Agricultural AI Platform Dominance and Input Control

1. Introduction

Agricultural AI platform dominance and input control refers to a situation where an AI-driven agricultural platform becomes sufficiently powerful that it can influence or control important inputs needed by farmers, competitors, or agricultural-service providers.

An agricultural AI platform may combine:

farm and soil data;

satellite and drone imagery;

weather data;

machinery data;

seed and crop information;

fertilizer and pesticide recommendations;

irrigation decisions;

agricultural credit or insurance information;

farm-management software;

machinery and repair systems;

input purchasing;

distribution and marketplace services.

This creates a possible vertical integration problem. A platform may simultaneously operate the digital decision-making system and have commercial interests in the seeds, chemicals, machinery, fertilizers, financing, or other inputs recommended through that system.

The European Commission has specifically recognized that digital agriculture can exhibit first-mover advantages, network effects, data advantages and algorithmic advantages, including in digitally enabled agricultural prescriptions. (European Commission)

Importantly, agricultural AI dominance is not automatically unlawful. Article 102 TFEU prohibits abusive conduct by a dominant undertaking; simply becoming successful or dominant is not itself an infringement. (Competition Policy)

2. Basic Meaning

A simple model is:

Farm Data → AI Platform → Recommendation → Agricultural Input → Farmer Dependency

For example:

An AI platform collects millions of hectares of farm data, recommends a particular fertilizer or seed, provides machinery software, controls access to farm-management data, and makes it difficult for farmers to use competing suppliers.

The competition-law concern becomes stronger if the platform has:

substantial market power;

control over an important input or infrastructure;

high switching costs;

network effects;

exclusive arrangements;

self-preferencing;

discriminatory access;

tying or bundling;

refusal to provide necessary data/interoperability;

discriminatory AI recommendations.

3. Why Agriculture Is Particularly Sensitive

Agricultural production contains several interconnected markets.

Traditional agricultural chain

Seeds → Fertilizer → Pesticides → Machinery → Farming → Storage → Processing → Distribution

AI-enabled agricultural chain

Farm Data → AI Model → Prediction → Recommendation → Input Purchase → Machinery Operation → Crop Production

A single platform could therefore influence several stages simultaneously.

The European Commission states that EU competition rules apply throughout the food supply chain, while agricultural products also receive specific treatment under the CAP framework and Article 42 TFEU. (Competition Policy)

4. What Is "Input Control"?

Input control means having significant control over something necessary for agricultural production or for competing in an agricultural service market.

Examples include:

Physical inputs

seeds;

fertilizer;

pesticides;

machinery;

irrigation equipment.

Digital inputs

farm data;

machine-generated data;

APIs;

AI models;

software;

digital prescriptions;

satellite information;

agronomic databases.

Infrastructure

cloud systems;

agricultural marketplaces;

machinery operating systems;

authentication systems;

diagnostic software.

Commercial inputs

financing;

insurance;

distribution;

logistics;

agricultural procurement.

The EU Joint Statement on competition in generative AI identifies concentrated control over important inputs such as data, computing resources and technical capabilities as a potential source of bottleneck power. The same economic reasoning can be relevant when examining AI-intensive agricultural ecosystems. (Competition Policy)

5. Relevant Markets

An agricultural AI investigation may involve several relevant markets rather than one giant "agriculture" market.

Possible markets include:

agricultural AI software;

precision-agriculture services;

farm-management platforms;

digital agricultural prescriptions;

agricultural data services;

agricultural machinery software;

agricultural inputs;

seeds;

fertilizers;

pesticides;

agricultural marketplaces;

cloud/AI infrastructure.

The Commission normally begins an Article 102 investigation by defining the relevant product and geographic markets and then assessing dominance using market shares together with entry barriers, buyer power, resources and vertical integration. (Competition Policy)

6. How AI Can Create Agricultural Market Power

A. Data network effects

More farmers → more data → better AI → more farmers → still more data.

