Competition Law and Future Dominance Theories in Intelligence Economies

1. Introduction

An intelligence economy is an economy in which competitive advantage increasingly depends on the ability to collect, process, predict and act upon information through artificial intelligence, machine learning, foundation models, autonomous systems, cloud computing, algorithms, data networks and computational infrastructure.

Traditional competition law generally asks whether a firm possesses substantial market power in a defined product and geographic market. In intelligence economies, that approach may become insufficient because market power can arise from control over intelligence inputs and technological ecosystems, rather than merely from market share.

The future theory of dominance therefore needs to examine not only who sells the most, but also:

• who controls scarce computational resources;
• who possesses uniquely valuable training data;
• who controls foundation models;
• who controls AI distribution channels;
• who determines interoperability standards;
• who benefits from feedback loops;
• who controls AI agents and autonomous decision systems;
• who can restrict competitors' access to essential AI inputs; and
• who can leverage dominance from one layer of the AI stack into another.

The CMA's foundation-model work specifically identified the need to consider how competition could evolve across foundation-model markets, while the FTC has highlighted possible concerns involving cloud resources, engineering talent, switching costs, exclusivity, bundling and access to sensitive information.

2. Traditional Theory of Dominance

Traditional dominance analysis normally proceeds through four stages:

A. Relevant market

The authority defines the relevant:

1. product/service market;
2. geographic market; and
3. competitive conditions.

B. Market power

Indicators include:

• market share;
• barriers to entry;
• financial strength;
• access to distribution;
• technological advantages;
• customer dependence;
• network effects.

C. Durable dominance

A high market share alone does not necessarily establish unlawful dominance. The important question is whether market power can be maintained or reinforced.

D. Abuse

Examples include:

• exclusionary pricing;
• tying;
• bundling;
• refusal to supply;
• discriminatory access;
• exclusive dealing;
• self-preferencing;
• predatory conduct;
• exploitative conduct.

The intelligence economy complicates each of these stages.

3. Why Intelligence Economies Require New Dominance Theories

3.1 Market share may become less informative

A company may have relatively modest revenues in an emerging AI market while controlling an indispensable input.

For example:

A firm may not dominate the AI-application market but may control a critical cloud infrastructure, GPU ecosystem, operating system, model API or dataset.

Consequently, dominance may exist at an upstream infrastructure layer even where downstream market share is fragmented.

4. The Future "Intelligence Stack" Theory of Dominance

A useful analytical model is to divide the intelligence economy into layers:

Layer 1 — Compute

• GPUs;
• AI accelerators;
• data centres;
• cloud computing;
• electricity and cooling infrastructure.

Layer 2 — Data

• training datasets;
• behavioural data;
• search data;
• transaction data;
• sensor data;
• proprietary enterprise data.

Layer 3 — Models

• foundation models;
• large language models;
• multimodal models;
• specialised AI models.

Layer 4 — Distribution

• operating systems;
• app stores;
• browsers;
• search engines;
• enterprise software;
• cloud platforms.

Layer 5 — Applications

• AI assistants;
• autonomous vehicles;
• financial AI;
• healthcare AI;
• legal AI;
• industrial AI.

Layer 6 — Agents

Future AI systems may independently:

• negotiate contracts;
• purchase products;
• allocate resources;
• select suppliers;
• trade securities;
• optimise logistics;
• interact with other AI agents.

The most important future competition question may therefore be:

Which firm controls the interfaces between these layers?

5. Six Major Future Theories of Dominance

Theory 1 — Compute Dominance

A firm may acquire market power by controlling access to computational capacity.

AI development requires enormous computing resources. If access to sufficiently powerful computing becomes concentrated, control over compute can become a competitive bottleneck.

Potential abusive practices include:

• discriminatory cloud pricing;
• preferential access for affiliated AI companies;
• capacity reservation;
• exclusive cloud arrangements;
• refusal to provide sufficient compute;
• tying AI models to cloud services.

The FTC's investigation into major cloud-AI partnerships specifically examined whether such arrangements could affect access to computing resources, engineering talent and other AI inputs.

Future legal principle

Control over indispensable computational infrastructure may become an important indicator of dominance.

6. Theory 2 — Data Dominance

Data can generate market power through:

• scale;
• uniqueness;
• quality;
• real-time availability;
• historical depth;
• network-generated feedback.

A future dominant firm may possess a data advantage that competitors cannot reproduce even with substantial investment.

Example

A search engine possesses billions of queries.

Those queries improve:

search → user interaction → data → model improvement → better search → more users → more data.

This creates a data-feedback loop.

The competitive concern is not merely the quantity of data but whether rivals can realistically obtain equivalent data.

