Ecorobotix moves 10 plant detection models to commercial status for 2027
Ecorobotix moves 10 plant detection models to commercial status for 2027. Its 2026 models cut inputs by 86 per cent across more than 11,000 hectares.
Ecorobotix announced its 2027 plant detection line-up on 29 October 2026 from Yverdon-les-Bains in Switzerland. Ten models move to commercial status from March, among them garlic, leek and potato, and growers can test a further 55 models at no charge, including pepper, zucchini and yellow nutsedge. The ARA ultra-high precision sprayer now carries more than 120 recognition and treatment models.
This is a software release rather than a machine release. ARA pairs the company's plant-by-plant artificial intelligence with ultra-high precision nozzles, and each model teaches the machine to separate one species from another inside a specific crop at working speed. Commercial status means a grower can buy the model and run it on a paying field. The free pool is where next season's candidates are proved.
The evidence behind the 2027 line-up comes from the season just finished. Models that reached commercial status in 2026 were run on more than 11,000 hectares in over 20 countries, with an average input reduction of 86 per cent compared with broadcast spraying. More than 1,000 ARA units are now in use worldwide and more than 600,000 hectares have been treated in total. Treated area for the carrot model, the company's most established, rose by more than 80 per cent.
The weed to watch is yellow nutsedge. Ecorobotix cites Agroscope research putting yield losses at 40 to 71 per cent in Swiss sugar beet and potato fields, and says the new algorithm is the first precision spraying approach able to detect that weed and place a non-selective herbicide directly on it. Simon Gasser, agronomist and crop algorithm product manager at Ecorobotix, presented the nutsedge model as the technical first in the line-up. Luc Détroz, a potato grower in Belgium, ran the potato model across 300 hectares and controlled goosefoot and knotweed in one pass. "This clearly saved the crop," he said.
Why this matters for AgTech founders: the asset here is the model library, not the sprayer. A thousand machines in the field generate the labelled imagery that makes the next model cheaper to build, and every crop added widens the set of farms for which the capital outlay pencils out. The commercial structure deserves as much attention as the technology. Ten models are sold, 55 are given away, and the free pool serves as the trial programme and the data pipeline for the following season at the same time. Anyone selling hardware into specialty crops should read the carrot figure rather than the headline count: one mature model growing its treated area by more than 80 per cent in a single season is what recurring revenue on an implement actually looks like. Depth in a handful of crops beats a long list of beta claims.
Ecorobotix moves 10 plant detection models to commercial status for 2027
The Swiss company puts garlic, leek and potato on the paid list from March and opens 55 further models for free testing. Models that went commercial in 2026 cut inputs by an average of 86 per cent.
Ecorobotix announced its 2027 plant detection line-up on 29 October 2026 from Yverdon-les-Bains in Switzerland. Ten models move to commercial status from March, among them garlic, leek and potato, and growers can test a further 55 models at no charge, including pepper, zucchini and yellow nutsedge. The ARA ultra-high precision sprayer now carries more than 120 recognition and treatment models.
This is a software release rather than a machine release. ARA pairs the company's plant-by-plant artificial intelligence with ultra-high precision nozzles, and each model teaches the machine to separate one species from another inside a specific crop at working speed. Commercial status means a grower can buy the model and run it on a paying field. The free pool is where next season's candidates are proved.
The evidence behind the 2027 line-up comes from the season just finished. Models that reached commercial status in 2026 were run on more than 11,000 hectares in over 20 countries, with an average input reduction of 86 per cent compared with broadcast spraying. More than 1,000 ARA units are now in use worldwide and more than 600,000 hectares have been treated in total. Treated area for the carrot model, the company's most established, rose by more than 80 per cent.
The weed to watch is yellow nutsedge. Ecorobotix cites Agroscope research putting yield losses at 40 to 71 per cent in Swiss sugar beet and potato fields, and says the new algorithm is the first precision spraying approach able to detect that weed and place a non-selective herbicide directly on it. Simon Gasser, agronomist and crop algorithm product manager at Ecorobotix, presented the nutsedge model as the technical first in the line-up. Luc Détroz, a potato grower in Belgium, ran the potato model across 300 hectares and controlled goosefoot and knotweed in one pass. "This clearly saved the crop," he said.
Why this matters for AgTech founders: the asset here is the model library, not the sprayer. A thousand machines in the field generate the labelled imagery that makes the next model cheaper to build, and every crop added widens the set of farms for which the capital outlay pencils out. The commercial structure deserves as much attention as the technology. Ten models are sold, 55 are given away, and the free pool serves as the trial programme and the data pipeline for the following season at the same time. Anyone selling hardware into specialty crops should read the carrot figure rather than the headline count: one mature model growing its treated area by more than 80 per cent in a single season is what recurring revenue on an implement actually looks like. Depth in a handful of crops beats a long list of beta claims.
