The agricultural autonomy industry has a consensus problem.

Ask almost anyone in it whether autonomous systems have a role in the future of farming, and the answer is yes. Ask the same people why adoption keeps falling short of projections, and you get a different answer every time: the technology is not ready, the use cases are too narrow, farmers are just resistant to change.

What you will not hear is this: the economics have never been properly worked out, and the industry has been treating an economic problem as an agronomic one. For dealers being asked to represent autonomous platforms, and for OEMs deciding where to invest, that distinction matters enormously.

The Framing Problem

The debate is almost always conducted in agronomic terms. Compaction, row spacing, timeliness, chemical reduction, labor substitution. These are real considerations. But agronomic benefits only matter if the economics work first. The industry is solving for the wrong variable.

Timeliness is the clearest example. The argument for smaller, lighter autonomous systems often rests on their ability to extend the available operational window. But weather and crop stage are the binding constraints, not the length of the working day. A planting or spraying window does not open because the machine can run at 2 a.m. The response to time pressure is volume. Either you have enough capacity working fast enough to cover the acreage in time, or you do not. That is an economics question, not an agronomy one. Smaller machines change what the problem looks like. They do not change what solves it.

The labor argument has the same problem. In commercial broadacre farming, one operator already runs a large, efficient machine. The robot is not replacing a team. It is replacing one person. The economics of that are categorically different from replacing hand labor in specialty crops where a single field can require dozens of workers. The pitch adjusts for this; broadacre leads with agronomy and timeliness, specialty leads with labor. What both versions tend to leave out is the explicit economic comparison between the two.

The compaction argument gets raised with equal confidence. Lighter machines, fewer passes, better long-term yield. The logic sounds clean. The reality is more constrained. No autonomous system today covers the complete crop cycle, which means a conventional machine still enters the field for some operations and the compaction differential narrows from the start. In vegetable systems, where most current robot activity is concentrated, production runs predominantly on beds for agronomic reasons. The wheel tracks between beds carry no crop. Compaction is not the constraint there. In North American broadacre, where compaction would be most relevant, no-till and direct seeding already covers more than 35% of U.S. cropland, and GPS-guided controlled traffic has narrowed the differential further. The compaction argument has real validity in European small-farm systems with more conventional tillage practices. It does not close the economic case in commercial broadacre.

What Dairy Robots Actually Tell Us

The most useful reference point the industry has for honest robot economics is not a field trial. It is 30 years of commercial robotic milking.

Iowa State economist Larry Tranel has been tracking the numbers longer than most dealers have been selling the machines. His framing is precise: "Cash flow of a robot tends to be very negative in the first seven years, then pretty positive for rest of the life of the AMS, but that is dependent on many variables, especially repair costs across the whole life of the robot."

That pattern reveals something structural. The economics of robotic milking require the system to remain supported and operational for long enough that the back half of its useful life can recover the investment the front half cannot. Shorten the supported life and the math collapses. The payback horizon is not a financing detail. It is the economic architecture of the whole investment.

A USDA study published in January 2026 confirmed that robotic milking delivers around 13% net return improvement on average. The economics work, over time, with the right institutional backing. But adoption after 30 years remains limited and concentrated in specific conditions. Iceland and Sweden lead global adoption at around 30% of farms. The US sits at roughly 5%. The difference is not technology access or farmer sophistication. Each robot unit milks around 50 to 60 cows. That fits a small European or Icelandic herd precisely. A large US dairy operation would need 50 or more robots to replace a rotary parlor that two or three people already run efficiently. Above that scale threshold, the economics of robotic milking do not work, and they have not worked for 30 years of trying.

Lely and DeLaval do not lead their pitch with return on investment. They lead with cow welfare, farmer quality of life, and labor availability as a survival question. That is not modest marketing. It is an honest characterization of what the economics actually support. The category succeeds where it does because the non-financial benefits are compelling enough to carry marginal economics across a long horizon, backed by the institutional infrastructure of established OEMs who guarantee the service life the payback horizon requires.

Field robots have none of those foundations. The relevant question is not whether the economics eventually work. It is whether break-even arrives within the period you can credibly assume the platform will be supported. For early-generation field robots, a five-year support horizon is more realistic than fifteen. Those five years have to carry the full weight of financing, depreciation, operating costs, and ROI. That test looks very different depending on what problem the robot is actually solving.

