Data Sovereignty: Who Owns the Farm's Intelligence

What Agricultural Data Actually Is

A Midwest operator in 2015 receives a new iPad with a Climate FieldView subscription pre-loaded. The dealer walked them through the interface during the winter meeting. It tracks everything. GPS coordinates tagged to bushels per acre at 1-to-5 metre resolution. Soil moisture readings across field zones. Nutrient application records from every variable-rate pass. Combine sensor outputs: grain quality, kernel count, loss rate per hour. The platform collects these at every pass of every piece of enrolled equipment. The operator activates the subscription, logs in, and the data starts accumulating in a database they will never directly access. The terms of service scroll past on the screen. The operator does not read them. The data leaves the farm.

Agricultural field data is not incidental. It is a structured intelligence asset. A yield map at 5-metre resolution across a 1,000-acre field holds roughly 80,000 georeferenced data points per harvest season. Stack five years of that onto soil moisture logs, fertility records, and planting population maps and the result is a field-level production model with predictive power over input response, pest pressure, and yield potential. That model has four categories of commercial value: input recommendation, insurance pricing, commodity market positioning, and platform lock-in. The farmer generated every data point. The farmer paid the sensor cost and the subscription fee to collect it. In 2015, few operators tracked where it went after that.

The terms of service were precise about this even then. Climate FieldView's 2023 terms (section 4.2) grant Bayer Crop Science a broad licence to use, copy, process, and transmit field data for platform improvement. The definitions of platform improvement in the document are not operator-auditable. They encompass model training, agronomic research, and product development, categories that include Bayer's own seed and crop-protection recommendations. The operator signed. The operator did not read. The data left the farm as surely as the grain did, and unlike the grain, it left without a price on the scale ticket.

Rent-Extraction Arithmetic: Three Vectors, One Direction

Bayer Crop Science acquired Climate Corporation, the parent company of Climate FieldView, from Monsanto in 2013 for $1.1 billion (Monsanto press release, 2013; Bloomberg, 2013). At the time, the platform had no embedded revenue logic beyond subscription fees. By the time Bayer completed its acquisition of Monsanto in 2018, Climate FieldView had enrolled millions of acres and the integration with Bayer's seed and crop-protection portfolio was direct. John Deere's Operations Center reported 150 million enrolled acres globally in its 2023 Annual Report. Corteva acquired Granular, a farm management and data platform covering millions of additional acres, in 2021. The three platforms together hold field-level data across a substantial share of North American production acreage. The precision agriculture market was valued at $9.5 billion USD in 2023 and is projected to reach $16.4 billion by 2028 (MarketsandMarkets, 2023). That figure includes hardware, software, and data services. The data-monetisation piece is what the sovereignty arithmetic lives inside.

The first extraction vector is input recommendation. FieldView analyses a field's multi-year yield and soil-moisture history and generates agronomic recommendations. Those recommendations draw from Bayer's product portfolio. The farmer pays a platform subscription fee and then receives guidance toward purchasing Bayer's own herbicides, fungicides, and seed. The Environmental Working Group documented this relationship in 2021 reporting on FieldView's data-sharing architecture. The farmer is paying twice: once for the intelligence layer that generates the recommendation, once for the product the recommendation selects. A soil test and an independent agronomist breaks this loop. The platform design does not encourage that path.

The second vector is crop insurance. Field-level yield history telemetry flows from precision-ag platforms into actuarial models. Insurers who access that data can price crop insurance risk at per-field resolution, identifying chronically low-yielding zones and structuring premiums accordingly. The farmer who installed sensors, paid for the subscription, and generated five years of yield maps has now furnished the actuarial input for the premium they pay on those same acres. Reuters (2021) and the Financial Times (2022) both documented commercial relationships between ABCD grain traders (ADM, Bunge, Cargill, Louis Dreyfus) and precision-ag data brokers designed to sharpen pre-harvest yield estimates. When commodity buyers have field-level data on crop progress, the information asymmetry between the buyer and the farmer, already structural given ABCD traders' roughly 70-90 percent combined global grain trade share, widens further.

The third vector is straightforward: platform lock-in. An operator with five years of field data in Deere's Operations Center faces real friction switching to a different data ecosystem. Historical records are not fully portable. The agronomic models built on that data stay with the platform. The switching cost is not contractual but practical. Every additional year of data deposited raises the cost of exit and deepens the platform's analytical advantage over the farmer it is ostensibly serving. Sovereignty names this for what it is: a retention mechanism paid for by the operator's own field intelligence.

The Cooperative Alternative: Three Exits

OpenTEAM (Open Technology Ecosystem for Agricultural Management) launched in 2019 as a farmer-owned interoperable data platform consortium. Its founding member organisations include the Rodale Institute, Practical Farmers of Iowa, and the National Young Farmers Coalition. Governance rests with farmer-member organisations. Data is not pooled into a centralised corporate repository. Each operator determines what they share, with whom, and under what terms. The distinction from a proprietary platform is not cosmetic. An operator using OpenTEAM who chooses to share their yield data with a university research programme retains the right to revoke that sharing next season. An operator whose yield data sits inside FieldView cannot reach backward into Bayer's training datasets and extract what was contributed. The structure is not equivalent, and the asymmetry is not an accident.