This can create a feedback loop:

Users ↑ → Data ↑ → Model quality ↑ → Attractiveness ↑ → Users ↑

The Commission's investigation into digital agriculture involving Bayer/Monsanto identified both first-mover advantages and network effects, with data and algorithms being important competitive factors. (European Commission)

B. First-mover advantage

A platform that enters early may accumulate:

farmer relationships;

historical crop data;

machinery connections;

agronomic models;

distribution networks;

supplier contracts.

A later entrant may find it difficult to reproduce these advantages.

C. Switching costs

Farmers may become dependent upon:

proprietary farm-management software;

stored historical data;

machinery compatibility;

customized AI models;

digital prescriptions;

subscription systems.

Changing platforms could require significant technical and financial costs.

7. Input Foreclosure

One major theory is input foreclosure.

Suppose an AI platform controls a critical agricultural input and also competes in downstream agricultural services.

It might allegedly:

refuse to supply the input to competitors;

increase the input price for competitors;

reduce quality;

delay access;

provide better terms to its own downstream business;

restrict API access;

restrict data portability.

The Commission recognizes refusal to supply an indispensable input as one possible form of Article 102 abuse, subject to the applicable legal conditions. (Competition Policy)

8. Customer Foreclosure

The reverse problem can also occur.

The AI platform may control farmers' access to suppliers.

For example:

"Farmers using our AI platform must purchase recommended fertilizer only from our affiliated supplier."

Potential mechanisms include:

exclusivity;

loyalty discounts;

bundled subscriptions;

contractual restrictions;

preferred recommendations;

algorithmic ranking.

This could potentially make competing input suppliers unable to reach enough farmers.

9. Algorithmic Self-Preferencing

This is particularly important.

Suppose an AI platform recommends agricultural products.

It ranks:

Platform-owned fertilizer: 95% suitability

Independent fertilizer: 70% suitability

If the ranking is genuinely based on agronomic performance, there may be a legitimate explanation.

But if the platform artificially modifies the algorithm to favor its own products, the competition concern becomes stronger.

The legal analysis would examine:

dominance;

discriminatory treatment;

ranking rules;

actual or potential foreclosure;

objective justification;

effects on competitors;

effects on farmers.

The Google Shopping judgment provides an important modern Article 102 framework for analyzing exclusionary conduct involving a dominant digital platform and preferential treatment of its own services. Google and Alphabet v Commission, C-48/22 P (2024) is therefore highly relevant by analogy, although it was not an agricultural case.

10. Six+ Important Case Laws

Important note: There are currently relatively few EU court judgments directly concerning an AI agricultural platform controlling agricultural inputs. Therefore, the cases below are principally analogical competition-law authorities. The cases establish principles that could be applied to such an agricultural AI dispute.

Case 1: Bronner v Mediaprint

Case: Oscar Bronner GmbH & Co. KG v Mediaprint Zeitungs und Zeitschriftenverlag GmbH & Co. KG, C-7/97
Court: CJEU
Date: 26 November 1998

Facts

Bronner sought access to Mediaprint's newspaper home-delivery system.

The question was whether a dominant undertaking could be required to provide access to infrastructure it had developed for itself.

Principle

The Court established strict conditions for treating infrastructure as indispensable.

The existence of alternative solutions must be considered, including whether technical, legal or economic obstacles make alternatives impossible or unreasonably difficult. (curia)

Agricultural AI relevance

Imagine an agricultural AI company controls the only realistically usable:

farm-data infrastructure;

machine-data platform;

agricultural API;

digital prescription infrastructure.

A competitor seeking access could potentially invoke refusal-to-supply principles.

But merely saying "the platform is useful" would not automatically satisfy the strict indispensability requirements.

