Plants may not speak in words, but they constantly transmit electrical signals through their tissues. Electric Plant is betting that those signals can become a practical early-warning system for agriculture, with its new Bombadil model turning faint changes in plant voltage into readable information about health and stress. The startup has built an AI translator for plant electrical activity, pairing it with a sensor called Bud that attaches directly to living plants.
Electric Plant says Bud captures electrical activity in plant tissue and sends the data to Bombadil in the cloud. The model then estimates seven physiological stress indicators, potentially giving growers a way to adjust irrigation or fertilizer before problems become obvious in leaves or fruit. The company says Bombadil can interpret signals across roughly 125 plant species, including crops, trees, ferns, and flowers.
The underlying science is less fanciful than the phrase “plant language” might suggest. Researchers have documented several forms of plant electrical signaling, including action potentials, variation potentials, and system potentials. These signals can be triggered by changes in light, temperature, water availability, touch, or physical damage, and can travel through a plant’s vascular system.
Bombadil’s technical foundation is described in a company-authored paper published on arXiv. The model was pretrained without labels on more than four million hours of electrical recordings from 100 plant species across laboratories, greenhouses, commercial orchards, and natural forests. A later prediction layer is designed to track seven indicators of physiological stress, including measurements related to photosynthetic performance and water stress.
Electric Plant was founded by Justin Mast and Lee von Kraus, a neuroscientist with experience working with DARPA, after several years of collecting plant-signal data. The company emerged from stealth with $11.2 million in funding, positioning Bud and Bombadil as tools for growers rather than novelty devices for listening to houseplants. Its stated goal is continuous, direct measurement of plant condition, replacing occasional visual checks with a stream of biological data.
That ambition comes with an important qualification: Bombadil’s headline capabilities currently rest largely on the startup’s own dataset and technical publication, not broad independent validation. The next test will be whether its readings remain reliable across new species, climates, growing systems, and real-world farms. If they do, Electric Plant could give agriculture a sensor-driven view of crop stress that arrives before conventional symptoms—and offer a remarkably literal way for humans to start paying attention to what plants are telling them.
