Every frontier technology goes through a phase where the demos are astonishing and the business models are vague. Brain-computer interfaces are squarely in that phase. The science is real and accelerating, the headlines are loud, and almost nobody can tell you, precisely, which company you would want to own ten years from now.
That gap is exactly where venture lives. Our job is not to be impressed. It is to be specific.
Start with the constraint, not the dream
The temptation with BCIs is to underwrite the dream: telepathy, memory, cognition on demand. The discipline is to underwrite the constraint. Today, the binding constraints are bandwidth, biocompatibility, surgical risk, and regulatory time. Each of these compresses the near-term market to a narrow set of high-value medical use cases where the alternative is far worse.
That is not a limitation. It is a map. The first real markets for BCIs look like restoring function: movement, speech, and vision for people who have lost them. These are markets where willingness to pay is high, where regulators have clear precedent, and where the bar is set by human need rather than novelty.
The first dollar in a frontier market is almost always a medical dollar. The platform dollar comes later, and to a different company.
Where value actually accrues
It helps to break the stack into layers and ask, at each one, whether value is likely to concentrate or commoditise.
- Electrodes and materials. The physical interface to neural tissue. Hard science, slow to move, and potentially very defensible. Whoever solves long-term, high-channel-count, low-immune-response recording owns a chokepoint.
- Implantable hardware and signal chain. The device itself: power, wireless, packaging, surgical delivery. Capital-intensive and regulated, which keeps out tourists.
- Decoding and software. Turning noisy neural signals into intent. This is where modern machine learning compounds, and where a data advantage could become a moat, if the data is proprietary.
- Applications. The clinical and, eventually, consumer products built on the layers below. Highest visibility, often lowest defensibility.
The naive instinct is to back the application, because it tells the best story. The investor instinct is to ask which layer has a structural reason to stay scarce. In BCIs, scarcity lives lower in the stack than the pitch decks suggest.
The data-flywheel question
Software people love to assume a data flywheel: more users, more neural data, better decoders, more users. It can be real here, but only if the data is both proprietary and transferable across people. If every implant requires bespoke calibration and the learnings do not generalise, the flywheel stalls and the decoder is a feature, not a company.
So the question we would put to any decoding-layer founder is blunt: does your model get meaningfully better with the next thousand patients, and can you keep that improvement to yourself? If the answer is soft, the moat is soft.
What we would watch for
A few signals would move a BCI company from interesting to investable for us:
- A regulatory path that is specific and already in motion, not aspirational.
- Ownership of a layer that stays scarce as the market grows.
- A clinical wedge with real willingness to pay, and a credible bridge from that wedge to a larger surface.
- A team that talks about constraints with more energy than it talks about the dream.
The Banyan view
BCIs are a textbook case of a technology that looks obvious as a theme and is genuinely hard as an investment. The theme is consensus. The edge is in being precise about which layer compounds, and patient enough to let the medical market do the early work while the platform quietly forms underneath.
Curiosity got us into this. Conviction will come from the constraint.