Matthias Puls took over as CereGate's chief executive just over a year ago. The Munich company had spent six years building a therapy that adds no new implant to a Parkinson's patient's head. The hardware is already there: a Boston Scientific deep brain stimulation generator seated in the subthalamic nucleus. CereGate's software adds a second programme to the clinical workflow, bursts of stimulation timed as a walking cue. The patient switches it on when a step will not start.

On 14 October, CereGate said it had closed an oversubscribed Series C. The amount was not disclosed. Neither was the valuation, the ownership split, or the name of the lead investor, described only as "a global US-based medtech company." Impact Ventures of Budapest joined as a new backer. Existing investors followed: 468 Capital, re.Mind Capital, Heal Capital, HTGF, TruVenturo.

The missing number is the interesting part. A strategic medtech lead financing a software layer that runs on implants already in patients is making a distribution bet. If that lead builds deep brain stimulation systems itself, the round ties a hardware maker to a therapy that expands what its installed base can do. Inside BCI raised exactly that question and did not get an answer.

Puls framed the money around clinical work. The round, he said, lets CereGate "take our therapy through the clinical work that matters most, in an indication where few options remain once medication and standard stimulation stop working."

The cue, not the implant

Freezing of gait is a specific failure. A Parkinson's patient is walking, then stops, unable to restart despite intending to move. Deep brain stimulation can control the symptoms that respond to dopamine and still leave this gap. CereGate's own release says freezing responds poorly to established forms of stimulation. That admission is the whole business case, because it means the installed base of implants is not finished working.

CereGate describes the cue as the same idea as the rhythmic sounds and visual markers used in Parkinson's rehabilitation, moved inside the body. Its claim is that the signal can travel through the implant instead. A timed bursting pattern is a different intervention, delivered by the same generator.

The commercial logic follows the clinical one. Boston Scientific has already sold the generators. A software programme that improves walking on those generators reaches patients without another surgery, another hardware approval cycle, or a new implant to manufacture. The constraint sits in the clinic, not the factory: someone has to tune the pattern for each patient.

A programme the clinician tunes

The trial registry is unusually precise about what the software does. CereGate's system has one job: guide a clinician through the stimulation parameter space, propose settings, and document each patient's response. The clinician decides what to test.

It does not send commands to, nor control the output of, the DBS System.— ClinicalTrials.gov, NCT05292794

That sentence changes how the product should be read. CereGate sells a configuration and documentation layer. The stimulation still comes from Boston Scientific hardware, through two programmers named in the protocol. The protocol pairs conventional tonic stimulation of the subthalamic nucleus with additional bursting stimulation, delivered as a cue. CereGate calls this a computer-brain interface because information travels into the nervous system. Most BCI companies work in the other direction, decoding neural activity into control signals for a cursor or a robotic arm.

Encoding has a lower technical bar in one respect. The nervous system already knows how to interpret sensory input. The 2024 spinal cord work showed patients could learn artificial patterns in minutes. The harder question is whether a cue delivered through a movement-disorder implant improves walking outside a laboratory.

Forty-one patients, no control arm

The US study, FOG-LIGHTS, plans to enrol up to 41 people across up to eight sites. Seven are listed, including Northwestern University, Vanderbilt University Medical Center, the University of Miami and Kaiser Permanente's movement disorders programme in Redwood City. Every participant receives the therapy and knows it. There is no randomisation and no comparison group.

41 patients planned

FOG-LIGHTS enrolment target

Up to 41 participants at up to eight US sites, single arm and open label · ClinicalTrials.gov, 2026

Participants keep their conventional stimulation and use the added cue on demand for about 60 days, with follow-up visits at day 60 and day 61. The primary endpoint is the mean percent change in excess arrhythmicity on a figure-of-eight course with turns and barriers, measured with an Opal wearable sensor while patients are on medication and on stimulation. Eligibility requires less than 8% arrhythmicity before therapy, which keeps the sample inside a narrow gait profile.

The endpoint deserves a note. Excess arrhythmicity compares a patient's gait rhythm against age-matched controls, measured on a course that forces turns and obstacles. It is an objective number produced by wearable sensors, which limits some of the optimism bias that patient-reported walking scores invite. It does not remove the bias that comes from every participant knowing the device is on.

The study began in April 2022. Primary completion is estimated for June 2027, with the full study closing a month later. The current evidence is a protocol. Results arrive in 2027.