The European Commission's 2026 DMA measures concerning access to Google Search data illustrate the increasing regulatory importance of data access in AI-related competition.

7. Theory 3 — Algorithmic Dominance

Traditional dominance is based substantially on economic resources.

Intelligence economies introduce another source:

superior algorithmic capability.

A firm may gain an advantage through:

• superior prediction;
• better recommendation algorithms;
• better reinforcement learning;
• superior model architecture;
• faster inference;
• superior optimisation;
• proprietary AI agents.

The difficult question is whether algorithmic superiority constitutes merely competition on the merits or whether it is being reinforced through exclusionary conduct.

8. Theory 4 — Feedback-Loop Dominance

This is likely to become one of the most important future theories.

An AI platform may operate a self-reinforcing cycle:

More users → more data → better model → better service → more users.

The same phenomenon can occur with:

More developers → more applications → more users → more data → better ecosystem → more developers.

This produces a form of dynamic dominance.

Traditional static market-share analysis may underestimate such power because the relevant competitive advantage lies in the firm's ability to continually strengthen its position.

9. Theory 5 — Ecosystem Dominance

A company may not dominate a single market but may control an ecosystem consisting of:

• operating system;
• cloud;
• search;
• browser;
• app store;
• AI assistant;
• advertising;
• payment infrastructure;
• enterprise software.

The concern becomes ecosystem leverage.

A dominant firm in one layer can potentially transfer its power to an emerging AI market through:

• tying;
• bundling;
• default settings;
• self-preferencing;
• interoperability restrictions;
• discriminatory API access;
• exclusive agreements.

The European Commission's Google Android decision provides an important precedent because it examined tying and the leveraging of dominance from one digital layer into another.

10. Theory 6 — Agentic Dominance

The most radical future development concerns AI agents.

Suppose millions of consumers delegate decisions to AI agents.

The agent may decide:

• which search engine to use;
• which products to purchase;
• which bank to select;
• which airline to book;
• which healthcare provider to contact;
• which software to install.

The company controlling the dominant AI agent could therefore become an economic gatekeeper.

Competition would shift from:

controlling consumers

toward:

controlling the artificial intelligence that acts on behalf of consumers.

This could produce a new form of delegated market power.

11. Important Case Laws

Case 1 — United Brands v Commission

United Brands v Commission, Case 27/76

The Court of Justice established important principles concerning dominance under Article 102 TFEU.

Relevance to intelligence economies

The case demonstrates that dominance involves the ability to behave to an appreciable extent independently of competitors, customers and ultimately consumers.

Applied to AI:

A foundation-model provider could potentially possess substantial power where customers cannot realistically discipline it because alternatives are technologically or economically inadequate.

Future significance

The concept of economic independence may become more important than simple market share when analysing AI ecosystems.

12. Case 2 — Commercial Solvents v Commission

Istituto Chemioterapico Italiano and Commercial Solvents v Commission, Joined Cases 6/73 and 7/73

The case concerned the use of dominance in an upstream market to restrict competition downstream.

Relevance

This provides a foundation for the future concept of:

AI-stack leveraging.

For example, an infrastructure provider might use upstream dominance over:

• cloud infrastructure;
• compute;
• AI chips;
• model hosting;

to disadvantage downstream AI competitors.

Principle

Dominance at one level does not give a firm unrestricted freedom to exploit that position in related markets.

13. Case 3 — Bronner v Mediaprint

Oscar Bronner GmbH & Co. KG v Mediaprint, Case C-7/97

This is one of the principal EU precedents concerning refusal to supply and essential facilities.

The Court adopted stringent conditions before requiring a dominant company to provide access to infrastructure.

Relevance to AI

Future disputes may involve:

• AI compute;
• model APIs;
• datasets;
• cloud infrastructure;
• interoperability interfaces;
• AI distribution platforms.

The central question would be whether the input is genuinely indispensable and whether refusal substantially eliminates effective competition.

Future significance

Bronner provides an important caution:

Not every valuable AI resource should automatically be classified as an essential facility.

14. Case 4 — Microsoft

Microsoft Corp. v Commission, Case T-201/04

The General Court upheld major aspects of the Commission's finding concerning Microsoft's conduct involving interoperability information and tying.

Relevance

Microsoft is particularly important for intelligence economies because AI competition is highly dependent on interoperability.

Future disputes may concern:

• interoperability with AI assistants;
• access to APIs;
• operating-system functionality;
• cloud portability;
• AI agent interoperability;
• model switching.

The case therefore supports the proposition that technological control can have competition-law consequences where it is used to restrict rival access or reinforce market power.

15. Case 5 — Google Shopping

Google Search (Shopping), Case AT.39740

The European Commission found that Google had abused dominance by favouring its comparison-shopping service in general search results.

