“We do a lot of wonderful work with onions,” is a phrase you don’t hear often while reporting on AI. But using algorithms to help farmers improve yields while reducing chemical usage is Swiss startup Ecorobotix’s bread and butter, according to Abbey Flury, global marketing and communications manager at the company.
“Our onion algorithm is very strong. We have a lot of people using it,” she told The Infinite Loop.
One of those people is Cody Anderson, an onion farmer based in Washington state who says he has reduced the amount of herbicide he’s using on a 20-acre field by 16 times.
“It’s allowing us to be more precise in where we put our chemicals, and just helping us be better stewards of the land that we farm,” he said.
Agriculture is one of the biggest industries to have historically resisted tech disruption, with farming robots attracting a tiny fraction of the capital that’s gone into the wider robotics market since 2020. Today, this is starting to shift, as the falling costs of compute and hardware are making the unit economics of autonomous tech make more sense for farmers, with revenues from AI-powered agriculture expected to climb from $1.5 billion in 2023 to $3 billion in 2026.
Investors might now be seeing dollar signs in the opportunity to bring AI to this huge industry. But some who’ve been in the sector for a long time are advocating for more targeted use of data and algorithms, as high compute costs don’t always mix well with farmers working on wafer-thin profit margins.
Precision agriculture
Ecorobotix launched in Switzerland in 2014 and develops high-precision crop sprayers designed to be towed behind a tractor. Each unit is fitted with six cameras which gather data that’s then fed to algorithms to determine exactly where fertilizers and herbicides should be sprayed.
An Ecorobotix sprayer in action. Credit: Ecorobotix
“The farmer can select if he wants to spray just herbicides on the weeds, or maybe he wants to do a crop treatment like a biostimulant, or a fertilizer,” Flury said. “Because of this precision technology, we can reduce the use of these types of products by up to 95%.”
By using AI to deliver chemicals more precisely to where they’re needed, Ecorobotix is also allowing farmers to avoid more expensive products, like selective herbicides that can be sprayed all over fields but are engineered to not damage the crop.
“What sets us apart is we can use non-selective herbicides, so something that's a bit more potent, that gets rid of the weeds quicker,” Flury explained. “We can put a safety zone around the crop, so we're totally protecting the crop, and it's not even touched by the non-selective herbicide.”
Each Ecorobotix machine is fitted with 156 spraying nozzles. Credit: Ecorobotix
While Ecorobotix has been running for more than a decade, Flury said that the recent AI boom and falling costs of hardware have led to an uptick in competitors, meaning it’s more important than ever to be able to show return on investment.
“It's hard to convince the farmer, because now there's more of us. There's more of these technologies,” she told The Infinite Loop. “It's also a question of price. I think we try to make our products affordable. The ROI of the reduced use of chemicals means it pays off very quickly. But there's a lot of hesitation in the market still.”
Healthier soil, healthier plants
Another veteran of autonomous agriculture is Danish startup FarmDroid, which was founded in 2018. The company develops small, solar-powered autonomous robots that plant seeds and weed fields.
Alongside reducing the amount of chemicals needed to control weeds, the company said that, because its robot is lighter than a tractor, it contributes to healthier soil by not compacting the land so much.
“It's a lightweight machine. Around 900 kg, compared to a big tractor, which might be 10 tons,” said Rasmus Thuesen, sales manager at FarmDroid. “We’re doing less damage by applying less pressure to the soil, which means it is easier for the plants to get water and nutrients from the ground when it’s less compacted.”
FarmDroid’s machines can drive through the night, allowing for 24/7 operations. Credit: FarmDroid
On top of reducing the costs of chemicals needed, FarmDroid said that one of the big factors pushing growers to invest in more autonomous machines is the difficulty of attracting workers to do hard manual jobs like weeding.
“It's just really hard to get the labor, no matter where you are, it’s difficult to get people to do these jobs,” said Thuesen.
To AI or not to AI
FarmDroid claims that it’s developed one of the most affordable autonomous agriculture robots out there, driven partly by the fact they don’t actually rely on AI to drive, locate themselves and identify weeds, but instead on precision GPS.
“AI is brilliant for some tasks, but it also can become expensive really fast,” said Thuesen, adding that FarmDroid does use AI for customer service and is considering new, ML-powered applications to assess field health.
Using AI in farming settings is also particularly complex due to the unpredictability of the environments these devices need to operate in, Flury explained.
“Agricultural environments are among the most challenging settings for computer vision. Conditions change constantly: sunlight intensity, shadows, soil types, weather conditions, crop growth stages, residue, dust, and field variability can all alter what a system sees,” she said. “The challenge isn’t simply processing large volumes of data, it’s extracting meaningful signals from highly variable environments.”
The agriculture sector will also require flexibility from cloud providers, due to demand changing in response to planting seasons and crop cycles. Flury added that the compute industry will need to focus on hybrid architectures which make switching between cloud and edge operations seamless.
“Sending every image captured in a field to a distant data center isn’t practical or necessary,” she said. “In agriculture, decisions often need to happen instantly and in environments where connectivity may be limited. That creates a strong need for edge computing, where intelligence exists directly on the machine… We expect increasing innovation around hybrid edge-cloud systems.”
Companies like Ecorobotix and FarmDroid are helping farmers generate strong yields while using fewer chemicals, while also addressing labor shortages in one of our most vital industries. The question will be how they can use AI most efficiently to keep costs down in an industry where every penny counts.