Case 2: IMS Health v NDC Health

Case: IMS Health GmbH & Co. OHG v NDC Health GmbH & Co. KG, C-418/01
Court: CJEU
Date: 29 April 2004

Facts

IMS Health controlled a pharmaceutical sales-data structure used by companies to organize regional pharmaceutical sales information.

A competitor sought access.

Principle

The Court considered when refusal to license/access a protected system could constitute abuse of dominance.

The case developed the exceptional circumstances framework for compulsory access to an indispensable resource. (Infocuria)

Agricultural AI relevance

Suppose an agricultural AI platform controls a unique data architecture containing:

historical crop data;

farm-level production information;

machine data;

agricultural prescription information.

If competitors cannot realistically reproduce the system, an IMS Health-type analysis could become relevant.

However:

Data valuable ≠ automatically indispensable.

The legal test remains demanding.

Case 3: Microsoft v Commission

Case: Microsoft Corp. v Commission, T-201/04
Court: General Court
Date: 17 September 2007

Facts

Microsoft was found to have abused its dominant position in connection with interoperability information and its relationship with competing work-group server products.

Principle

The case is a major authority concerning refusal to provide information necessary for effective interoperability.

Agricultural AI relevance

Consider an agricultural platform that controls:

tractor operating data;

sensor interfaces;

API access;

machinery diagnostics;

farm-management software.

If competing agricultural applications cannot effectively operate because the dominant platform refuses necessary interoperability information, Microsoft provides an important analogy.

The key issue would be whether the conduct substantially restricts competition and satisfies the relevant conditions for an abuse.

Case 4: Intel v Commission

Case: Intel Corp. v Commission, C-413/14 P
Court: CJEU
Date: 6 September 2017

Facts

Intel's rebate arrangements were examined under Article 102 TFEU.

Principle

The judgment is important for the assessment of exclusionary effects associated with rebates and pricing practices.

The economic effects of the conduct can matter significantly in determining whether competition is actually capable of being foreclosed.

Agricultural AI relevance

Suppose an agricultural AI platform says:

"Farmers using our AI subscription receive a 30% discount on our affiliated fertilizer."

The economic analysis might need to examine:

duration;

scale;

customer coverage;

competitor ability to compete;

effective price;

foreclosure;

efficiencies.

An AI recommendation combined with financial incentives could therefore create a potentially complex algorithm + rebate + input-control structure.

Case 5: Google Shopping

Case: Google and Alphabet v Commission (Google Shopping), C-48/22 P
Court: CJEU
Date: 10 September 2024

Facts

The case concerned Google's conduct in comparison-shopping services and the treatment of competing services within Google's search ecosystem.

Principle

The case is a major modern authority concerning exclusionary conduct by a dominant digital platform and the competitive significance of preferential treatment within a platform ecosystem.

Agricultural AI relevance

This is highly relevant by analogy to:

Agricultural AI platform → agricultural recommendations → competing inputs

For example, an AI platform might:

rank its own seeds first;

rank its own fertilizers first;

suppress competitors;

give affiliated products greater visibility;

manipulate recommendation scores.

The relevant question would not simply be:

"Did the platform prefer its own product?"

It would require a broader Article 102 analysis of dominance, conduct, competitive effects and possible justification.

Case 6: Deutsche Telekom v Commission

Case: Deutsche Telekom AG v Commission, C-280/08 P
Court: CJEU
Date: 14 October 2010

Facts

The case concerned pricing relationships between upstream and downstream services and the possibility of exclusion through a margin squeeze.

Principle

A vertically integrated dominant undertaking can potentially use pricing structures between different levels of a supply chain in a manner that disadvantages competitors.

Agricultural AI relevance

Imagine:

AI platform → agricultural data/infrastructure → agricultural service

The platform could theoretically charge independent agricultural-service providers high upstream access fees while competing downstream itself.

This could produce a digital agricultural margin-squeeze theory.

