Huang Li has spent two decades building infrared night-vision cameras. His newest bet is stranger: a chip that plugs straight into the brain.
The company is Wuhan Zhonghua Brain-Computer Fusion Technology (武汉衷华脑机融合科技发展有限公司), founded in 2021. Huang is better known as the founder of Guide Infrared (高德红外), a major Chinese maker of thermal-imaging gear. What Zhonghua (衷华) showed in 2023 is an implantable brain-computer interface (脑机接口) that the firm claims can both read neural signals and write stimulation back — a two-way link most rivals are still chasing one direction at a time.
What the system actually claims to do
Most consumer and research brain-computer interfaces (脑机接口) only listen. Zhonghua (衷华) says its "植入式脑机接口系统" (implantable BCI system) is different: it both records electrical activity from neurons and sends current back into the tissue. The company calls this 双向读写 — two-way read and write.
The headline number is the channel count. Zhonghua (衷华) says its 微针 (microneedle) electrode array reaches a maximum of 65,536 channels — what it markets as a "6.5万级" (65,000-class) array. For comparison, the firm states this is more than twenty times the channel count of comparable Neuralink hardware. That is a company-made comparison, and it should be read as one (more on that below).
The system also includes:
- A bidirectional neural measurement and modulation chip that both senses and stimulates.
- On-site signal processing, meaning the chip does some computation inside the body rather than streaming everything out.
- A one-time implantation design the company says avoids the repeated insertions other approaches need.
- A stated portfolio of more than 100 patents across the electrode, chip, and system.
The expert sign-off
On 25 August 2023, the system went through a scientific appraisal organized by a third-party evaluation center. The panel was large and credentialed: eleven experts, including Chinese Academy of Sciences academician Zhao Jizong (赵继宗) and Chinese Academy of Engineering academician Li Peigen (李培根). Their conclusion, as reported by the Wuhan municipal government portal, was that the system has completely independent intellectual property and reaches an internationally leading level in the brain-computer interface (脑机接口) field.
That kind of panel verdict matters in China's research ecosystem — it is a formal stamp that the work is real and original. But an appraisal is not the same as a peer-reviewed journal paper or an independent bench test. The distinction is important, and we return to it in the limitations section.
Why two-way matters
A brain-computer interface (脑机接口) that only reads is already useful: a paralyzed person can move a cursor or a robotic hand. But the harder, more useful problems — restoring movement after spinal injury, calming a seizure focus, or nudging a damaged circuit back toward normal — often require sending signals in. That is the promise of 双向 (bidirectional) hardware.
Zhonghua (衷华) lists a long target list of conditions: paralysis from spinal injury, disorders of consciousness, depression, dementia, and even blindness and deafness where the sensory pathway is intact but unused. These are aspirations, not approved treatments. The company says its next step is clinical work with hospitals across roughly thirty categories of neurotechnology medical devices.
Where this fits in China's BCI push
China has been stacking brain-computer interface (脑机接口) programs at the national, city, and company level. Zhonghua (衷华) is one of the more aggressive private entrants, and its 云脑计划 (Cloud Brain Plan) talks about building brain-science centers anchored in Wuhan with links to Beijing and Shenzhen.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, the field is moving past a simple contest of who records the most channels and toward which team can turn raw neural data into something a patient or a device can genuinely use.
That is the real test. A high channel count is a hardware achievement; turning it into a stable, safe, long-term clinical result is a different mountain.
Honest limitations
- Figures are company-disclosed. The 65,536-channel count and the "20+ times Neuralink" comparison come from Zhonghua (衷华) and the municipal government's report of the company's claims, not from an independent publication.
- No peer-reviewed validation yet. The 25 August 2023 appraisal was an expert panel review, not a journal paper with reproducible methods and raw data.
- No published human or animal efficacy data. We have not seen trial registrations, peer-reviewed results, or safety dossiers for the implant in humans.
- The condition list is promotional. Paralysis, depression, dementia, blindness, and deafness are named as targets; none are demonstrated outcomes.
What readers can do now
- Treat channel-count claims with skepticism. Until independent labs or journals reproduce the numbers, read "65,536 channels" as a specification claim, not a proven capability.
- Watch for a clinical registration. A real next step would be a registered trial (for example on ClinicalTrials.gov or China's trial registry) — that is the signal to take the system seriously.
- Follow the appraisal panel, not the press release. Academicians Zhao Jizong and Li Peigen lend credibility, but their review is a starting point, not proof of bedside readiness.