The patient had been living with a spinal cord injury for a decade, the result of a car accident that took most of the function in his hand. On a Monday in July, at Huashan Hospital in Shanghai, surgeons slipped a coin-sized wireless chip under his skull and rested it on the dura mater, the tough membrane covering the brain. They did not push a single electrode into brain tissue. When they switched it on, the device pulled clean epidural signals, and an AI decoder began turning his intent to grip into commands for a soft robotic glove.
What made the operation historic was not the surgery. It was the invoice. This was the world’s first commercial implantation of an invasive brain-computer interface — a device sold as a prescribed medical product rather than administered inside a research protocol. Six weeks later, the financial scaffolding around it arrived: on Aug. 12, state-owned PICC Property and Casualty announced China’s first commercial insurance policy covering BCI implantation surgery, underwritten in partnership with a hospital affiliated with Zhejiang University School of Medicine in Hangzhou. Hong Kong-listed BCI stocks rose on the news.
The device and the pipeline
The implant is called NEO, developed by Shanghai-based Neuracle Medical Technology with a team at Tsinghua University’s School of Biomedical Engineering led by Professor Hong Bo. It carries eight recording electrodes and sits epidurally, which spares cortical tissue at the cost of a noisier, lower-bandwidth signal — a deliberate trade that Chinese developers have made repeatedly, and one that separates them sharply from Neuralink’s 1,024-electrode intracortical threads.
China’s National Medical Products Administration approved NEO for commercial sale on March 13, 2026, after clinical trials in 32 patients. Within days, the National Healthcare Security Administration issued the device a medical insurance reimbursement code, closing the gap between market authorisation and payment that typically strands new hardware for years. Production, hospital onboarding and patient screening followed inside four months.
Hong has set a pace to match. He has said the team plans to implant NEO in 30 to 50 spinal cord injury patients over the coming year across roughly 10 centres nationwide. Mao Ying, director of Huashan Hospital and head of the National Centre for Neurological Disorders, who was part of the implantation team, offered a blunt verdict on the early data: “The results are even better than we expected.”
The pricing is the part that should make Western neurotech founders sit up. Beijing’s municipal government has set a standardised price of 7,980 yuan — about $1,200 — for invasive BCI implantation, with partial insurance coverage for urban employees; other provinces have issued guided prices between 6,000 and 6,600 yuan. A Shanghai-backed commercial scheme covers up to 150,000 yuan (roughly $21,450) of total treatment cost. Neuracle filed in June for a STAR Market IPO seeking 2.5 billion yuan, around $370 million.
The rest of the sector is moving in formation. Venture funding for Chinese BCI companies reached roughly 6.47 billion yuan by early July, against 1.5 billion yuan for all of 2025. Shanghai StairMed closed a 500 million yuan round led by Alibaba with Tencent participating, and has implanted a 256-channel wireless system as part of its registration pathway. NeuroXess is building a Jiangxi factory targeting 10,000 brain chip sets a year. In April, the Ministry of Education approved Brain-Computer Science and Technology as a formal undergraduate major, with Tianjin University enrolling its first cohort this autumn. Sixty-two hospitals had opened BCI centres by the end of last year.
Analysis
Strip away the surgery and the hard problem here is a machine learning problem. Eight epidural electrodes produce a signal far weaker and blurrier than what penetrating arrays capture, so nearly all the performance has to come from the decoder — models that map smeared field potentials onto intended movements, and that must hold calibration as electrode impedance drifts, scar tissue forms and the user’s own cortex adapts. Chinese firms are effectively betting that decoding models improve faster than surgical hardware does, and that a cheap, low-risk implant paired with a very good model beats an expensive, high-risk implant paired with a merely good one. That is a defensible bet in 2026, and it is the same bet the rest of the AI industry keeps winning.
The commercialisation model diverges just as sharply. Neuralink runs the American playbook: private capital, a small tightly controlled trial cohort — around 21 Telepathy users worldwide as of January, with patient 26 implanted in May — and reimbursement as a distant problem. China ran approval, insurance coding, provincial price schedules, a state insurer’s surgical policy, an industrial hub, a national degree programme and an IPO through a coordinated pipeline in five months. Paradromics, an Austin competitor, noted in March that the US has one approved product while China has a national industrial strategy and reimbursement infrastructure already forming.
The unresolved piece is neural data. A commercially reimbursed implant generates a continuous stream of cortical recordings flowing to a manufacturer, a hospital and now an insurer with a financial interest in outcomes. China has moved faster on payment codes than on rules governing who may retain, train on or resell that data — and the same asymmetry exists in the US.
What to watch
Three things. Whether Hong’s 30-to-50-patient rollout produces published outcomes rather than press statements, since real-world durability over 12 to 24 months is what insurers will actually price. Whether StairMed’s 256-channel registration trial, targeting roughly 40 implants, clears the NMPA and moves China onto Neuralink’s bandwidth turf. And whether PICC’s pilot policy expands beyond one Hangzhou hospital into standard coverage — the difference between a milestone and a market.
“The results are even better than we expected.”— Mao Ying, Director, Huashan Hospital