NeuroBridge Raises $420M to Bring Neural-Visual Implants to Market by 2029

A New Challenger in the Neuro-Prosthetics Race

Two years ago, five engineers walked out of Neuralink's San Francisco office with a shared frustration: the company's all-brain approach to neural interfaces was missing an entire sensory pathway. Today, NeuroBridge is challenging that philosophy head-on with a $420 million Series B raise led by Fidelity Ventures, with participation from Founders Fund and a consortium of medical device investors.

The Optic Nerve Bridge

NeuroBridge's flagship product, the OB-1, doesn't sit in the brain. Instead, it threads a 0.3mm flexible electrode array through the vitreous humor of the eye, docking directly onto the retinal ganglion cell layer. A wireless inductive link connects to a palm-sized processing unit that runs NeuroBridge's proprietary vision reconstruction AI — a spiking neural network trained on 14 million hours of first-person visual experience footage.

The key innovation is temporal coding. Previous retinal prosthetics like Second Sight's Orion attempted to restore vision through crude spatial stimulation patterns. OB-1 instead uses predictive temporal coding: the AI model anticipates visual flow and delivers stimulation patterns in advance, reducing the perceived lag from 340ms to under 40ms.

"We stopped trying to tell the brain what the eye sees," said Dr. Yuna Sato, NeuroBridge's co-founder and CTO. "We started asking the brain what it expects to see next, and we give it that."

Clinical Data: 23 Patients, 18 Months

The company has implanted OB-1 in 23 patients across three clinical sites — UCSF, Moorfields Eye Hospital in London, and Seoul National University Bundang Hospital. All patients had complete or near-complete vision loss due to optic neuritis or diabetic retinopathy, conditions where the optic nerve remains structurally intact but photoreceptor function is destroyed.

Results exceeded expectations. Mean visual acuity improved from light perception only to 20/400 — still legally blind, but sufficient for independent navigation. Critically, 20 of 23 patients reported that objects appeared "three-dimensional" rather than the flat phosphene patterns of previous devices. Four patients could read large-print text without audio cueing.

The Processing Unit: AI at the Edge

The 40nm custom ASIC at the heart of the processing unit runs a quantized 1.3B parameter vision reconstruction model entirely on-device. No cloud. No latency. No data leaving the patient's body.

The model was trained using a novel self-supervised technique the company calls VisualEcho, which learns temporal context from unlabeled video rather than relying on paired image-caption datasets. The result is a model that understands spatial relationships — that a coffee cup is taller than a coin, that a staircase descends — without ever being explicitly taught those concepts.

Timeline and Regulatory Path

NeuroBridge plans to file for FDA Breakthrough Device Designation by Q1 2028, targeting a PMA (Premarket Approval) pathway rather than the 510(k) route used for lower-risk devices. The company estimates commercial launch by late 2029, with a per-implant price point of $185,000 — roughly half the cost of competing brain-based approaches.

Investor Perspective

"Neural interfaces have been brain-first for a decade. NeuroBridge proves the periphery matters," said a partner at Fidelity Ventures who asked not to be named. "The unit economics of an eye-based device are dramatically better than a brain implant. This could scale."

NeuroBridge will use the new funding to expand manufacturing capacity, run a 120-patient pivotal trial, and build out its European regulatory team ahead of a planned CE Mark submission in 2028.