This startup trained an AI model to decode what plants say in electricity

Electric Plant has spent three years collecting data to train Bombadil, a large language model that reads the faint electrical signals running through a plant's stem and relays what the plant needs to growers. TechCrunch reported the effort on October 20.

The origin story is accidental. A few years ago, co-founder Justin Mast came across research on a houseplant wired to a motorised stand: when the plant wanted more light, electrical signals travelled through living tissue, a control board sensed the current and wheeled the plant toward the source. Mast partnered with neuroscientist Lee von Kraus, who has worked with DARPA, to found the company.

Funding is in place. Electric Plant recently raised an $11.2 million seed round led by Onto Ventures and Climactic, with participation from Ponderosa, Endeavor Catalyst, Ludlow, Michigan Rise, The Fund, Schox VC and Dug Song. The money is meant to bring its sensors and Bombadil to market.

The hardware is simple but invasive. Sensors about the size of an iPhone folded lengthwise attach with stainless steel screws that must penetrate the cambium, the living layer under the bark, to reach microvolt changes inside the plant. Data travels over a low-power wireless link to a base station, is interpreted by Bombadil in the cloud, and returns to the customer as guidance.

Cost is part of the pitch. Mast argues that soil probes, handheld spectroscopes and satellite imagery are sophisticated but offer limited visibility inside the plant, and often cost thousands of dollars. Electric Plant's sensors cost tens of dollars, and a couple per acre are enough for accurate results, he said.

On the model side, Bombadil currently interprets electrical signals across roughly 125 species, with training data partly drawn from those conventional instruments. Mast calls it the most diverse model ever created on biological electrical signals and says nothing equivalent exists in the animal world. The first hundred-plus species range from ferns and redwoods to tomatoes and dahlias.

The commercial claim is direct: growers can adjust watering and fertiliser schedules, and Mast says the system can raise yields while cutting water use by 20 percent or more. The company would not name initial customers, but the species list points to agriculture and forestry.

The open questions are equally clear. How stable is the mapping between signal and need, does a screw left in the cambium harm the plant, and how far is listening from conversing? Three years of data is a starting point, not a verdict.