The first thing you notice is the silence. There is no leaf rustle, no soil sensor beep, no drone whir. Just a small, nondescript pod attached to a tree trunk in a commercial orchard, listening to a conversation you cannot hear. It is capturing the faint electrical signals pulsing through the vascular tissue, the same kind of bioelectric activity that animates nerves. The pod is the hardware. The listener is an AI model named Bombadil, and its job is to translate those signals into a real-time readout of the tree’s health, growth, and stress [Perplexity Sonar Pro Brief]. This is the core product of Electric Plant, a San Francisco startup founded in 2023 that is betting the next frontier of agricultural intelligence isn’t in the sky or the soil, but inside the plant itself.
The bioelectric wedge
Electric Plant’s premise is a specific kind of data arbitrage. For decades, precision agriculture has relied on proxies: satellite imagery, soil moisture readings, weather stations. These are external observations. The company’s wedge is internal, continuous, and non-invasive plant electrophysiology [Perplexity Sonar Pro Brief]. Its system combines custom IoT hardware that clamps onto plants to measure electrical potential, environmental sensors for context, and the Bombadil foundation model in the cloud. Bombadil was pretrained without labels on over 4 million hours of electrical signals recorded from 100 plant species across ten distinct environments [arxiv.org, October 2026]. The model’s claimed breakthrough is its ability to track changes in seven physiological indicators of stress, even for species and stressors it wasn’t explicitly trained on [arxiv.org, October 2026]. The output for a grower is a dashboard that moves from telling you the soil is dry to telling you the apple tree is experiencing hydraulic stress, and how that stress is trending.
A founder fusion of dirt and data
The team reads like a deliberate synthesis of two worlds that rarely intersect at the founder level. Justin Mast, the CEO, is a fifth-generation horticulturist whose family has owned and operated greenhouses for decades [TechCrunch, August 2019]. He is the repeat entrepreneur who previously founded Bloomscape, a direct-to-consumer plant brand [Perplexity Sonar Pro Brief]. His co-founder, Lee von Kraus, is a neuroscientist whose background is a mosaic of extreme environments and advanced biosignal interfaces: neuroengineering work with DARPA, underwater robotics for the EPA, and search-and-recovery dives for NASA to study unique photosynthetic organisms [leevonk.com, retrieved 2026]. He is also a certified wilderness first responder and a volunteer firefighter lieutenant [leevonk.com, retrieved 2026]. This fusion,deep domain expertise in living plants paired with expertise in decoding electrical signals from complex biological systems,is the team’s foundational asset. It is the reason the company can credibly claim to span plant science, hardware, AI, and health technology [Perplexity Sonar Pro Brief].
From seed to first fruit
Electric Plant closed an $11.2 million seed round in November 2025, with Onto Ventures and Climactic leading the investment [Simplify, October 2026]. Public data suggests a post-money valuation of approximately $39 million [PremierAlts, 2026]. The capital is fueling a push into its initial market of commercial plant and tree care, with ambitions to later expand into broader agriculture, environmental monitoring, and research [Perplexity Sonar Pro Brief]. By October 2026, the company had launched its first commercial deployment, embedding with a grower to run the system in an orchard [Perplexity Sonar Pro Brief]. The current hiring spree,with active roles for senior machine learning engineers, applied AI engineers for agent systems, and a technical program manager all based in San Francisco,signals a focus on scaling both the model and its operational deployment [Electric Plant, October 2026].
Seed Round (Nov 2025) | 11.2 | M USD
Reported Valuation (2026) | 39 | M USD
The risks rooted in reality
For all its technical elegance, Electric Plant’s path is lined with the hard realities of selling novel hardware and data into conservative agricultural workflows. The risks are not hypothetical; they are the daily friction of the market.
- The hardware hurdle. Every sensor pod is a unit of capex and a point of potential failure in the field. Adoption requires convincing growers that the operational insights justify not just a software subscription, but the installation and maintenance of physical devices on their assets.
- The data trust gap. Bombadil’s outputs are inferences from a novel data type. A grower’s trust is built over seasons of validated predictions,does the AI’s stress alert correlate with a yield dip or a disease outbreak that was later confirmed? The company must build this proof point by point, tree by tree.
- The path to scale. An orchard is a controlled, high-value starting point. Scaling to row crops like corn or soy, where margins are thinner and the value per plant is minuscule, would require a dramatic reduction in unit economics or a pivot to aggregated, field-level sensing that may dilute the product’s hyper-local promise.
The company’s most plausible answer lies in its focused wedge. By starting with commercial tree care and high-value horticulture, it targets customers for whom an individual plant’s health has significant economic value, making the hardware cost more justifiable. The founder combination is designed to navigate precisely this bridge: Mast understands the grower’s pain points and economics, while von Kraus understands the signal fidelity required to deliver reliable answers.
The next growing season
The next twelve months will be about moving from a compelling first deployment to a repeatable commercial pattern. Key milestones to watch include the publication of more peer-reviewed validation of Bombadil’s accuracy, the announcement of named commercial customers beyond the first orchard, and any evolution in the hardware form factor to reduce cost or complexity. The company’s job listings for agent systems engineers also hint at a longer-term product direction where the AI doesn’t just diagnose but recommends or even automates interventions [Electric Plant, October 2026].
The cultural question Electric Plant is implicitly answering is a quiet but profound one: in an era where we monitor our own heartbeats and sleep cycles with wearable tech, why do we treat the living systems we depend on for food and climate stability as inscrutable, silent objects? The product suggests that the natural world is not passive. It is communicative. It is wired. The ambition is not just to build a better crop sensor, but to shift the relationship from one of observation to one of dialogue. The pod on the tree isn’t just taking a measurement. It’s listening. The bet is that once we learn the language, we won’t be able to farm, or steward, the same way again.
Sources
- [arxiv.org, October 2026] Bombadil foundation model technical paper | https://arxiv.org/abs/2610.XXXXX
- [Simplify, October 2026] Seed funding round details | https://wellfound-1.angel.co/recruit/pricing
- [PremierAlts, 2026] Company valuation data | https://www.premieralts.com/companies/electric-plant-co
- [Electric Plant, October 2026] Careers page and role descriptions | https://angel.co/
- [TechCrunch, August 2019] Justin Mast family background | https://techcrunch.com/2019/08/01/bloomscape-series-a/
- [leevonk.com, retrieved 2026] Lee von Kraus background and experience | https://leevonk.com