Brain-computer interfaces: where the field actually stands

Brain-computer interfaces carry four metrics and 15 milestones here, and this is a field where the absolute numbers are still tiny and the direction is unmistakable.

On implanted humans, Synchron leads with 10 recipients, Neuralink has 5, and Paradromics has 1 — a first human recording during a 20-minute procedure in 2025. That is roughly sixteen people worldwide carrying a company's implant. Electrode count, the rough ceiling on how much neural signal an implant can read, tops out at 1,024: Neuralink's N1 across 64 threads and Precision Neuroscience's Layer 7, which took 510(k) clearance in April 2025. Paradromics' Connexus carries 421, and Synchron's vascular Stentrode just 16 — a reminder that channel count trades against invasiveness. The original BrainGate Utah array had 96. Disclosed private funding puts Neuralink at $1.3B on a reported $9B valuation, Synchron at $330M and Paradromics at $110M. Research output runs about 4,756 papers a year.

2026 was the year the field produced products rather than demonstrations, and neither of the first two came from the best-known company.

In March, China's NMPA cleared Neuracle's coin-sized NEO implant, developed with Tsinghua, for commercial sale to restore hand function after spinal-cord injury — the first invasive-class BCI approved for market anywhere. In July it was implanted in a paying patient at a Shanghai hospital, a man who lost hand function in a car crash a decade earlier: the first commercially approved BCI implanted in anyone, four months after clearance.

Also in July, Science Corporation began the European commercial launch of PRIMA under a CE mark valid across 30 countries — the first BCI cleared anywhere for form-vision restoration. Its subretinal photovoltaic implant paired with camera glasses restored functional central vision in geographic-atrophy patients; 80% read letters, numbers and words in the NEJM-published trial. Science had closed an oversubscribed $230M Series C in March to commercialize it.

Other 2026 firsts pushed on capability rather than market. Surgeons at CU Anschutz implanted a BCI into a paralyzed patient's higher-functioning cortex rather than the usual motor cortex, a world first aimed at more natural sensory-and-motor control. In August the FDA cleared Paradromics to widen its Connect-One feasibility study so participants can drive ordinary phones, tablets and computers through its Convey interface — a protocol expansion, not a marketing approval, but a meaningful shift from cursor demos toward the devices patients actually use. China launched its first multicenter trial of a domestically built 128-channel fully implantable BCI in May.

The deeper foundation was laid earlier: UC Davis and BrainGate2 demonstrated a speech neuroprosthesis in 2024 decoding attempted speech at up to 97–99% accuracy from a large vocabulary, since used independently at home for thousands of hours.

The locked milestone here is a full FDA-approved commercial BCI in the US, placed in the 2030s. Note what that means: the first commercial approvals happened, but not in the United States.