds alone are 1,999 RMB.
From a form factor perspective, it’s still a typical consumer electronics combo: earbuds, wearable, mobile connectivity. But product-wise, Guangfan wants it seen as an always-on, context-aware AI assistant.
Guangfan chose earbuds as its entry point because they’re less intrusive to users’ existing habits. CEO Dong Hongguang noted in an interview that smart glasses are still held back by optical display tech, weight, and aesthetics—making them a tougher sell. Earbuds, on the other hand, are a mature category with a huge existing user base. Adding AI features on top of what people already use is the path of least resistance.
“We’re adding to an existing category, not trying to teach users to accept a brand-new one,” Dong said.
The Guangfan AI Full-Sense earbuds aren’t your typical Bluetooth earbuds. Their standout feature? A tiny camera built right in. That comes from the team’s core belief about future interaction: visual perception is a must for proactive services.
“The camera’s most important role isn’t taking photos—it’s letting the AI know what you’re dealing with in the moment,” Dong explained. He compared it to the early days of GPS: GPS wasn’t just for navigation; it let computers understand location, which then gave birth to the massive O2O ecosystem of food delivery and ride-hailing.
That’s why Guangfan’s product design puts a bigger emphasis on proactive AI compared to its peers. The system pulls together vision, voice, location, and your calendar info to close the loop for context-aware tasks.
In demos, the device recognized meeting invites and asked if you want to add them to your calendar. It also proactively suggests when to leave based on real-time traffic and your commute habits—and even helps call a ride.
In the team’s vision, when a user stares at a painting in a museum or picks up a product in a supermarket, the AI uses the camera to sync visual info—no need to describe it out loud—and proactively gives background info or compares prices across the web.
Powering all that sensing and action is Guangfan’s self-developed AI OS. Dong, who used to be a key player in Xiaomi’s mobile OS and Quick App ecosystem, knows the importance of system-level orchestration. In his view, a true AI OS must handle four core tasks: multi-device hardware coordination, native app orchestration, edge-cloud computing distribution, and multi-modal interaction support.
In his view, the earbuds, watch, and charging case are just sensing and execution layers. The real magic is how the system orchestrates tasks across these devices. For example, the earbuds handle real-time voice and visual capture, the watch takes care of biometrics and lightweight interactions, and the case provides some computing power and connectivity. Users don’t need to understand how it all works together—the system automatically distributes tasks.
Packing sensors, chips, and other components into such tiny wearables isn’t easy. Dong admitted that over the past six months, the team has spent a lot of effort improving yield rates and tackling the mismatch between software and hardware iteration cycles.
To tackle AI hallucination, the team uses engineered constraints. In critical scenarios like payments, ride-hailing, and scheduling, the system adds confirmation steps. Plus, it combines small models and a rules engine to keep AI reasoning within safe bounds.
Facing the possibility that tech giants might jump into this space, Guangfan Tech strikes a tone of strategic optimism.
Dong believes that wearables are highly personal and decorative, so the market won’t see a winner-take-all scenario. Startups with shorter decision chains and unique insights into hardware-software integration can carve out a differentiated space. He thinks the industry is still in the pre-explosion phase and needs big players to step in and help educate the market together.
Dong predicts the tipping point for AI hardware will come faster than it did for smartphones—he expects an “iPhone 4 moment” within one to two years. At that point, the competition will revolve around the AI OS and its ecosystem moat.
From a commercialization standpoint, first-gen AI hardware is almost always about trial and error. Dong has previously said that he’s not overly focused on sales volume for the first product; he cares more about word of mouth. In his eyes, a startup’s opportunity lies in building user recognition of the brand’s innovative spirit.
For now, the Guangfan AI Full-Sense Wearable still needs more time in the market to prove itself in terms of experience stability, scene coverage, and user retention. Butthe team believes that for a product that has never been clearly defined before, what’s more worth paying attention to is its engineering complexity, product paradigm exploration, and the breakthrough significance of going from concept to mass production.
The difficulty of this task lies first in building from scratch a native AIOS for future interaction, and then integrating hardware modules that lack mature packaging solutions into a compact wearable device. Pushing software and hardware engineering forward in parallel, the difficulty of mass production will increase exponentially. Additionally, for a true AIOS, you also need to build a developer ecosystem, application standards, and a business loop, and define a series of AI-native applications—which also means challenges.
In this phase of rapid AI capability growth, Guangfan has chosen to first build out this generation of AI terminal product form and push it to market for testing. “The ceiling for general-purpose hardware is high enough, but before the real tipping point arrives, you at least have to be at the table,” Dong said.
