The agriculture industry has been slow to embrace new technologies and is still largely non-automated and non-digitized. But technology adoption now seems to be hitting an inflection point with artificial intelligence and robotics capabilities aimed at farming operations making a giant leap, as illustrated by the recent introduction of a fully autonomous tractor by Deere.
These advances couldn’t come at a better time as agriculture has been facing serious challenges this year including labor shortages, cost inflation and food security, challenges that are widely expected to last. Labor shortages in American farms have been widespread for a long time and find their roots in an ageing agricultural workforce, decreased interest in farming, reluctance to live in rural areas and stricter immigration laws. The issue is that shortages keep worsening, putting at risk farm productivity/yields and the world’s food supply.
According to the 2021 Global Harvest Automation Report, harvest labor represents around 20 to 50% of speciality crop production expenses and around 80% of total labor costs. Moreover, as most growers anticipate labour costs to increase by 10 to 30% over the next three to five years, a decrease in profitability is becoming a serious issue, especially because business viability is also ever more threatened by more stringent regulatory requirements.
Robotics and automated can then help alleviate the pressure on farming operations, amid increasing requirements to maintain food supply and prices at affordable levels to feed a population growing to 10 billion people by 2050 and to meet a food demand expected to rise by 50%.
As the adoption of robotics, AI and automation is becoming an urgent matter, noticeable initiatives are emerging. These include Western Growers (an association of farmers growing fresh produce in Arizona, California, Colorado and New Mexico), which aims at driving sustainable and domestic food security, by automating 50% of harvest in the US by 2031. Nevertheless, the challenges to achieve such a goal are significant, especially because the required agronomics is unique to each type of crop, and because each one is harvested in a very different way. These factors require that each culture automation system make use of distinct techniques in addition to specific AI software.
When it comes to self-driving, Deere has decided to do away with LIDAR contrary to most car makers, considering that laser-based detection of obstacles, along with assessment of their distance, direction, and velocity, was not relevant for a tractor operating in a field and decided to focus on a combination of stereo cameras and GPS guidance. The image processing and collected data are then handled by an Nvidia GPU and an artificial intelligence that has been trained to identify and navigate around most common obstacles (animals, plants…) and to stop and send an alert to the farmer when it cannot handle a situation. Overall, given the specificities of tractors compared to road vehicles (farming environment, dust, heat…), Deere could not simply borrow EV self-driving technologies and had to design or customize its own hardware and AI software.
Other noteworthy developments include the Trident 5550, which uses self-driving, high-tech cameras, and AI to interpret a stream of photos to analyze impediments and spread supplies in agricultural fields. The technology was designed by Raven Industries, which CNH purchased for $2.1 billion last year.
Self-driving has then made massive progress and is now gradually becoming a commercial reality in the agriculture landscape. But there’s still a lot to do in other agriculture segments with, for instance, fresh fruit picking remaining largely manual, then suffering from high harvesting costs and grappling with the challenge of assembling large armies of seasonal pickers.
Harvesting delicate crops without a human hand requires vision, autonomous motion, decision-making, reliability, dexterity, and speed. While machine vision can today identify different fruits against complex and varying backgrounds, the robotic path planning, picking strategy and the robotic arm’s motion control remain a work in progress.
Hence, an automated system like a robot is currently unable to match a human fruit or vegetable picker’s performance in terms of speed and accuracy at a reasonable price. It is estimated that human pickers are way faster today – around 2-3 seconds per picked strawberry vs. 8-10 seconds for a robot. But this speed gap is expected to narrow in the future and robots can also bring many productivity efficiencies to the table such as the ability to work 24/7 and the presence of several arms, offsetting the slowness of each arm.
As such, it’s more a question of when robots replace seasonal pickers and take over labor-intensive activities, not if.
Interestingly, a flurry of autonomous machines and robots are now moving beyond the prototype phase. Better Food Ventures and The Mixing Bowl, who combine their financing and thought leadership expertise in the disruptive agtech space, have identified nearly 250 farm robotics startups automating various crop farming activities both indoors and outdoors. Among these, Four Growers is developing robotics technologies, addressing greenhouse autonomous harvesting of tomato, pepper and cucumber crops at a competitive price. Farm-ng provides another example, with its all-electric micro-tractor named Amiga, which adapts well to the agricultural techniques of most farms, and which reduces the need for manual labour during harvests, in addition to saving on maintenance and fuel costs. Burro is another illustration of the space, manufacturing tiny, autonomous robots that can help agricultural laborers with a variety of transportation duties.
While the value proposition of autonomous machines for farmers is obvious (boosting farm productivity and easing human labor availability challenges), their impact on the financials and business models of agriculture machinery makers should be material as well as some features such as self-driving technology are widely expected to be sold on a subscription basis, with several consequences in sight.
First, for Deere, AGCO, CNH and peers, self-driving adoption and revenue growth could dramatically accelerate as affordable monthly payments (then lowered upfront costs), tailored offers (pricing based on the features customers unlock) and regular over-the-air software updates entice more farmers to try out the technology. Interestingly, most machinery vendors are currently eying the retrofit market. This will allow farmers to adopt the technology without flipping their entire fleet and spending millions (e.g. a Deere autonomous tractor carries a price tag of $500,000), while vendors will get the opportunity to quickly monetize their installed base.
Second, subscription revenue are highly recurring and then likely to lessen the cyclical profile of agriculture machinery companies.
And finally, software technology and subscriptions are likely to be much more profitable than the core business with operating margins expected to approach or exceed 30%. We calculate that for Deere, which forecasts to derive 10% of its revenue from software fees by the end of the decade, the impact on its operating margin could be in a 100-200bps range.
In conclusion, automation and robotics in agriculture are still in their early stages but appear as a secular trend that will spark a major equipment upgrade cycle and sustain revenue growth for machinery vendors, well beyond the current soft commo cycle.