Eighteen of twenty, on a different implant

The peer-reviewed support comes from spinal cord stimulation, not deep brain stimulation. In a 2024 paper in Brain Stimulation, CereGate researchers worked with 20 patients who had received spinal cord implants for chronic neuropathic pain. They established a stable computer-brain interface in 18 of them.

Those 18 learned to read rhythmic patterns (13 participants), decode Morse code (3), and hold a balance-board position using only spinal stimulation as feedback (18, plus five with a sensor cap). Median calibration took 79 minutes. Median training for a task took under two minutes. Average task performance was 88%, with wide individual spread.

The authors disclosed their interest: several are CereGate employees, others sit on its advisory board. The sample was small, the tasks lacked control conditions, and the implants were in the spine. The DBS trial is testing whether the same logic holds in the brain. That transfer is plausible and unproven.

The research lineage is Budapest, not a California incubator. The 2024 study drew patients from the National Institute of Mental Health, Neurology and Neurosurgery in Budapest. CereGate's founder, Bálint Várkuti, is an author, and the company has kept the academic ties close enough that board members and employees appear on the byline. That is normal for a clinical-stage spinout. It also means the published evidence was produced by people with a stake in the result.


CHRONICLE: CereGate
────────────────────────────────────────
2019   ·  🏭  Founded in Munich
2020   ·  💰  Series A follows early proof-of-concept work
2024   ·  🧪  Brain Stimulation: 18 of 20 SCS patients get an interface
◉ NOW  ·  🧠  Series C; FOG-LIGHTS enrolling at seven US sites
🔥 NEXT ·  📊  Primary completion due June 2027

ClinicalTrials.gov NCT05292794; CereGate release, 14 October 2026.

Who pays for the installed base

CereGate describes its platform as hardware-agnostic and says it has partnerships with device makers it will not name. For now, the US trial accepts only two Boston Scientific pulse generators, Gevia and Genus R16. CereGate says the platform has been tested in more than 85 people. The public record supports that only in aggregate: 20 in the spinal cord study, an earlier proof-of-concept cohort, and an unspecified number in the current DBS programme. Aggregates are not cohorts.

The strategic logic is circular in a useful way. A device maker that leads the round gets a software upgrade path for generators it has already sold. CereGate gets a channel into movement-disorder clinics without building a sales force for hardware it does not make. The loser, if the approach works, is the assumption that better stimulation requires a new implant.

💡
What is still undisclosed
The amount and valuation. The lead investor's identity. Whether the unnamed hardware partners have signed anything binding. And the FDA Breakthrough Device status: HTGF announced the designation in January 2022, while CereGate's own materials describe it as pursued.

What would change the verdict

Four things would move this from a well-financed programme to a validated therapy. The identity of the lead investor, because it determines whether the round is a commercial channel or a financial punt. Publication of the FOG-LIGHTS primary endpoint, because an open-label gait measure with no control arm is vulnerable to expectancy. A clear regulatory route and submission date. Evidence that the cue generalises beyond one manufacturer's generators.

The FDA Breakthrough Device question is the easiest to resolve and the most telling. A designation announced by an investor in 2022 should not still be ambiguous in 2026.

As we wrote in October, mapping a fruit fly's wiring diagram became a machine-learning benchmark almost overnight. That work reads the brain. CereGate is testing the reverse direction, writing a usable signal into a system that is already wired. Both directions need the same thing before they become medicine: a result that survives a control group.

The Series C buys time. The trial buys evidence, and not until 2027. A cue that helps someone restart walking is a real clinical outcome. An oversubscribed round is a real financial one. They are not the same claim.

The FOG-LIGHTS protocol as registered
Up to 41 participants, up to eight US sites, one open-label arm, and a gait endpoint measured on a turning course. The registry also states plainly that the software does not control the implant.
The primary record for the US trial, including eligibility, endpoint definition and the June 2027 completion date.
Converting a spinal cord implant into a sensory channel
Twenty patients, eighteen stable interfaces, and tasks built on rhythm, Morse code and body posture. The proof of concept behind the company's claims, published in January 2024.
Peer-reviewed and open access, with a competing-interest statement that lists CereGate staff and advisers among the authors.
Seven US sites, one open-label arm
An independent read on the Series C, the trial's geography, the two supported Boston Scientific generators, and the unresolved Breakthrough Device status.
The most precise public description of what the software does not do, published the day after the funding announcement.