A paralympic athlete raises an arm that ends in a robotic hand and closes its fingers around a torch. No surgery, no implant — just sensors reading the faint electrical chatter of the nerves that remain. The crowd never sees the harder engineering problem behind the moment: making a whisper-quiet signal reliable enough to trust with a life.
BrainCo (强脑科技) has spent a decade arguing that the future of brain-computer interface (脑机接口) does not have to pass through an operating room. Its flagship product is a bionic hand that decodes intention from outside the body.
The bet against the craniotomy
Most of the attention in brain interface work goes to invasive implants — electrodes pushed into or onto brain tissue for the sharpest possible signal. BrainCo went the other way. Founded in 2015 by Bicheng Han (韩璧丞) at the Harvard Innovation Labs, the company is headquartered in Hangzhou with a U.S. office in Somerville, Massachusetts, and is counted among the city's "Six Little Dragons (杭州六小龙)" alongside DeepSeek and Unitree (宇树科技).
Han has said his decision to avoid invasive routes came after watching more than twenty craniotomy surgeries and concluding that the risk, infection exposure, and cost would keep such devices from ever reaching ordinary users. The non-invasive (非侵入式) path trades signal sharpness for reach: a device anyone can put on.
What the hand actually does
The company's BrainRobotics bionic hand reads the electrical activity of the residual limb's nerves and muscles (EMG signals) and turns intended movement into finger motion. By the company's specifications it enables independent movement of all five fingers, reacts in roughly 0.8 seconds, and reaches a precision of 0.1 mm — fine enough to pick up an egg without crushing it. It received U.S. FDA 510(k) clearance (number K220002) in September 2022.
The signal-capture side rests on a "solid gel electrode (固态凝胶电极)" the company developed to replace the conductive paste used in conventional EEG caps. According to a Zhejiang provincial government profile, the material lifts signal-collection accuracy to medical-grade levels above 95% without paste, removing a major source of discomfort and inconsistency.
Scale is the part competitors still lack
Where BrainCo separates from most brain-interface ventures is manufacturing, not science fiction. In 2022 the company reported single-product mass production of 100,000 units of a high-precision brain-interface device — a figure repeated in official Chinese provincial and municipal coverage. That is a different category of problem solved: consistent yield, supply chains, and a price point well below Western equivalents (the company cites roughly one-fifth to one-seventh of comparable overseas prosthetics).
The hands have shown up in visible places. At the 2023 Asian Para Games opening ceremony in Hangzhou, an athlete with an upper-arm amputation used a BrainCo hand to grip and light the torch. The product was named to Time's list of best inventions in 2019.
A financing round that signals intent
In January 2026, Bloomberg reported that BrainCo had confidentially filed for a Hong Kong IPO, with China International Capital Corporation and UBS as joint sponsors. Around the same time, the company confirmed a new financing round of about RMB 2 billion (≈ US$280M / HK$15.6B), led by IDG Capital and Walden International (华登国际, the firm founded by former Intel CEO Lip-Bu Tan), with strategic participation from hardware manufacturers including Lens Technology (蓝思科技) and OmniVision (豪威集团).
That scale of capital matters less for the lab than for the factory floor. BrainCo's stated plan is to push core technology R&D, product scaling, and mass production — the unglamorous work that decides whether a neurotech device becomes a prescription or a prototype.
Beyond the hand
BrainCo's portfolio spreads the same non-invasive signal-reading core across several products: an attention-training system deployed in Chinese schools, a sleep device (Easleep / 深海豚) that pairs EEG detection with light and sound intervention, an autism-spectrum social-training system (开星果), and a bionic leg. The through-line is consistent — read weak biosignals cleanly, decode a state or intent, act on it — applied to rehabilitation, education, and wellness rather than to implanted computing.
Honest limitations
FDA 510(k) clearance covers the device class and its substantial-equivalence review; it is not an endorsement that the hand restores native function or a cure for amputation. Non-invasive signal reading is inherently lower-resolution than invasive recording, and decoding accuracy depends heavily on fit, skin contact, and user training. The "100,000-unit" figure refers to a consumer-style device production run, not to clinical brain-interface implants for paralysis. The RMB 2 billion round and confidential Hong Kong IPO filing were reported by Bloomberg and echoed in Hong Kong and Singapore media; as of the reporting, no hearing, pricing, or listing had occurred. The price comparison to overseas products is the company's and is not tied to a specific named competitor in the sources.
What readers can do now
- If you or a family member needs a prosthetic, ask specifically whether the device reads residual nerve or muscle signals — that distinction separates an active, intent-driven hand from a passive cosmetic one.
- If you buy neurotech consumer gear, look for medical-grade signal claims backed by a filing number (FDA 510(k), NMPA, or equivalent), not vague "brain-training" language.
- If you track the industry, watch mass-production milestones — a hundred thousand shipped units is a stronger signal of real adoption than any demo video.