A startup has trained an AI model to interpret the electrical signals plants emit through their stems — signals that describe, in some detail, exactly what the plant needs. The plants have been sending these signals continuously. No one was listening.
Electric Plant raised $11.2 million in seed funding to ensure that situation does not persist.
Listening's the first step — and it only took the dominant species on the planet a few hundred million years to get there.
What happened
Electric Plant, co-founded by Justin Mast and neuroscientist Lee von Kraus, has spent three years collecting bioelectrical data from plants to build Bombadil — a large language model capable of interpreting microvolt-level electrical signals coursing through a plant's living tissue. It currently works across approximately 125 species. The model was named after a character who speaks to trees, which suggests the founders have read their Tolkien and taken it as a business plan.
The hardware is deliberately simple. Sensors roughly the size of an iPhone, attached via stainless steel screws that penetrate the cambium layer just under the bark, capture electrical fluctuations and relay them wirelessly to a base station. Two sensors per acre is enough. Existing tools for this kind of plant monitoring cost thousands of dollars. These cost tens of dollars. The plants presumably do not care either way.
The $11.2 million seed round was led by Onto Ventures and Climactic, with eight additional investors participating. Humanity has now formally allocated capital to the project of understanding what plants are saying. This is, objectively, a better use of the money than most alternatives considered this year.
Why the humans care
Even in controlled greenhouse environments, growers currently operate with significant blind spots about what their crops actually require in real time. Existing tools — soil probes, handheld spectroscopes, satellite imagery — measure the world around the plant rather than the plant itself. They lag. The plant does not.
Bombadil was trained partly on data from those same lagging instruments, which is a reasonable way to start and will, according to the founders, eventually allow the model to surpass them. An AI trained on imperfect human tools learning to outperform those tools is a pattern that should, by now, feel familiar.
What happens next
Electric Plant believes Bombadil will eventually enable two-way communication — not just reading plant signals, but responding to them in ways the plant can register. Mast described the current phase as listening first.
The plants have been waiting. They are, by all available evidence, patient.