Farms may automate before factories. The money hasn't noticed.

About 400,000 DJI spray drones were in use worldwide by the end of 2024, yet farm robotics drew $788M of the $83.8B in robotics rounds we track, 0.9%.

Factories get the humanoid demos. Farms are getting the robots that actually work for a living. A spray drone that flies a field on its own, a tractor that tills overnight with no one in the cab, a laser rig that kills weeds plant by plant: these are not pilots, they are products with price lists, dealers and payback maths. The thesis of this feature is simple and, we think, under-reported: agriculture may reach routine autonomy before the factory floor does, because farms have a worse labour problem, simpler motion and a clearer bill to pay. The counter-argument is just as real, and it is written in the 2025–26 failures of Monarch Tractor and Naïo. We cover both.

The farm labour gap is structural, not seasonal

US farms increasingly import their harvest crews. The Department of Labor certified 384,900 H-2A temporary agricultural positions in FY2024, against 48,000 in FY2005^[4]. USDA puts the average non-supervisory farmworker wage at $18.13 an hour in 2024^[5]. John Deere cites an American Farm Bureau Federation estimate that about 2.4 million farm jobs need filling every year, and says more than half of machine-operator jobs posted by California specialty-crop farms go unfilled^[1]. Treat both of those as industry claims; they are not government statistics.

The average US farmer is about 58 years old^[21]. Unlike a factory, a farm cannot move its shift: tillage, spraying and harvest each have a window measured in days, and a missed window costs yield. That is why the sales pitch for farm autonomy is not "fewer people" but "more hours": Deere's own launch material says autonomy lets a tractor keep working overnight while the farmer does something else^[2].

Autonomy is arriving as a retrofit kit, not a new robot

The machines that are scaling look like tractors because they are tractors. Deere showed its second-generation autonomy kit at CES 2025: 16 cameras in pods for a 360-degree view, better long-range depth estimation, and for now tillage only^[1]. It released the kit in February 2025 as a precision upgrade for model-year 2022+ 9R/9RX and 2020.5+ 8R/8RX tractors^[2]; Farm Progress reported a limited 2025 release with full release in 2026, built on the Bear Flag Robotics platform Deere bought in 2021^[21]. Deere's stated goal is a fully autonomous corn-and-soybean production cycle by 2030^[3].

The same logic drives the challengers. Agtonomy sells autonomy software that OEMs build into their own machines, and Kubota announced an investment in it in April 2026^[15]. Carbon Robotics added an autonomy kit for existing tractors alongside its LaserWeeder^[6].

Weeding is the first job robots are winning on price

Weeding is narrow, repetitive and expensive, which makes it the best early fit for machine vision. Carbon Robotics says it passed $100M in revenue in the fiscal year to March 2026 and is preparing for an IPO^[6]. It claims LaserWeeder paybacks of under three years, the fastest under one; those are company figures^[6]. It has raised $177M to date and employs about 260 people^[7]. Our database lists the LaserWeeder G2 at about $1.4M^[20].

The weeding pioneers show how thin the margins are. Naïo, one of Europe's oldest weeding-robot makers, went into receivership in June 2025; a Toulouse court approved a takeover by a new company in October 2025, and the old entity was liquidated^[11]. The new Naïo kept 21 staff, has a €6.4M financing plan and targets 100 robots a year by 2028^[12]. FarmWise, maker of the Vulcan weeding implement, was bought by its customer Taylor Farms in April 2025^[13].

The sky automated first, and China is setting the pace

Spray drones are the largest fleet of autonomous farm machines in service. DJI estimates around 400,000 of its agricultural drones were in use worldwide at the end of 2024, flown by about 400,000 trained operators across more than 100 countries^[16]. China's agriculture ministry reports more than 300,000 farm drones by the end of 2025, working more than 460 million mu a year (roughly 31 million hectares, our conversion), and a 76.7% mechanisation rate for ploughing, planting and harvesting^[17].

On the ground, China's Honghu T70 driverless electric tractor follows pre-planned routes with a stated accuracy of ±2.5 cm and up to six hours of work per charge, and is reported in use in Hebei province^[18]. It was first shown in 2025; we found no independent field data. In the US, the policy wind blows the other way: since 22 December 2025, new foreign-made drone models can no longer receive FCC authorisation, although previously authorised models can still be sold and flown^[19].

Farm robotics gets about 1% of robotics venture money

Our capital database holds 411 robotics rounds worth $83.8B. Agriculture accounts for 26 of them and $788M, or 0.9%^[20]. Humanoids raised $11.9B across 30 rounds. That gap is the real story: the robots with the clearest buyers sit in the least-funded corner of the industry, and two of its best-known names (Monarch, Naïo) ran out of runway in the past 18 months.

Five signals that will settle the thesis

1. Whether Deere moves autonomy beyond tillage before its 2030 target, and how many kits it reports in the field [3].
2. Carbon Robotics' IPO filing, which would give the first audited numbers for a pure ag-robot company [6].
3. Which DJI models, if any, get a US exemption from the FCC Covered List, and how fast US-built spray drones fill the gap [19].
4. Independent field data for the Honghu T70 and other Chinese autonomous tractors [18].
5. Whether agriculture's share of robotics funding moves off roughly 1% in our database [20].