Why the Residual Problem Compounds

The challenge is not just the absence of a secondary market. It is the risks that multiply underneath it.

The first is company survival. A University of Nebraska analysis of 18 agtech shutdowns in 2025 named the pattern precisely: a cost-adoption mismatch, where the capital and operational burden of a technology exceeds the farmer's capacity to adopt it, even when technical performance is validated. Technology working. Capital committed. Commercial scale not reached.

The second is acquirer continuity. Companies that survive are often acquired by necessity or by design, which raises a question: how confident are you in the preservation of the product under new ownership?

The third is software obsolescence. Precision agriculture platforms from a decade ago are already running into support lifecycle problems as OEM software stacks evolve. A farmer who assumes the machine has a useful life independent of software integrity is making an assumption the industry is already questioning.

The fourth is the complete absence of price discovery. In an established asset class, decades of comparable transactions anchor residual assumptions. Certified refurbishers, extended warranty products, and third-party parts networks all emerge from that history. None of it exists for autonomous field platforms today.

There is a fifth problem that compounds all of them. Early adopters pay the highest purchase price at the moment the manufacturer is least able to offer commercial terms. Volume is low, production costs are high, and no secondary market exists to anchor resale value. The capital risk is greatest exactly when commercial confidence is lowest.

These risks multiply at exactly the moment dealers are being asked to stake their service reputation on new product lines, and customers are being asked to make significant capital commitments in a category that has not proven commercial-scale economics. The logical response is the total loss model: build the economic case assuming zero residual value at the end of a credible support horizon. If the investment cannot be justified on that basis, it cannot be justified at all.

The Switching Cost Nobody Discusses

Deploying an autonomous system alongside an existing conventional fleet is not optional during any transition. The economics of both systems are degraded while the parallel operation runs. Brand switching in conventional equipment is hard enough. Adding an autonomous system from outside the established fleet is not a brand switch. It is a structural change, with all the compatibility gaps, redundancy, and support friction of a mixed fleet, plus the additional risk of backing a platform that may not exist in 15 years.

If a customer commits meaningfully to an autonomous platform and the company behind it fails five years into a 15-year asset life, they have not just lost the investment. They have paid the switching cost twice: once to get in, and again on the way out, into a market that still has no consistent solution to move toward. In a swarm deployment that risk is even more concentrated. One machine with a faulty sensor loses you a day. A fleet of 10 units whose platform provider winds down loses you the operation. These are not comparable risks.

Where the Economics Hold Today

The systems that survive the payback horizon test will likely share one or both of two characteristics: a problem acute enough to compress the break-even timeline, and institutional backing capable of guaranteeing the support life the investment requires. Either can carry the case independently. Together, they substantially improve the odds.

John Deere's autonomous 8R, 8RX, 9R, and 9RX tractor platforms apply this logic differently: autonomy layered onto platforms with decades of market transactions, established dealer networks, and functioning second-life markets. CNH and AGCO are arriving with similar solutions. The autonomy kit adds cost; the base asset retains value independent of it. Sabanto extends this further, adding autonomous capability to existing Deere, Fendt, and Kubota tractors with no new iron required. Different organizations, different routes, the same conclusion: protect the underlying asset, layer the autonomy on top. If the autonomy layer is discontinued, the tractor remains. The switching cost does not attach in the same way, and neither does the stranded asset problem.

The Question for Dealers and OEMs, For Now

The economics of autonomous systems will improve. Secondary markets will develop. Institutional support will mature around the platforms that survive. The total loss assumption is a product of where the market is today, not where it will be in a decade.

But right now, when a startup approaches a dealer group about representing its autonomous platform, the agronomic case is usually compelling and the economics question is usually left to the customer to work out. That is the wrong order. If a platform fails commercially two years into a five-year support commitment a dealer made to its customers, the reputational exposure does not land on the startup. It lands on the dealer.

One question should come before all the agronomic arguments: does the break-even arrive within the period this platform can credibly be assumed to exist, and is the institutional backing behind it deep enough to get you there?

Platforms where the economics work on honest assumptions, backed by institutions with the commitment to honour them, are worth serious investment today. Everything else is interesting technology. The market simply has not caught up with it yet.