What the Implants Have Shown

A brain-computer interface, in the form now being tested in people, is a device that records the electrical activity of neurons and translates it into a command a computer can act on — a cursor moved, a word selected, a sentence spoken through a synthesiser. The premise is old; the demonstrations are new. Over the past two years a small number of people with paralysis have used implanted systems to do things they could not otherwise do, and the recordings of them doing so are real. The task is to hold that fact at its true weight — neither less nor more.

WHAT IS ACTUALLY DEMONSTRATED

The verifiable results are narrow and significant. A participant in Neuralink’s first human study, paralysed from the shoulders down, has used an implant to control a computer — playing games, browsing, sending messages — for over a year, after an early problem with the electrode threads was managed in software. A man with ALS used the same system to edit and narrate a video. Synchron, which threads a smaller electrode array into a blood vessel against the motor cortex rather than opening the skull, reports having implanted its device in dozens of patients with no serious adverse events recorded. In late 2025 the U.S. Food and Drug Administration cleared a third company, Paradromics, to begin a clinical study of a higher-bandwidth device for restoring speech. These are restorations of communication and control for people who had lost them. That is the achievement, and it is genuine.

The honest unit of measurement is restored agency for a handful of people, not symbiosis with a machine.

WHAT SMALL NUMBERS MEAN

What the evidence is not, yet, is what the valuations imply. Morgan Stanley has put a figure of around four hundred billion dollars on the eventual market. The clinical record supporting it consists of early-phase studies designed to test whether the devices are safe and whether they work at all — not large, controlled trials measuring how well, for whom, and for how long. Implant counts are in the dozens. Follow-up is measured in months to a few years. The thread-retraction issue in the first Neuralink device is a reminder that durability inside living tissue remains an unsolved engineering problem, not a footnote. None of this diminishes the results; it places them. Feasibility is the beginning of medical evidence, not the end.

THE REGISTER PROBLEM

The field is also fighting a framing imported from its loudest backers — the idea that the destination is a merger of human and machine cognition, a general-purpose neural interface for the able-bodied. That framing distorts how the evidence should be read. The trials underway are not about enhancement; they concern people with severe spinal-cord injury, ALS, and stroke regaining the ability to communicate. Judged as enhancement, the technology looks speculative and slightly alarming. Judged as assistive medicine, it looks like one of the more consequential developments in neurology in a generation. The same data supports a sober reading and an inflated one, and only the sober reading is currently earned.

THE HONEST TIMELINE

Realistic paths to commercial approval, on the companies’ own accounting, extend toward the end of the decade, with initial use limited to those who have no other means of communication. Cost and insurance coverage are unsettled. The international expansion now underway — Neuralink trials have opened in the United Kingdom, Canada and Abu Dhabi — broadens the safety data but does not shorten the road to proof of benefit. The work is moving quickly by the standards of implanted medical devices, which is to say slowly by the standards of the announcements around it. A reader who keeps two questions apart will understand this technology better than the valuation does: can it give speech back to someone who has lost it — increasingly, on current evidence, yes; and will it become a routine interface between healthy brains and computers within a decade — nothing in the trials says so.