BCI Monthly - March 2026
March 2026 ~ The race to market is real now.
March was one of the more consequential months the field has had in a while. Not because of any single breakthrough, but because of what happened in aggregate: serious capital, a regulatory first, and big tech showing up with a foundation model. The commercialization phase isn’t coming… it’s here.
Money
Science Corp closed a $230M Series C at a $1.5B post-money valuation. The PRIMA retinal implant has a CE mark application in and a European launch expected mid-2026 (Germany). That would make Science Corp the first BCI company with a vision restoration product actually on the market.
Cognito Therapeutics raised $105M Series C to push its gamma-frequency Spectris headset through Alzheimer’s trials. New EEG data at AD/PD 2026 showed attenuation of EEG slowing. This signal is early, but it’s something to watch.
Smaller raises: Mave Health took in $2.1M for consumer neurotech wearables (focus and stress). MindMaze Therapeutics picked up a “Buy” rating from Baader Bank. This sell-side attention for a Swiss-listed neuro-rehab company is a meaningful signal.
Total capital deployed into the sector this month: north of $335M.
Regulation
The headline: Borui Kang Medical Technology received Chinese regulatory clearance for a minimally invasive extradural BCI targeting hand grasp restoration in quadriplegia. Widely reported as the first commercially authorized BCI medical device of its kind in the world.
Meanwhile, NeuCyber publicly stated it trails Neuralink by ~3 years. This is an unusually candid admission that reads more like strategic positioning than modesty. Beijing is building its neural-interface bench in public view.
On the non-invasive side: a multicenter RCT in Nature Digital Medicine provided solid clinical trial evidence for at-home tACS in depression. They did this by building on Flow Neuroscience’s FDA approval for at-home tDCS late last year. The at-home neuromodulation category is real.
Research Worth Knowing
Meta’s TRIBE model is the thing people are talking about most. Meta AI introduced a foundation model for neural decoding that achieves 70x resolution gains in cross-individual, cross-language decoding via in-silico simulation. Big tech is no longer sitting on the sidelines.
Handwriting BCI: A Nature Communications paper showed that motor cortex encodes handwriting as a full multidimensional movement: 3D velocity, grip force, pressure, EMG. Incorporating all of those features meaningfully improved decoding. This demonstrated that the signal is richer than the interface has been capturing.
Ultrasound wristband: A non-invasive wrist-worn device used AI to decode 22 degrees of freedom in real-time hand tracking from muscle and tendon kinematics. 22 DoF non-invasively is a number that should get your attention.
Closed-loop neuromodulation had a strong month. EEG-gated rTMS and adaptive MI-BCI difficulty both advanced; the trend toward tighter integration between decoding and stimulation is compounding.
MI-BCI methods: New architectures including DSSICNN (graph-attention spectral), polarity-aware microstates, and MEAN multimodal fusion (EEG+ECG+EOG). The non-invasive decoding stack keeps getting better.
The Actual Bottleneck
Commentary this month correctly put the spotlight on decoder training — the data requirements, the calibration burden per subject, the lack of generalization across individuals. Meta’s TRIBE is partly a response to this. So is the cross-participant encoding work on transfer learning for cortical mapping. The implant question is mostly answered. The decoder question is not.
The hardware isn’t the problem. Calibration is.