The thesis
Dutch glasshouse horticulture has automated climate, irrigation, transport, packing and spraying. Harvesting is the exception, and it is the largest remaining labour line in the crop. It is done by agency workers from a pool that every glasshouse in the region draws on, and it tightens every season and every time the wage floor moves.
The reason it is still manual is that it is hard, not that it is unwanted. Both growers we interviewed rejected the harvesting robots they have seen — on readiness, not on principle. That is the correct shape of an opening: demand established, supply not yet good enough.
Vinea's specific bet is that the machine that wins is not the fastest picker but the one that pays for itself twice. Cameras good enough to cut a truss are cameras good enough to map the crop, and a per-row disease record is the thing both growers asked for before they asked about picking.
What is actually established
Two Dutch tomato growers interviewed in full — 5.5 hectares owner-operated, and 50 hectares across six sites. Both still in contact. The labour figures on this site come from them.
- Harvesting labour costs €5 per square metre per year at 5.5 hectares — grower-confirmed, harvesting only. The fifty-hectare operator is cheaper per square metre, but his figure was recorded all-in and never disambiguated, so no scale gap is claimed here.
- Harvest labour runs at 1.5 to 2.2 workers per hectare — well under the industry figure the first model was built on, which is worse for the displacement maths and worth knowing early.
- Both growers named disease scouting as wanted and unserved. Neither had been offered it.
- Both rejected existing harvesting robots. One from direct experience of the fruit being damaged.
The model
Robotics as a service: the machine is rented by the month, we carry the reliability risk, and the grower makes an operating decision rather than a capital one. That structure is a direct response to what growers said about spray robots breaking. The price is provisional and will stay provisional until a pilot sets it — the earlier bands were derived from a rate target the simulation did not reach, and they have been withdrawn rather than restated.
The risk register
Written before the pitch rather than after the question.
- Speed is unsolved
- The hardest problem in the category. Well-funded companies have spent years getting a harvesting robot to human pace. There is no reason to think Vinea skips that.
- No hardware
- Everything so far is design and simulation. The first real plant will invalidate something.
- The first simulation's numbers did not hold
- They were withdrawn rather than defended, and the simulation is being rebuilt. Treat the absence of figures on this site as the correct state, not an omission.
- Solo founder
- One person, actively looking for a technical cofounder. Both previous cofounders left in July 2026 with nothing signed, so there is no equity or IP entanglement.
- Sequencing may be wrong
- One of the two growers would rather have de-leafing than harvesting. Scouting is the single thing both wanted. Harvesting-first is a choice that the next few interviews could overturn.
- Incumbents ride the same rail
- Running the existing pipe rail is table stakes, not a moat. The differentiation has to come from harvest-plus-scout and from the commercial model.
- A machine that damages fruit is worthless
- Class-1 packability is the pass/fail, not how many trusses an hour. It is also not yet demonstrated.
The ask
Not yet public. If you are looking at this seriously, ask on the call and you will get a number and what it buys, rather than a range that moves.