Case 7: Post Danmark

Case: Post Danmark A/S v Konkurrencerådet, C-209/10
Court: CJEU
Date: 27 March 2012

Principle

The case concerns exclusionary conduct and the assessment of pricing practices by a dominant undertaking.

Agricultural AI relevance

An agricultural platform might offer apparently cheap or even below-cost AI services while recovering its costs through:

affiliated fertilizer;

seed sales;

machinery;

financing;

data monetisation.

The analysis could therefore need to consider the overall economic structure rather than examining the AI subscription price in isolation.

Case 8: Hugin

Case: Hugin Kassaregister AB v Commission, Case 22/78
Court: CJEU
Date: 31 May 1979

Facts

The case involved cash registers and spare parts.

Principle

The case is an important historical authority concerning market power in an aftermarket involving equipment and replacement parts.

Agricultural AI relevance

This is especially useful for:

AI-enabled agricultural machinery + proprietary software + spare parts + diagnostics.

For example:

Tractor → proprietary operating system → AI diagnostics → authorized parts → authorized maintenance.

A company may face competition-law scrutiny if control of the installed base is used to restrict competition in related aftermarkets.

11. Special Problem: AI Recommendations and Input Bundling

A particularly complex scenario would be:

AI subscription + seeds + fertilizer + pesticide + machinery

The platform might tell farmers:

"To receive the full AI optimization service, purchase our approved agricultural inputs."

This can raise questions concerning:

Tying

The AI service is tied to another agricultural product.

Bundling

Several products are offered together.

Exclusivity

Farmers are required or strongly incentivized to use affiliated inputs.

Self-preferencing

The platform's own inputs receive superior recommendations.

Data advantage

Purchasing affiliated inputs generates additional data that improves the platform.

This produces a potentially powerful feedback loop:

More farmers → more input sales → more data → better AI → stronger recommendations → more farmers.

12. Data as a Strategic Input

Agricultural data can include:

soil characteristics;

yield;

pesticide use;

fertilizer use;

weather;

machine telemetry;

planting dates;

crop diseases;

irrigation;

harvesting patterns.

The European Commission's Joint Research Centre has specifically studied competition problems associated with non-personal agricultural machine data, including the possibility that data-driven business models can lock farm data into machines and affect downstream agricultural-service competition. (JRC Publications)

Therefore, data control can become input control.

13. Farmer Lock-In

A farmer may become locked into a platform because the platform stores:

five years of farm history;

machine configurations;

field maps;

AI-generated prescriptions;

customized models;

yield predictions.

If transferring all this information to a competitor is difficult, switching costs increase.

The competition question is:

Are these switching costs simply the result of legitimate investment and innovation, or are they being deliberately reinforced through exclusionary conduct?

That distinction is important.

14. Interoperability Problem

Agricultural AI ecosystems may involve:

Tractor + drone + sensor + satellite + AI platform + farm software + input marketplace

If each component uses a different system, interoperability becomes critical.

A dominant platform could potentially:

restrict API access;

prevent third-party applications;

limit machine compatibility;

deny data portability;

degrade competing services.

This creates a possible competition problem alongside technical and regulatory issues.

15. Input Discrimination

Suppose an AI platform supplies data to two fertilizer manufacturers.

Company A is affiliated with the platform.

Company B is independent.

The platform gives:

This could raise an Article 102 concern if dominance and exclusionary effects are established.

16. Refusal to Share Agricultural Data

A competitor might argue:

"Without access to the platform's agricultural data, my AI cannot compete."

That does not automatically create a legal right to access the data.

Bronner and IMS Health demonstrate that compulsory access/refusal-to-supply theories have demanding conditions. (curia)

The investigation would need to examine:

Is the undertaking dominant?

Is the resource genuinely indispensable?

Are alternatives realistically available?

Is competition being eliminated?

Is there objective justification?

Would access be technically feasible?

What would be the effect on innovation?

17. Objective Justifications

An agricultural AI platform might have legitimate reasons for restricting access.