Our Sci (ourcientist.org) operates on a community-science model: an open data platform for food and agriculture research where contributors set their own data governance terms before committing any records. The approach routes around the principal-agent problem that proprietises precision-ag data. On a commercial platform, the platform is the principal and the farmer is the data supplier. On Our Sci, the farmer is the principal and the research community is the conditional recipient. The data flows in the same direction. The authority does not.

The AgStack Foundation, a Linux Foundation project launched in 2021, works at the infrastructure layer rather than the farm-management application layer. AgStack builds open-source agricultural data infrastructure: commons-based field-data protocols, geospatial reference layers, and interoperability standards that no single corporation owns. The logic mirrors the open-source software movement's argument against proprietary operating systems: if the infrastructure is owned, every application built on it inherits the ownership constraint. AgStack aims to establish a field-data commons so that no future FieldView or Operations Center starts from a proprietary lock-in position at the foundation. The commons-layer approach does not preclude commercial services built on top. It prevents those services from capturing the substrate.

The asymmetry of the proprietary platforms is not a technical constraint imposed by the complexity of field-level data. It is a commercial choice by the platform owner. OpenTEAM, Our Sci, and AgStack all demonstrate that the technical substrate supports farmer-owned governance. The extraction runs through the terms of service, not the sensor. Changing the terms is an institutional problem, and farmer-owned institutions are the institutional answer.

The Equipment Junction: One Fence, Two Layers

Equipment telemetry is the primary feedstock for precision-ag platforms. John Deere's Operations Center is architecturally downstream of the Deere tractor's CAN bus, GPS unit, yield sensor, and combine grain-quality analyser. The machine collects. The Operations Center aggregates. The platform monetises. The chain is vertical, and it lives inside a single corporate family. Deere manufactures the equipment, controls the diagnostic port, operates the data platform, and holds the field-intelligence asset that results.

This structural fact means the equipment-sovereignty argument and the data-sovereignty argument address the same fence at two different layers. Equipment sovereignty addresses the ECU fence: the proprietary controller area network that locks third-party repair, makes independent diagnostics illegal under the Digital Millennium Copyright Act, and keeps the machine's operational data inside the manufacturer's ecosystem. Data sovereignty addresses what happens to that operational data once it reaches the platform. Breaking equipment sovereignty without breaking data sovereignty leaves the telemetry stream intact inside a proprietary aggregation system. Breaking data sovereignty without breaking equipment sovereignty leaves the physical repair and diagnostic lock in place. The two spokes address the same extraction; the resolution requires both.

The implication for an operator navigating the sovereignty stack is practical. Choosing an open-repair-compatible tractor (Kubota, AGCO, or a pre-2010 Deere pre-dating the closed-ECU architecture) reduces the equipment-sovereignty exposure. Pairing it with an OpenTEAM-compatible data management workflow and a consent-model approach to any data sharing closes the data-sovereignty gap. Neither choice requires abandoning precision agriculture. Both choices require a practitioner who understands that the data infrastructure and the physical machine are the same sovereignty problem expressed at different resolution.

The Architecture Is the Argument

The precision agriculture market will reach $16.4 billion by 2028 (MarketsandMarkets, 2023). Most of that growth is hardware and software sold to farmers to collect better field data. The data is genuinely useful. Yield mapping at 5-metre resolution identifies low-performing zones that justify input reduction and management change. Soil-moisture telemetry informs irrigation timing that reduces water expenditure and improves crop quality. Variable-rate application of fertiliser guided by field-history models cuts input volumes and cost per bushel. The technology argument for precision agriculture is sound. The ownership architecture around it is the problem.

The Midwest operator who activated that Climate FieldView subscription in 2015 was not irrational. They were solving a real agronomic problem with a genuinely useful tool. The extraction was in the terms of service they did not read. The data platform collected field intelligence through sensor cost the farmer paid, subscription fees the farmer paid, and equipment the farmer bought. That intelligence then flowed through three monetisation channels: input recommendations biased toward the platform owner's product portfolio, actuarial data for the insurance premiums the farmer paid, and market intelligence that sharpened commodity buyers' position at the elevator. Every vector ran in one direction. The farmer's sovereignty over their own field intelligence was the resource being mined.

The exit is farmer-governed data infrastructure. OpenTEAM offers cooperative governance. Our Sci offers operator-controlled consent. AgStack offers commons-based infrastructure that prevents proprietary capture at the foundation. None of these is as polished as the incumbents' platforms. That is a feature gap, not a structural one, and feature gaps close. Governance structures do not close once they are built into a terms of service and accepted by a hundred million enrolled acres.

The data the farmer generates to improve their own operation is currently the raw material for someone else's competitive advantage. This is not a coincidence. It is the architecture. Data you pay to generate is data sold back to you.