Where the work is, and where the money went
US agriculture is counted well at the top and thinly exactly where a machine has to work. One published crop budget decomposes a specialty crop's labour line by line, and reading it next to where capital actually went is the most useful ten seconds in this review.
Which farm labour line is the biggest?
Two lines from the same budget for the same crop, so they are directly comparable: the thing field robotics was funded to automate, and the thing that costs the most.
The estate, ranked by where the value actually sits
Ten opportunities, ordered by the review's own reading of pool size against readiness. Pick one to see what is measured and what is not. Note how far down the list the thing most capital chased appears, and what the review says about the one at the bottom.
The ledger: what happened to the companies
Every closure, withdrawal and litigation the review tracks, with the grade of the evidence behind each. This is not a prediction about the field. It is the record of what has already been resolved, which is the only part of a forecast anyone can check.
| What | Who | Outcome | Grade |
|---|---|---|---|
| Purpose-built electric tractor | Monarch Tractor | Ceased operations 3 April 2026; IP acquired by Caterpillar, terms undisclosed | verified |
| Automated weeding service | FarmWise | Wound down after about $95.7M raised; acquired by Taylor Farms 4 April 2025, and the buyer was investor and customer | reported |
| Field robotics closures | Small Robot Company, Iron Ox | Liquidation Feb 2024 after $13.18M; Iron Ox closed after $102M | reported |
| Commercial harvest robotics | Abundant, advanced.farm, Tortuga | All three exited between June 2021 and March 2025 | reported |
| Field robotics, long-running vendor | Naio Technologies | 350+ cumulative; judicial recovery June 2025, relaunched with EUR 6.4M | reported |
| Monarch dealer litigation | Burks Tractor and two others | Ten tractors, about $800,000 financed, alleged not to operate autonomously | reported |
What would settle it
The review does not say the field is wrong. It says the record is too thin exactly where a machine has to work, and it names three measurements that would resolve the question, each of which someone could start taking this season:
| Measurement | Why it decides the question |
|---|---|
| A harvest-robotics deployment publishing acres covered and cost per acre | The largest labour line has no published per-acre economics from any commercial deployment. One number would move the whole ranking. |
| A national count of acres on which hand weeding is actually performed | The $612M pool is modelled at full adoption. Without an acreage count nobody knows the real denominator. |
| Per-acre economics from more than two farms in one valley | The only published weeding economics located are grower-supplied, from two farms in a single valley. |
Every figure on this page is read from Physical AI on the farm (TR-2026-32) and nothing is added to it. The two labour lines come from the one specialty-crop budget the review located that decomposes a crop's labour, so they are comparable to each other and are not a national average. Constraint vocabulary: the eight.