Relevance to intelligence economies

The underlying concept is highly relevant to AI:

self-preferencing by an infrastructure operator can affect downstream competition.

Future equivalents could involve:

• an AI assistant favouring its owner's shopping service;
• an AI search engine prioritising affiliated applications;
• an operating system giving privileged access to its own AI;
• a cloud provider preferentially allocating compute to its own models.

This represents a possible transition from search self-preferencing to AI-agent self-preferencing.

16. Case 6 — Google Android

Google Android, Case AT.40099

The European Commission examined Google's conduct involving Android, search, browser and app-store arrangements.

The Commission concluded that certain tying and contractual practices could restrict competition.

Relevance to AI

The case provides a framework for analysing:

AI + operating system + search + app distribution.

A future AI ecosystem could involve:

OS → default AI assistant → search → applications → advertising → data.

Competition law may therefore examine whether control over one layer is used to protect another.

17. Case 7 — Qualcomm

Qualcomm, Case AT.40220

The Commission's Qualcomm decision concerning exclusivity illustrates the competition-law importance of exclusionary arrangements by a technologically powerful supplier.

Relevance to intelligence economies

Future AI infrastructure markets may contain similar contractual strategies involving:

• exclusive chip supply;
• cloud exclusivity;
• model exclusivity;
• AI accelerator commitments;
• preferred access arrangements.

The critical question would be whether contractual arrangements substantially foreclose competitors rather than merely reflecting legitimate commercial competition.

18. Case 8 — Aspen Skiing

Aspen Skiing Co. v Aspen Highlands Skiing Corp., 472 U.S. 585 (1985)

The U.S. Supreme Court considered a refusal-to-deal theory involving a dominant ski operator.

Relevance

The case is frequently discussed in relation to exceptional circumstances in which a dominant firm abandons a profitable course of dealing in a manner that harms competition.

In an intelligence economy, analogous disputes could arise concerning:

• access to AI infrastructure;
• previously available APIs;
• interoperability;
• data access;
• platform integration.

However, Aspen Skiing should not be interpreted as establishing that every refusal to provide AI resources is unlawful.

19. Case 9 — FTC v Qualcomm

FTC v Qualcomm Inc., 969 F.3d 974 (9th Cir. 2020)

The Ninth Circuit considered allegations concerning Qualcomm's licensing practices and rejected the FTC's principal Sherman Act theory on the record presented.

Importance

This case illustrates the need to distinguish:

• legitimate intellectual-property monetisation;
• technological superiority;
• contractual leverage; and
• exclusionary conduct that violates competition law.

This distinction will be especially important in AI because companies may possess valuable:

• model weights;
• patents;
• training techniques;
• datasets;
• software;
• chip designs.

20. Case 10 — Google LLC v Commission

Google LLC v Commission, Case C-48/22 P

The Court of Justice's treatment of Google's Android-related conduct is significant for understanding how dominance, contractual restrictions and competition in digital ecosystems interact.

Future application

The same analytical framework may be relevant where an AI platform uses:

• defaults;
• contractual restrictions;
• pre-installation;
• ecosystem integration;
• technical restrictions

to protect an AI service against competing AI systems.

21. AI-Specific Regulatory Developments

Although not yet traditional reported dominance judgments, recent AI competition investigations are important evidence of how enforcement thinking is developing.

The FTC's 2025 report on cloud-service-provider/AI-developer partnerships examined Microsoft–OpenAI, Amazon–Anthropic and Google–Anthropic relationships. The FTC highlighted potential effects on access to computing resources, engineering talent, switching costs and commercially sensitive information.

The FTC has also identified possible future competition concerns involving:

• bundling;
• tying;
• exclusive dealing;
• discriminatory access;
• cloud/AI integration;
• API restrictions; and
• leveraging existing dominance into generative AI. 

The CMA's foundation-model review similarly examined how AI markets could develop and what competition principles could preserve contestability.

22. From Market Dominance to "Intelligence Infrastructure Dominance"

A major future development could be a shift from:

Traditional dominance

Market share + barriers + customer dependence

to:

Intelligence dominance

Compute + data + models + distribution + feedback + interoperability + autonomous decision power.

This means a competition authority may need to ask:

23. Dynamic Dominance

One of the most important future concepts is dynamic dominance.

A company might initially possess only moderate market power but have mechanisms that continually increase its advantage.

Example

More users

↓

More behavioural data

↓

Better AI model

↓

Better predictions

↓

More applications

↓

More developers

↓

More users

↓

More data

This creates a self-reinforcing intelligence loop.

Competition authorities may therefore need to consider future competitive effects rather than only present market shares.

