Can Robots secure Europe’s Food System?

A harvesting robot built for one crop's season can stand idle for 80% of the year. For a venture fund that invests for five years and holds for ten, Philip Stark of OysterBay said, that is the trouble with most farm robots. "We have unfortunately sometimes nature or biology that is not a big fan of a fund model of 10 years." The Ripple Philip hosted with Julius Strauss of FoodLabs kept finding the same answer. Robots make venture sense where the work never stops, and farmers would rather buy a harvested field than a machine.

At the start, every hand went up when Julius asked who expected robotics to take a much larger share of agtech by 2035. The next half hour was about how hard that will be.

Where the work never stops

Philip's two robotics investments both avoid the seasons. Goodbytes, a kitchen robot from Hamburg, can cook up to 3,000 meals a day. Its first customer was the University Hospital Tübingen, and it has just signed up the US Army. Eternal harvests tomatoes and cucumbers in greenhouses, which need harvesting all year. About 92% of greenhouses follow the Venlo design, Philip said, so greenhouses around the world look much the same and robots can be deployed widely.

Barns run all year too. A founder putting round-the-clock computer vision into turkey, chicken and pig barns said most are still run on the farmer's gut feeling. The images also help train seasonal workers from the Philippines and Vietnam who stay three or four months and arrive with no barn experience. New animal-welfare rules that demand documentation give farmers a reason to buy, although a new chicken barn takes at least three years to get approved.

Open fields have the biggest labour gap and the hardest engineering. An agrifood investor cited 300,000 open positions in US field agriculture last year, and 800 applicants. But a field robot has to get past a log that fell from a tree the day before, and the technology may not be ready within a fund's life. Julius added that farms, unlike car plants, have seasons with very different tasks, and a separate robot for each makes the farmer's upfront bill too high. Many founders, Julius said, can build good businesses that may not be venture-scale, because they have no path from one task to the next.

We have unfortunately sometimes nature or biology that is not a big fan of a fund model of 10 years.

Selling the harvest

Julius's tests for a venture case were pain big enough that farmers will pay, a clear path to reliability, and a platform that can take on more tasks. Farmers are approached by start-ups almost every week, and a robot that breaks during harvest, with no workers to step in, can be catastrophic for a farm. The agrifood investor added speed of deployment and whether a company can survive two or three rounds before its valuation is big enough to return at least 10x.

Capital is the sticking point. Follow-on investors balk at using equity to build hardware, Philip said. Greenhouse operators accept a payback of two to three years and some banks will underwrite robots once they work, but getting there is the hard part. Philip's advice was to buy reliable parts, often from China, and build the edge in software. Someone from an agri-fintech start-up said farmers find it hard to judge whether a robot pays, and traditional German banks need help before they will underwrite one.

An investor asked why the farmer should carry the capex at all, since a subscription priced on the productivity gain had gone down well in medical robotics. Julius said most founders in the field now sell robotics as a service. A farmer is unlikely to pay half a million euros upfront to a start-up founded two or three years ago with no service team. Full autonomy is also years away, so start-ups still need remote operators and people in the field when robots get stuck. What the farmer buys is the promise that the field will be harvested, ideally for less than manual labour costs.

Cooperatives complicate the sale. In Lower Saxony, one participant said, cooperatives and machinery rings buy most equipment, so one robot might serve 20 to 30 farms. Farmers have bought from the same dealers for generations, Julius said, but frustration is growing. Repairing your own tractor now voids the warranty, and a two-week wait for service at harvest time is a disaster. That opens a door for newcomers.

Europe's chance, Julius said, is not in actuators or batteries, a race it will not win. A participant from Hong Kong had just met a Chinese start-up expanding abroad because nobody in China will pay for its robots. It is targeting berries and other high-value crops in Europe, the market most willing to pay. Europe has farmland and no longer has the people willing to work it, Julius said, so it should source hardware abroad and build local champions that operate farms. It is not the answer anyone hoped for 10 or 15 years ago.

Julius closed by asking why every hand had gone up after so much talk of difficulty. One participant admitted to innovation bias, since most farmers want to keep working the way they always have. Another gave a simpler reason. "I think it's the plain fact that no one wants to be the farmer any longer."

This Ripple was hosted by Julius Strauss (FoodLabs) and Philip Stark (OysterBay) at The Drop 2026 on 16 September.