For example:

cybersecurity;

protection of farmer privacy;

protection of trade secrets;

agricultural safety;

prevention of inaccurate recommendations;

protection of proprietary algorithms;

prevention of manipulation;

regulatory compliance.

Therefore:

Restriction ≠ automatically abuse.

The authority would need to consider whether the justification is genuine, appropriate and proportionate under the applicable legal framework.

18. Agricultural AI Merger Concerns

Competition problems may arise before an abuse occurs.

Imagine:

Large agricultural-input company + leading agricultural AI platform

The merger could combine:

seeds;

chemicals;

machinery;

farm data;

AI;

distribution.

The Commission's Bayer/Monsanto investigation is particularly relevant because the Commission examined digital agriculture and digitally enabled prescriptions and identified first-mover advantages and network effects. (European Commission)

A merger assessment could therefore examine:

data accumulation;

network effects;

innovation;

vertical foreclosure;

access to agricultural inputs;

access to farmers;

interoperability;

future competition.

19. Competition Law vs Agricultural Regulation

Agricultural AI does not exist only under competition law.

Potentially relevant frameworks include:

Competition law

Article 101 TFEU;

Article 102 TFEU;

EU merger control.

Agricultural law

Common Agricultural Policy;

Common Market Organisation Regulation;

agricultural producer cooperation rules.

Digital regulation

Digital Markets Act;

AI-related legislation;

data/interoperability rules.

Data law

GDPR where personal data are involved;

rules concerning access and use of industrial/non-personal data.

The European Commission expressly recognizes special competition treatment for agricultural products under Article 42 TFEU and the Common Agricultural Policy framework. (Competition Policy)

20. Possible Article 102 Theories

An agricultural AI platform could theoretically face scrutiny for:

1. Refusal to supply

Refusing access to indispensable agricultural data or infrastructure.

2. Discriminatory access

Giving affiliated input suppliers better access.

3. Self-preferencing

Giving the platform's own agricultural inputs preferential AI recommendations.

4. Tying

Requiring use of affiliated inputs to obtain AI services.

5. Bundling

Combining AI, machinery, seeds and fertilizers.

6. Exclusive purchasing

Requiring farmers to purchase inputs exclusively through the platform.

7. Predatory pricing

Using AI services or input pricing to eliminate competitors.

8. Margin squeeze

Charging competitors excessive upstream access prices while competing downstream.

9. Excessive pricing

Potentially relevant in exceptional circumstances where the legal requirements are satisfied.

The Commission identifies exclusive purchasing, predatory pricing and refusal to supply indispensable inputs among possible Article 102 theories. (Competition Policy)

21. Article 101 Problems

The platform itself may not be the only concern.

Suppose an agricultural AI provider and input manufacturers exchange:

farmer-level demand forecasts;

expected prices;

crop forecasts;

supply plans;

future input prices.

If competitors use the same AI system to coordinate their behaviour, Article 101 issues may arise.

The central distinction is:

AI-generated parallel behaviour ≠ automatically an unlawful agreement.

Authorities must establish the necessary elements of an agreement, decision or concerted practice.

22. Economic Evidence

An agricultural AI investigation could require:

market-share data;

farmer switching data;

input prices;

recommendation logs;

AI training data;

API access records;

algorithmic ranking records;

contracts;

exclusivity clauses;

rebate structures;

internal communications;

customer complaints;

competitor foreclosure evidence;

counterfactual simulations.

AI systems create an additional problem:

The important evidence may be contained inside the model itself.

Therefore, investigators may need to examine:

training datasets;

model objectives;

recommendation functions;

ranking variables;

optimization criteria;

changes in model outputs;

human overrides.

23. Remedies

If unlawful conduct is established, possible remedies may include:

Structural remedies

divestiture;

separation of businesses.

Behavioural remedies

non-discriminatory access;

interoperability;

data portability;

prohibition of exclusivity;

non-discrimination requirements.

