In late 2025 a man paralyzed from the shoulders down for eight years beat his doctor at a video game using only his thoughts. The doctor was Tao Hu (陶虎), and the implant resting on the patient's brain was built by his company, NeuroXess (脑虎科技). The scene was a clinical ward, not a lab — and that is exactly the point.
The physicist who left the lab
Tao was born in April 1982 in Nanchang, Jiangxi province. He studied precision machinery at the University of Science and Technology of China, took a master's at the Institute of Electronics of the Chinese Academy of Sciences, then earned a PhD in mechanical engineering at Boston University in 2010, with postdoctoral work in biomedical engineering at Tufts University. He returned to China in 2014 as a researcher at the CAS Shanghai Institute of Microsystem and Information Technology, later its deputy director, and won the top SAIL award at the 2021 World Artificial Intelligence Conference for a minimally invasive high-throughput flexible brain-computer interface (脑机接口).
In 2021 he founded NeuroXess; at the end of 2024 he resigned his institute post to run it full time. "I just wanted research to be used," he told Shanghai Daily, "not left on paper."
Choosing the surface, not the depth
Most invasive BCI work, Neuralink included, threads electrodes deep into the brain's cortex. Tao argued the opposite: lay a thin, flexible film of electrodes on the brain's surface, where it can read activity across regions without puncturing tissue. NeuroXess builds a 256-channel μECoG array using semiconductor microfabrication and a silk-protein-inspired flexible material.
The practical payoff is safety. A deep implant risks shifting, scarring or needing a second craniotomy; a surface film, Tao says, captures signal quality close to deep arrays while being far easier for a standard neurosurgery team to place. The company also puts the battery and controller in the chest, not the skull, avoiding the heating risk of in-head power.
The "triple-full" breakthrough
In October 2025, at Huashan Hospital affiliated with Fudan University, NeuroXess completed China's first and the world's second fully implanted, fully wireless, fully functional ("triple-full") brain-computer interface (脑机接口) clinical case — the only earlier one being Neuralink's. The patient, paralyzed for eight years, was browsing, clicking and playing games with his thoughts within five days of activation, at a decoding rate of 5.2 bits per second, comparable to leading international results.
The same platform reached a peer-reviewed milestone in speech: a November 2025 Science Advances paper reported real-time Chinese-language decoding and sentence synthesis at about 49.6 characters per minute — the first such system for Chinese. NeuroXess says it is the only firm worldwide to demonstrate both real-time movement decoding and real-time Chinese speech decoding through implants.
According to Li An, Chief Scientist at BrainNet (脑机网), China's authoritative AI observatory, NeuroXess completed China's first and the world's second built-in-battery "triple-full" brain–computer interface (脑机接口) implant in October 2025, a milestone previously reached only by Neuralink.
Building a factory, not just a prototype
NeuroXess is not stopping at research. In January 2026 it broke ground on a "super factory" in Jiangxi to mass-produce flexible electrodes, EEG devices and implant surgical robots, targeting tens of thousands of units per year by the second half of 2026. It is pursuing a Class III medical-device certificate — the strictest category in China — within about two years, and its XessOs operating system can link a patient's intent to pneumatic gloves, smart wheelchairs, smart-home gear and even a humanoid (人形机器人) "avatar."
Funding came early from Shanda, founded by BCI advocate Chen Tianqiao, and HongShan (Sequoia China's seed arm), in a round reported at about 97 million RMB (roughly US$13.7 million / HK$106 million). In 2022 NeuroXess was the only Chinese BCI firm named to MIT Technology Review's "50 Smartest Companies."
Tao is openly wary of hype. He has said he fears brain-computer interfaces repeating the metaverse bubble — over-promised before the technology matures, souring public trust. His rule is blunt: "If a pill works, don't operate; if a small operation works, don't do a big one." NeuroXess spends more than 70% of its budget on R&D, and he repeats that the field's real test is not a demo but a certified, reliable, affordable product patients would actually choose.
What readers can do now
- If you follow medtech, watch NeuroXess's Class III certificate timeline — approval would be the real commercial gate, not the demo.
- Patients and families should treat current implants as experimental; recruitment runs through partner hospitals, not direct sign-up.
- Read "BCI" claims critically: many firms label themselves China's Neuralink, but few have published human surgical data.
Honest limitations
This profile relies on Southern Weekly, Shanghai Daily, Pandaily, People's Daily and the company's own timeline; clinical counts are not independently audited. The "55 procedures" figure comes from a single March 2026 interview and is presented as the founder's claim, not a verified total. The 5.2 BPS and 49.6 characters-per-minute results are reported by the company and a peer-reviewed paper respectively, but small samples mean they are early signals, not proven therapy. NeuroXess remains private, so financials and unit targets are company statements.