Nobody has automated it. Vinea is building the machine that does — an autonomous harvester for Dutch high-wire houses that cuts ripe trusses and maps the crop in the same pass.
No hardware yet. Nothing has run in a glasshouse. There are no performance figures on this site, and there will not be until the machine produces its own.
Scroll
The problem
The harvest is the last thing in a Dutch glasshouse still done entirely by hand.
Climate, irrigation, transport carts, packing lines and spraying are automated. Picking is not. Every ripe truss in the Netherlands is still cut by a person walking a trolley down a row, and that person is getting harder to hire every season.
Both growers we interviewed described the same squeeze, and neither described it as a headcount problem. Harvest labour is almost entirely Eastern European agency staff, and every glasshouse in the region competes for the same pool. Availability tightens as the season deepens — exactly when the fruit does not wait.
A 5.5-hectare owner-operated grower puts twelve people on 38-hour weeks and pays €5 per square metre per year to get the fruit off the plant.
The fifty-hectare operator pays less per square metre — he gets agency rates, shares labour across six sites and amortises supervision. We are not publishing his figure, because he gave it all-in and we never established whether it covers harvesting alone or every hour worked in the house, which makes it incomparable to the €5 above. The direction is the part that survives: the grower who pays the most to pick by hand is the small owner-operated one.
Two Dutch tomato growers, interviewed in full: 5.5 hectares owner-operated, and 50 hectares across six sites. Both still talking to us. Both named disease scouting before they named anything else.
What it does
It cuts ripe trusses. It maps the crop while it does.
A rail-guided trolley drives the row on the pipe rail the glasshouse already has — the same rail the pickers' trolleys use, which is why it needs no rebuild and no new infrastructure. Two arms work, one per side of the aisle. Neither of those is the interesting part; they are what it takes to be allowed in the building.
Drives the row
The trolley runs the heating pipe as a rail, so its position along the row is known without localising against a moving crop. At the headland it drops ground wheels, crosses to the next aisle and picks the rail back up.
Finds the ripe fruit
Deck cameras look along the row and build the work list. Wrist cameras take over inside the canopy, where the truss is occluded by leaf and the deck view is no longer good enough to cut against.
Cuts the truss, not the fruit
A cradle takes the weight of the whole truss from below before a blade severs the peduncle above it. The grower objection this answers is specific: existing harvesting robots cut the vine and let tomatoes fall off the stem. A truss that arrives intact is a truss that packs.
Places it in a crate
The arm carries the supported truss to a crate on the deck. Full crates go out at the headland.
Maps what it saw
The cameras that find fruit see every plant in the house several times a week. Vinea turns that into a per-row record of the crop — where disease is starting, where fruit is setting, where a row is behind. Nothing on the market does this today, and both growers asked for it unprompted.
Scouting is not a second product bolted on. It is what the harvesting cameras already produce, and the reason a grower who does not yet trust a harvesting robot will still take a machine into the row.
How this is different
Three bets, and none of them is the rail.
Driving the existing pipe rail is table stakes. Every trolley in a Dutch glasshouse does it and so does the competition. What follows is where Vinea actually differs — stated as bets, because that is what they are until a machine runs.
Harvest and scout on one platform
The harvesting robots on the market pick, and that is all. The scouting robots scout, and that is all. A grower who buys both buys two machines, two service contracts and two trolleys competing for the same aisle. Vinea's cameras earn their keep twice.
Grower-evidenced. ACRES: nothing on the market scouts for disease, and a harvesting robot that did would change the calculation. LEMA named harvesting and disease scouting together as the primary pain, unprompted.
Rented by the month, not bought
A grower who has watched spray robots break does not want to own the first generation of anything. Renting puts the reliability risk on us, where it belongs, and turns a capital decision into an operating one. Pricing is provisional until a pilot prices it.
Hypothesis. Both growers rejected current harvesting robots on readiness, not on desire — reliability is the barrier a rental model is meant to absorb.
The crop record compounds
Every deployed machine looks at hundreds of thousands of plants a season. That is the dataset the disease models need, and it does not exist yet because nobody has had cameras in the rows every day.
Bet, and the furthest out. It requires machines in glasshouses first, which is why it is third on this list and not first.
The objection
“Robots cut the vine, the tomatoes fall off the stem, and it runs at about five percent of a person.”
This is the most useful thing either interview produced, and it is not a rejection of automation. ACRES already runs transport carts, packing automation and spray robots. He wants de-leafing automated tomorrow. His objection to harvesting is mechanical and specific: the tool damages the product and the machine is too slow to matter.
Vinea's answer to the first half is the cradle-and-blade tool — support the truss from below, then cut the peduncle above it, so nothing is ever held by the fruit. The answer to the second half is that there is no answer yet. Speed is the hardest problem in this category, several well-funded companies have taken years over it, and we will not have a defensible figure until hardware runs.
Where this stands
Honestly: early.
- No hardware
- Nothing has been built. Nothing has run in a glasshouse.
- No performance figures
- The first simulation produced numbers that did not survive review, so it is being rebuilt rather than re-quoted. Nothing measured on it appears anywhere on this site, and no rate figure will appear until the new one produces its own.
- Two growers, in depth
- Two Dutch tomato growers interviewed in full, both still in contact. The labour economics on this site come from them, and they are the sole numbers here that are not public record.
- Solo, and building
- One founder. The design work, the simulation and this site are all his, done alongside full-time study.
Talk to us
The next thing that moves this forward is a conversation with a grower.
If you grow high-wire tomatoes in the Netherlands: forty minutes, no deck. What we want is your harvest hours, your loaded hourly cost, and the physical numbers of your rows. What you get is early access to a machine designed around your house rather than a demonstration greenhouse.
If you invest: the risk register is on the investor page, written before the pitch. Read that first and the call will be shorter.