Wearable devices have steadily added more sensors and features, but most still depend on a nearby smartphone or cloud servers to handle intensive processing. That reliance could start to change following new research from Tsinghua University and Peking University, where engineers have developed a flexible AI chip thin enough to rival a human hair and resilient enough to survive repeated bending.
The chip, called FLEXI, is designed to run artificial intelligence workloads directly on the wearable itself. Instead of sending raw data to a phone or remote servers, FLEXI processes information locally, reducing latency, energy use, and dependence on external devices. For wearables that need to operate continuously or discreetly, such as health monitors or smart patches, that shift could be significant.
What sets FLEXI apart is as much about its physical construction as its computing capability. According to the research published in Nature, the chip is fabricated on a flexible plastic substrate using low-temperature polycrystalline silicon circuits. This approach allows the entire system to bend, twist, or fold without breaking the underlying electronics. Unlike rigid silicon chips, FLEXI can conform closely to the contours of the human body, making it suitable for skin-mounted or textile-integrated devices.
Durability testing suggests the design is more than theoretical. The researchers subjected the chip to more than 40,000 bending cycles and folded it to a radius of just one millimeter, all without measurable loss in performance. That level of mechanical resilience addresses one of the main obstacles that has limited flexible electronics in the past.
Performance results are also notable given the chip’s size and power constraints. In health monitoring tests, FLEXI identified irregular heart rhythms with 99.2 percent accuracy and classified everyday activities such as walking and cycling with 97.4 percent accuracy. Just as important, it achieved these results while consuming less than one percent of the energy used by conventional rigid chips performing similar tasks. The research team estimates that, at scale, each chip could cost under one dollar to produce.
The implications extend beyond medical wearables. A flexible, low-power AI chip could enable earbuds that process voice commands locally, smart clothing that interprets movement without a phone, or lightweight augmented reality glasses that don’t rely on tethered devices. More broadly, it fits into a growing push across the electronics industry to move intelligence closer to the edge, where data is generated.
The researchers say their next goal is to integrate additional sensors and expand the chip’s capabilities, gradually increasing complexity while preserving flexibility and efficiency. While FLEXI is still a research prototype, it highlights how advances in materials and chip design could loosen the long-standing link between wearables and smartphones, opening the door to devices that are smaller, more autonomous, and less dependent on the cloud.
