Precision agriculture ran AI's experiment for thirty years. The per-unit efficiency gains were real. The systemic benefits never arrived. The platform capture became permanent. Nobody in the AI debate is reading the agricultural data.
In 1996, John Deere released the first commercially available GPS guidance system for tractors. The pitch was identical to the one AI companies make today: precision technology would reduce waste, lower input costs, improve outcomes, and pay for itself. Thirty years later, GPS-guided variable-rate application has been adopted across roughly half of U.S. corn and soybean acreage. The technology works exactly as promised. The systemic results are the opposite of what was promised.
A peer-reviewed study published in npj Sustainable Agriculture in January 2026 analyzed the full body of research on precision agriculture's environmental claims and concluded that they were "not fully tested nor supported by evidence." The HEAL Food Alliance's thirty-year retrospective, released in September 2025, was more direct: fertilizer and pesticide use increased during the precision agriculture era, not decreased. The per-acre application became more precise. The total volume went up.
This is not a failure of technology. It is the Jevons paradox operating at agricultural scale, and it is the clearest preview available of what AI will produce in enterprise over the next two decades.
The Rebound
William Stanley Jevons observed in 1865 that James Watt's more efficient steam engine did not reduce coal consumption — it made coal-powered industry profitable in more applications, increasing total consumption. A ScienceDirect meta-analysis of global agricultural data from 1961 to 2014 confirmed the same pattern: higher energy efficiencies in agriculture did not result in lower energy use. At the global level, the findings fit the definition of Jevons' paradox.
Precision agriculture made the per-unit application of fertilizer and pesticide more efficient. That efficiency made it economically rational to cultivate marginal land, increase planting density, and push for higher yields — all of which increased total input use. The per-acre cost of applying chemicals dropped. The number of acres treated expanded. The chemicals became cheaper to apply precisely, so they were applied to more precisely identified targets across more precisely mapped fields. Precision was the mechanism of expansion, not reduction.
The same structure is visible in AI's first enterprise data. Per-token inference costs dropped eighty percent in a year. Enterprise AI spending doubled. Block eliminated forty percent of its workforce and reinvested the savings into more AI deployment. The efficiency gain did not reduce the resource consumed. It made the resource cheap enough to consume in more places.
The Displacement
GPS guidance did not just change how farmers applied inputs. It changed what farmers needed to know. Before precision agriculture, a farmer's value was embodied knowledge — reading soil conditions, recognizing pest pressure from plant appearance, adjusting application rates based on decades of field-level experience. GPS guidance moved those decisions from the field to the data layer. The farmer's role shifted from practitioner to equipment operator.
Iowa State University's Center for Agricultural and Rural Development surveyed farmers about precision agriculture's impacts. The statement that drew the highest agreement — seventy-eight percent — was that precision agriculture increases profits for machinery and technology companies. Seventy-one percent agreed it would lead to fewer and larger farms. The farmers understood the structural dynamics better than the industry promoting the technology. They adopted it anyway, because the individual economics were compelling even as the collective consequences were visible.
The parallel to AI-driven workforce restructuring is exact. Individual firms capture productivity gains. The knowledge that those workers carried — judgment, pattern recognition, institutional memory — disperses. Nobody's quarterly earnings reflect the loss.
The Data Crop
Civil Eats reported in March 2026 on what it called the hidden crop: farm data. Precision agriculture technologies collect GPS coordinates from tractors, seeding rates from planters, pesticide volumes from sprayers, moisture readings from soil probes, and yield estimates from combines. Bayer, Deere, Corteva, and Trimble aggregate this data, increasingly using AI to detect patterns by weaving farmers' information with satellite imagery and weather feeds.
The companies insist farmers own their data. The contracts tell a different story. Click-to-sign software licenses running to ten thousand words grant business partners broad rights to exploit and share anonymized data pools. A farmer in eastern Washington running seventy-five hundred acres of wheat and canola generates a continuous stream of operational intelligence that flows to four corporations whose data platforms he depends on but does not control.
In Nebraska, the Agriculture Data Privacy Act became the first legislation anywhere claiming privacy rights for farm business data — an acknowledgment that existing frameworks offer farmers no protection. There is no HIPAA for agriculture. The data flows upward, the dependency flows downward, and the regulatory framework has not caught up to either.
The Lock-In
The 2026 Farm Bill contains a provision that reimburses farmers ninety percent of the cost of adopting precision agriculture technologies through the Environmental Quality Incentives Program — fifteen percentage points above the normal EQIP cap. The technologies eligible include GPS, yield monitors, data management software, and IoT telematics. The standards governing those technologies would be set not by the USDA but by the technology industry itself.
This is the subsidy structure that makes the trajectory irreversible. Public money accelerates adoption past the point where farmers can choose alternatives. The equipment is financed on multi-year cycles. The data dependencies compound. The agronomic knowledge that preceded precision agriculture — the field-level, embodied, generational knowledge — atrophies because the economic incentives to maintain it have been legislated away.
The pattern is structural, not agricultural. When efficiency technology enters a complex system with subsidy acceleration, four effects converge: per-unit gains are real and measurable, total resource use increases through expanded application, the knowledge base shifts from practitioners to platforms, and public investment locks in the trajectory past the correction point. Precision agriculture demonstrated all four over thirty years. AI is demonstrating the first three in three.
The thirty-year preview is not a prediction. It is a result. The experiment has been run. The data is in. The question is whether anyone building the next efficiency revolution will read it.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)