Paradromics put a 420-electrode brain chip smaller than a dime into a human and pulled it out in 20 minutes. Now it wants to restore speech.

A brain-computer interface smaller than a dime, carrying 420 microelectrodes, was implanted into a human brain on May 14, and removed 20 minutes later. That wasn't a malfunction. That was the plan, and Paradromics says it worked exactly as intended: the device recorded neural activity at the level of individual neurons, proved it was safe in human tissue, and set the Austin-based startup on course for the most ambitious goal in the BCI race, restoring speech in patients with paralysis.

The procedure took place at the University of Michigan, led by neurosurgeon Dr. Matthew Willsey and Dr. Oren Sagher during a patient's scheduled epilepsy resection surgery. Piggybacking on an existing operation was deliberate, it gave the team a narrow window to demonstrate human compatibility without requiring a standalone surgical intervention. The FDA has since granted Paradromics Investigational Device Exemption status for its Connect-One Early Feasibility Study, the first approved clinical trial specifically designed to test whether a fully implantable BCI can convert neural signals into text or synthesized speech. No other company in the space has cleared that specific hurdle yet.

"We've shown in sheep that our device is best in class from a data and longevity standpoint, and now we've also shown that it's compatible with humans." - Matt Angle, CEO and founder, Paradromics

The speech restoration target separates Paradromics from its closest rivals in a meaningful way. Neuralink's first patient, Noland Arbaugh, received his implant in January 2024 and has since been using it for up to eight hours a day to control a cursor and play chess, impressive motor control applications, but not communication. Synchron, backed by Jeff Bezos and Bill Gates, has implanted its stent-based device in multiple patients and demonstrated typing via thought. Precision Neuroscience has placed its thin-film array on the brain surface of surgical patients. Each approach captures neural signals differently, but none has an FDA-cleared trial specifically targeting speech output from a fully implantable system.

The technical distinction behind Paradromics' approach matters here. Its Connexus device embeds microelectrodes directly into brain tissue, recording from individual neurons rather than averaging signals across larger regions. Angle described the difference to CNBC as placing microphones inside a stadium versus outside one, inside, you catch individual conversations; outside, you only get the roar of the crowd. For speech restoration, that resolution is likely critical: reconstructing the neural patterns that produce specific phonemes and words requires single-neuron precision, not population-level approximations.

Founded in 2015, Paradromics has spent a decade getting here. The company has raised nearly $100 million as of early 2025, according to PitchBook, and announced a strategic partnership with Saudi Arabia's Neom in February, though the investment amount was not disclosed. Compared to Neuralink, which has raised over $700 million and benefits from Elon Musk's public profile, Paradromics has operated quietly, spending nearly three years on preclinical sheep studies before touching a human brain. The FDA exemption suggests that methodical approach paid off.

What the Connect-One trial will actually test is whether the Connexus system can sit permanently inside a patient's skull, not for 20 minutes, but for months or years, and reliably decode the neural signals that the brain generates when a person tries to speak. The target population is people with ALS, stroke, or spinal cord injuries who have lost the ability to produce speech but retain the neural intention to do so. For that group, a working speech BCI wouldn't be a convenience. It would be the difference between isolation and communication.

Neuralink's second patient implant, announced in late 2024, encountered hardware complications that required software workarounds, a reminder that durability inside a living brain remains the hardest unsolved problem in the field. Paradromics' next 12 months will show whether its decade of preclinical caution translates into the kind of long-term stability that actually gets a BCI from a clinical trial into a medical device.