24. Contestability as a Dominance Indicator

Future competition law may increasingly focus on contestability.

Important questions include:

1. Can new AI firms obtain compute?
2. Can they obtain training data?
3. Can they access distribution?
4. Can users switch models?
5. Can enterprises migrate between clouds?
6. Can developers change APIs?
7. Can AI agents interact with rival platforms?
8. Can competitors obtain interoperability?
9. Can startups attract AI engineers?
10. Can firms scale without depending on the incumbent?

The FTC and CMA's recent AI work demonstrates that access, switching and ecosystem structure are already central competition concerns.

25. The "AI Essential Facility" Problem

Future litigation may attempt to apply essential-facility principles to:

• foundation models;
• large datasets;
• AI compute;
• AI inference infrastructure;
• model APIs;
• interoperability interfaces.

But courts are likely to face a difficult balancing exercise.

Too broad an approach

Could discourage:

• investment;
• innovation;
• R&D;
• proprietary AI development.

Too narrow an approach

Could permit incumbents to control essential technological infrastructure and eliminate emerging competitors.

The Bronner framework therefore remains highly relevant.

26. AI Dominance and Merger Control

Dominance theories will also affect merger review.

A transaction may appear harmless when measured only by current revenues but potentially eliminate a future competitive constraint.

Examples include acquisitions involving:

• AI startups;
• model developers;
• data companies;
• chip designers;
• cloud providers;
• AI-agent companies.

Authorities may therefore examine:

potential competition + innovation competition + access to critical AI inputs.

The FTC's AI partnership study illustrates why investment structures can matter even before traditional horizontal dominance becomes obvious.

27. New Forms of Abuse

Future intelligence-economy dominance cases may involve:

1. Compute foreclosure

Restricting competitors' access to AI computing.

2. Data foreclosure

Preventing rivals from obtaining competitively important datasets.

3. Model foreclosure

Using a dominant model to exclude competing models.

4. API discrimination

Giving affiliated AI applications better access.

5. AI self-preferencing

Ranking the firm's own AI services above competitors.

6. Agent steering

Configuring AI agents to favour affiliated businesses.

7. AI tying

Making access to one AI service conditional on purchasing another product.

8. Cloud-model tying

Conditioning model access upon use of a particular cloud.

9. Interoperability foreclosure

Preventing competing AI systems from interacting with dominant infrastructure.

10. Predictive exclusion

Using privileged predictive information to anticipate and neutralise emerging competitors.

28. A Proposed Future Dominance Test

A useful analytical framework for intelligence economies would examine seven factors:

1. Control

Does the firm control an important intelligence input?

2. Scarcity

Can competitors realistically reproduce that input?

3. Dependency

Are customers or competitors dependent upon it?

4. Feedback

Does usage generate additional competitive advantages?

5. Switching

Can users and developers switch easily?

6. Interoperability

Can competing systems connect with the incumbent ecosystem?

7. Foreclosure

Can the firm use its position to prevent effective competition?

This framework supplements rather than replaces conventional market-definition and dominance analysis.

29. Future Competition-Law Paradigm

The emerging model can be represented as:

Data
↓
Compute
↓
Foundation Model
↓
AI Platform
↓
Distribution
↓
AI Agents
↓
Consumer/Business Decisions
↓
More Data
↓
Greater Intelligence

The final stage creates a feedback loop that can make dominance increasingly durable.

Thus:

The central competition-law problem of the intelligence economy may not simply be monopoly over a product; it may be control over the infrastructure through which economic decisions themselves are increasingly made.

30. Key Case-Law Principles at a Glance

31. Conclusion

Future dominance theory in intelligence economies will likely move from a purely market-centred conception of power toward an infrastructure-and-ecosystem conception of power.

The decisive competitive resources may increasingly be:

compute + data + models + algorithms + distribution + interoperability + feedback loops + AI agents.

The most significant legal challenge will be distinguishing legitimate technological superiority from strategic foreclosure.

Existing cases such as United Brands, Commercial Solvents, Bronner, Microsoft, Google Shopping, Google Android, Qualcomm and Aspen Skiing provide the doctrinal foundations, but their principles may need to be adapted to AI-specific characteristics.

Recent regulatory work reinforces this direction: the CMA has specifically studied contestability in foundation-model markets, while the FTC has identified potential competition risks arising from cloud/AI partnerships, switching costs, access to compute and technical information.

The future question for competition law is therefore not merely:

"Is this AI company dominant?"

but increasingly:

"Does this company control a critical intelligence bottleneck through which competitors, developers, businesses or consumers must pass—and is that control being used to make the intelligence ecosystem less contestable?"

That question provides a useful foundation for a future doctrine of intelligence-economy dominance.