China's Race to Commercial Brain Chips: From Paralysis to Mind-Controlled Machines

This story opens with Mr. Zhang: an accident more than eight years ago left him paralyzed from the neck down, yet today he steers his wheelchair with his mind and even shows off a few tricks. Last year surgeons implanted a brain-computer interface built by the Shanghai startup NeuroXess. Zhang is a pioneer of China's new wave of clinical trials, and these trials aim to shorten the road to a commercial product.

So how does the technology work? Every movement we make is governed by electrical signals between neurons in the brain; interfaces read those signals and translate them into commands that drive an external device. The range is wide: some systems need implants that penetrate brain tissue, some sit on top of it, and some merely touch the scalp. Because brain data is noisy, companies lean on artificial intelligence algorithms that filter out background chatter. The concept was born in the 1970s, when a study showed EEG signals could be mapped onto movement, and human trials followed in the early 2000s. That convergence is now accelerating with generative models, and according to Bciintel, neural decoding is merging with language models to produce fluent language from faint signals.

From money to device: funding and NeuroXess

The direction of money is clear: brain-chip companies in the United States have collectively raised about 2.75 billion dollars , while Chinese and European rivals managed a fraction of that. Even so, the Chinese front is accelerating: funding for brain-chip startups in the country doubled in 2025 from the year before. These funding figures come from TheInsight, drawing on the April report by the Fourth Wave Technology Think Tank; the report counts 80 Chinese companies raising 3.8 billion yuan, roughly 560 million dollars, across 17 deals in the first quarter of 2026. Bciintel data completes the global picture: Neuralink closed a 650-million-dollar Series E in June 2025, Synchron raised 200 million dollars and OpenAI-backed Merge Labs took 252 million dollars.

NeuroXess bets on a less invasive design: a strip of electrodes resting on the cortex collects signals, a connected brain chip processes them, and an internal control unit transmits the data wirelessly to a computer, with the battery in the same package. The company has tested different versions in more than 50 patients; Zhang, implanted since October 2025, is the longest-running case and trains several hours a day on new skills. This technical profile comes from NeuroXess, based on the company's 256-channel flexible-interface trials with Huashan Hospital; the team started decoding within two days and demonstrated mind-controlled devices at under 60 milliseconds of latency. Zhang is also learning to run smart-home devices and a hand exoskeleton with his thoughts under a Xiaomi collaboration; according to the company, this path could one day help him walk again.

A correction on approval, and the business model

Every object placed in the brain carries risk: over time scar tissue can form around the implant, damaging cells and muffling the signal. NeuroXess chose the non-penetrating strip form to reduce scarring. The regulatory bar is high, with this class of product assessed in the toughest medical-device category. This policy framework comes from Georgetown, based on the December 2025 implementation opinions of seven ministries, which count interfaces among the new high-quality productive forces. The state support described on screen is generous too: the 15th Five-Year Plan named interfaces one of six strategic industries of the future, clinical-trial reviews were accelerated, and a 165-million-dollar brain-science fund was opened.

An important correction is needed here: the video correctly says China granted the first invasive interface a commercial approval in March 2026, but the license went not to NeuroXess but to Shanghai-based Neuracle Technology and its NEO hand-movement device. The coin-sized unit sits outside the brain's membrane, decodes signals in real time and triggers a pneumatic glove for grasping, picking up objects and drinking; the product, aimed at spinal-cord-injury patients aged 18 to 60, was tested in 36 clinical procedures. This approval record comes from TheInsight, which reports the company has started clinical training at top-tier hospitals. The detail proves the race runs on real licenses and hospital infrastructure, not laboratory shows alone.

The biggest test is commercial: will these devices ever become a viable business? Payment is the puzzle; what insurance coverage will look like and whether it covers costs is unknown. According to Bloomberg's July 2025 reporting, Neuralink expects about 50,000 dollars of revenue per surgery; NeuroXess disclosed no price but broke ground on a factory it says will build 10,000 devices a year. This market picture comes from Bciintel, resting on a 3.2-billion-dollar market in 2026 and projections of 6 to 12 billion dollars by 2030. The real investor question is market size: as a purely medical device the returns may stay limited, while a broader release would change the math.

Privacy, the military and the merge

Mass adoption, whether through surgical or non-surgical systems, magnifies the question of neural privacy : who gets access to this data and what happens to it? In response to the acceleration, some jurisdictions are passing laws to protect neural data. This legal landscape comes from Mondaq, based on the first neural-data laws Colorado and California enacted in 2024; in 2025 US senators urged the trade commission to act against the sale of such data. The nightmare scenarios have some basis in reality: the CIA ran mind-control experiments during the Cold War, DARPA funded mind-driven drone-control research, and China funds military brain research too. Elon Musk's claim of offsetting artificial-intelligence risk aside, the host's thesis is clear: first the technology must reach the market, then it may move beyond medicine and weave humans and machines ever tighter.