--- title: "China's invasive brain chip is about to meet humans — and it is as fast as Neuralink's" date: 2026-09-30 category: Brain–Computer Interface site: NeuroAI canonical: https://neuroai.site/a/na-beinao2-invasive-bci language: en --- # China's invasive brain chip is about to meet humans — and it is as fast as Neuralink's > BeiNao-2 (北脑二号), a 512-channel fully-invasive wireless BCI from the Beijing Brain Institute's spinout NeuCyber (芯智达), is set to start human trials by end-2026 after primate tests hit about 7.2 bits per second. A monkey sits still and, by thought alone, moves a cursor to a red dot on a screen — faster, its handlers say, than a human hand could. The chip in its skull has recorded single neurons for over two years. Now China wants to put that chip in people. The system is **BeiNao-2 (北脑二号)**, and it represents the riskier, higher-performance half of Beijing's two-track brain-computer interface (脑机接口) strategy. ## What BeiNao-2 (北脑二号) is - A **512-channel wireless fully-implantable** engineering prototype. - **Fully invasive (侵入式)**: the electrode is implanted into brain tissue and records single neurons or tiny regions — far finer than the surface-level reading of extradural devices. - Developed by **NeuCyber (芯智达)**, spun out of the Beijing Institute for Brain Science and Brain-Inspired Research (北京脑科学与类脑研究所); the institute's director is Luo Minmin (罗敏敏). - After long animal experiments, the plan is to launch a **researcher-initiated trial (IIT) by the end of 2026**. ## The speed claim The developers report a maximum information rate of about **7.2 bits per second** (Beijing municipal and NCSTI sources; People's health channel cites 7.18 bit/s). They note this is close to the roughly **7.3 bits per second** reported for Neuralink's comparable system — making BeiNao-2, on its own numbers, a rare Chinese invasive BCI that sits next to the U.S. leader on the metric that matters most: how fast intent becomes action. In the primate demo, a monkey controlled a cursor "smoother than human manual operation." The electrode, described as thin as a hair, has recorded five to six hundred neurons at sub-millisecond resolution for more than two years without losing fidelity. ## Why two "BeiNao" at once Beijing runs both routes in parallel, which is unusual globally: - **BeiNao-1 (北脑一号)** is **semi-invasive (半侵入式)**, sitting extradurally — safer, faster to patients, now in confirmatory multicenter trials since March 2026, with registration expected around 2027. - **BeiNao-2** is fully invasive — higher risk, but the resolution to attempt finer motor and speech decoding. The stated logic is pragmatic: ship the safer product to help patients now, while using the invasive one to push the performance boundary. ## What it could restore Because it reads single neurons, BeiNao-2 is aimed at **finer decoding** than extradural systems can manage — tighter motor control and, potentially, more natural speech or cursor work. A disclosed (not yet approved) registration plan attributed to institute advisors targets around **2030** for formal trials, with about **50 cases across 10 hospitals**, including stroke hemiplegia, paraplegia, and ALS patients. We treat that as intent, not an approved protocol. ## The bigger bet: performance versus safety Every invasive BCI faces the same tradeoff. The deeper you implant, the cleaner the signal and the finer the control — but the greater the surgery and the longer the unknown about how tissue reacts over years. BeiNao-2 leans all the way into the high-performance end, while BeiNao-1 stays at the safer edge. That two-track design is the interesting part: it is not a hedge, it is a portfolio. If the extradural route reaches patients first and proves useful, the invasive route still gets to chase the harder decoding problems without holding patients hostage to its risk. ## How the world is reading it Western coverage tends to frame Chinese invasive BCI against Neuralink, and the 7.2-versus-7.3 bit/s figure invites exactly that comparison. But the more useful read is that several countries now have viable invasive programs, and the bottleneck is no longer "can we record neurons" — it is "can we do it safely in humans, for years, at scale, and prove it helps." On that metric, no one has won yet, and BeiNao-2 is still early. ## Honest limitations - BeiNao-2 is **still pre-human** — all confirmed results are from animal experiments; the end-2026 IIT is a plan, not a started trial. - The **7.2 vs 7.3 bits/sec** comparison is the developers' own; there is no independent head-to-head benchmark, and bit-rate alone does not capture clinical usefulness. - The **2030 registration / 50-case** plan is a disclosed intent, not an approved study design. - "Fully invasive" carries **higher surgical risk** than extradural devices, and human long-term biocompatibility is unknown. - All figures come from institute and affiliated-company disclosures reported by state media (NCSTI, Beijing government, People's Daily), not peer-reviewed publications. ## What readers can do now - **If you follow clinical BCI**, watch for the end-2026 IIT announcement from the Beijing Brain Institute / NeuCyber — that is the moment BeiNao-2 leaves the primate lab. - **If you compare BCI routes**, weigh invasive (BeiNao-2, Neuralink) against extradural (BeiNao-1, NEO) on the risk-performance tradeoff; the higher bit-rate is real, the higher risk is real too. - **If you invest**, the nearer-term catalyst is BeiNao-1's expected 2027 registration — BeiNao-2 is a 2026-trial, 2030-registration story, not a near-term product. --- Published by NeuroAI (https://neuroai.site/) — https://neuroai.site/a/na-beinao2-invasive-bci Free to quote with attribution and a link to the original.