Brain implants decode well enough now. The fight is over who pays.

Three things happened to brain-computer interfaces this autumn, and none of them was a better decoder. On September 24, Precision Neuroscience closed a $250 million Series D. On October 2, the American Medical Association published the first CPT codes written for brain-computer interfaces. And on October 28, Shanghai's NeuroXess announced a RMB 700 million Series B. Our thesis: decoding performance is no longer what keeps these devices from patients. Regulatory class, long-term durability evidence and payment now set the pace, and on two of those three China is ahead.

Start with performance, because it is no longer the hard part. Our earlier survey of the field covered the electrode-count race. Since then, Neuralink reports that self-supervised pretraining on more than 50,000 hours of unlabeled neural data kept participants' decoders working for weeks without recalibration, and set an 11.32 bits-per-second cursor record. That is company-reported, not peer reviewed. But it fits the pattern. Cursor control at or above able-bodied mouse speeds, and speech decoding in several labs, are now demonstrations that repeat. What none of the US systems has is a permanent-implant marketing authorization.

Precision is the clearest example of the gap between money and market. Its Layer 7 array received FDA 510(k) clearance in April 2025, but only for implantation of up to 30 days, as a recording and stimulation electrode. The new round brings total funding to $430 million and follows more than 100 patient procedures at 18 institutions, the company says. The New York Times reported a valuation just over $1 billion; the company did not disclose one. The money is for moving from a 30-day Class II component to a permanent, wireless system, which likely means a Class III premarket approval. CEO Michael Mager has said that will take years. Investors are paying for a regulatory step, not a technical one.

The billing codes matter more than they look, and less than headlines will suggest. The AMA's five new Category III codes, 1123T through 1127T, cover implantation of a brain-computer interface electrode array and unit, each additional array, revision, removal and replacement. They take effect January 1, 2027. The descriptor limits them to devices that record brain activity to control external digital interfaces, and excludes deep brain stimulation and seizure monitoring. Two caveats are in the AMA's own text. Category III codes carry no relative value units, so payment is left to each payer, and inclusion "does not imply any health insurance coverage or reimbursement policy." The procedure language also assumes cranial access by twist drill, burr hole or craniotomy, which does not obviously fit an endovascular implant like Synchron's.

Coverage is the harder problem. In April, CMS and FDA announced the RAPID coverage pathway, which would post a proposed national coverage decision the same day an eligible Breakthrough device is authorized, with coverage as soon as two months later. CMS paused its earlier TCET pathway for new candidates while it builds RAPID. The proposed notice, whose comment period closed October 13, requires an IDE study that enrolls Medicare beneficiaries and excludes devices that fall outside a Medicare benefit category. That second condition is the open question for BCIs. An implant whose job is to drive a cursor or a speech synthesizer is not obviously durable medical equipment, a prosthetic or a drug. King & Spalding also notes that the separate add-on payment route for Breakthrough devices was repealed for devices designated after September 30, 2026, so a newly covered implant that cannot show substantial clinical improvement could end up paid only inside the hospital's procedure bundle.

China has already run this sequence end to end. Its regulator approved Neuracle's NEO, an epidural implant that drives a pneumatic glove for people with cervical spinal cord injury, on March 13, 2026. According to BioSpectrum Asia's reconstruction, a national insurance consumable code followed on March 22 and Shanghai procurement listing on March 24. The approval rested on 36 implants with 18 months of follow-up. The pricing groundwork came first: a March 2025 national guideline created separate billing items for implanting, removing and fitting a BCI. The numbers still do not close. Paradromics estimates a full NEO procedure at ¥300,000–500,000 against a government procedure fee of about $900, and notes the device itself was not yet reimbursable. China has solved the coding sequence faster than anyone. It has not yet shown who pays for the hardware.

The Chinese pipeline behind NEO is deep and is now raising money at scale. NeuroXess says it has completed 65 clinical cases, started a registration trial at nearly 20 hospitals in July, and is building a 14,300-square-meter plant sized for 10,000 systems a year. The yuan figure converts to roughly $100 million; reported dollar values differ slightly by exchange rate. StairMed completed enrollment of its multicenter registration trial in August, reports 42 long-term implants, and signed for IPO tutoring on October 30. Its "over 5,000 cumulative implant days" works out to about four months per patient on average. That is real data, but it is not the multi-year durability record a permanent implant ultimately needs.

Durability is the evidence gap on both sides. Precision's clearance is for 30 days. NEO's dossier covers 18 months. Neuralink reported 21 trial participants in January, its last company-confirmed count; higher figures circulating since are unverified. For a device that is meant to stay in the skull for a decade, nobody yet has controlled, multi-year evidence on signal stability, revision rates and explant safety at scale. Claims also outrun data at the frontier. A Colorado team implanted arrays in higher-level cortex this spring, and some coverage called it the world's first cognitive BCI. The university's own account says only "among the first," and frames cognitive and mood treatments as a future goal.

What to watch next is administrative, not neural. Whether CMS finalizes RAPID, and whether it says BCIs have a benefit category. Which US company files the first permanent-implant PMA, and with what follow-up. Whether commercial payers attach rates to 1123T in 2027. Whether China's insurance code turns into actual reimbursement for the device. The race is no longer about who decodes best. It is about who first turns a working implant into something a hospital can buy and an insurer will pay for, and keeps proving it works years after surgery.