Electric Plant, a startup that uses artificial intelligence (AI) to interpret plants’ electrical signals, has raised US$11.2 million in seed funding to develop technology that helps growers understand what their crops need. The company announced its launch on Oct. 20, backed by Onto Ventures, Climactic and other investors.
Its AI model, Bombadil, draws on three years of data collection and aims to help growers identify crop stress and make better decisions about watering and fertilizer. Co-founder and chief executive Justin Mast developed the idea after reading about a houseplant connected to a motorized stand that moved toward light in response to the plant’s electrical signals. He later teamed up with a neuroscientist Lee von Kraus to turn that curiosity into a business.
Plants produce small changes in electrical voltage as they photosynthesize, transport water and respond to stress. Researchers have studied these signals for years, but Electric Plant says Bombadil can interpret them at a scale that could make the technology useful to growers. The company’s claims about improved crop yields and water savings still need independent validation, and existing research shows that accuracy can vary significantly between species and growing conditions.
How AI reads plants’ electrical signals
The system has two parts: a low-cost sensor and the AI model that interprets its data. The sensor, about the size of an iPhone folded lengthwise, attaches to a plant with stainless steel screws that reach the cambium, the layer of living tissue just under the bark, so it can detect changes measured in microvolts (millionths of a volt).
It sends its readings through a low-power wireless network to a base station. Bombadil then analyzes them in the cloud and returns advice that growers can use to adjust watering or fertilizer. Mast says a few sensors per acre are enough and that each costs tens of dollars, while traditional tools often cost thousands and may not reflect plants’ needs in real time.
Bombadil was trained on recordings from more than 100 plant species, ranging from ferns and redwoods to tomatoes and dahlias. Mast calls it the most diverse model ever built on biological electrical signals.
The company’s press release says the model learned without labels, meaning people did not tell it what each signal represented, and can estimate seven stress markers, including a widely used measure of water stress called the Crop Water Stress Index. Mast claims such a system could raise yields while cutting water use by 20% or more, but those figures come from the company itself and still need independent validation.
What research says about plant electrical signals
Independent research supports the basic idea but also warns that results vary. A 2019 study in Scientific Reports found that machine learning could predict drought stress in greenhouse tomato plants with up to 98.5% accuracy by reading their electrical signals. Other teams have used similar approaches to detect nutrient shortages in crops.
A more recent exploratory study from late 2025, however, found accuracy of only about 66% for tomatoes and 50% for apricot trees. The authors described those results as insufficient for practical use, although they saw potential in the approach.
That contrast suggests that results depend heavily on the species, the setting and the type of stress. A model trained in one environment may struggle in another. Electric Plant is also not alone in this field: PhytlSigns, a similar sensor produced by Swiss company Vivent, recently received 1.8 million euros to expand into field crops.
The idea of plants having a “language” also requires caution. The description works as a metaphor: the system translates electrical signals into practical recommendations rather than words. Even so, plant electricity has moved from scientific curiosity to commercial application, and field results will show whether Bombadil can deliver on its promises.