Technical remedies

API access;

standardized interfaces;

data export;

interoperability protocols.

Commercial remedies

removal of tying;

changes to rebates;

transparent input-ranking rules.

The appropriate remedy depends on the specific infringement and evidence.

24. Important Distinction: Innovation vs Exclusion

A successful agricultural AI company may legitimately gain an advantage because it has:

better algorithms;

better data;

better engineering;

lower costs;

more accurate predictions;

superior farmer service.

Competition law does not require successful firms to eliminate every competitive advantage.

The concern arises where market power is maintained or extended through conduct that is outside competition on the merits.

The Commission's current Article 102 framework emphasizes exclusionary conduct and is based on Union Court jurisprudence and Commission enforcement experience. (Competition Policy)

25. Hypothetical Example

Suppose AgriAI has:

70% of the digital farm-management market;

access to data from 5 million hectares;

ownership of a fertilizer company;

control of a machinery API;

an agricultural marketplace.

Its AI recommends fertilizer.

The platform gives its own fertilizer:

"Optimal — 98%"

while independent suppliers receive lower rankings.

At the same time, independent suppliers must pay high fees for API access.

Farmers cannot easily export their historical farm data.

Possible competition questions

Market 1: agricultural AI services
Market 2: fertilizer
Market 3: agricultural data services
Market 4: machinery software

Possible theories:

self-preferencing;

tying;

discriminatory access;

refusal to supply;

foreclosure;

interoperability restrictions;

data-related switching costs.

But an authority would still have to establish the relevant legal elements rather than assuming that the conduct is unlawful merely from the hypothetical facts.

26. Case-Law Comparison Table

27. Key Legal Formula

Agricultural AI Platform Dominance

Farm Data + Network Effects + AI Advantage + First-Mover Advantage + Switching Costs + Vertical Integration

↓

Potential Platform Market Power

↓

Control Over Agricultural Inputs/Data/Infrastructure

↓

Self-Preferencing / Tying / Exclusivity / Refusal / Discrimination / Interoperability Restrictions

↓

Potential Foreclosure of Competitors

↓

Article 102 TFEU Analysis

↓

Dominance + Exclusionary Conduct + Competitive Effects + Objective Justification

28. Important Exam Points

Remember these 10 keywords:

AI platform

Farm data

Network effects

First-mover advantage

Input control

Vertical integration

Self-preferencing

Interoperability

Foreclosure

Article 102 TFEU

29. Six Most Important Cases for Revision

If you need only six cases for an exam answer:

Bronner – C-7/97 → indispensability

IMS Health – C-418/01 → access to critical systems

Microsoft – T-201/04 → interoperability

Intel – C-413/14 P → exclusionary rebates/effects

Google Shopping – C-48/22 P → digital self-preferencing/platform conduct

Deutsche Telekom – C-280/08 P → vertical foreclosure/margin squeeze

Best way to remember:
Bronner = Access → IMS = Data/System → Microsoft = Interoperability → Intel = Rebates → Google = Platform → Deutsche Telekom = Vertical Pricing.

Conclusion

Agricultural AI platforms can create a new form of competition concern because data, algorithms, machinery, agricultural inputs and distribution can become integrated into one ecosystem. The strongest legal issues arise when a platform with substantial market power uses control over data or digital infrastructure to restrict competing agricultural inputs or services.

The central legal distinction is:

Agricultural AI dominance itself is not unlawful; the competition-law concern is the use of dominance to foreclose competitors or distort competition through control of indispensable inputs, data, access, recommendations or distribution.

The European Commission's digital-agriculture work already identifies data, algorithms, first-mover advantages and network effects as important competitive characteristics of the sector. (European Commission)

And the key doctrinal safeguard remains: AI recommendation, data control or vertical integration alone does not establish an Article 102 infringement; dominance, abusive conduct and the relevant competitive effects must be established on the evidence. (Competition Policy)