Competition Law and Antitrust Implications of Artificial General Intelligence Markets

1. Introduction

Artificial General Intelligence (AGI) refers to AI systems intended to perform a broad range of cognitive tasks across domains rather than being limited to a particular application. Although a fully developed AGI market has not yet emerged in a legally settled form, competition-law analysis can already be applied to the markets and infrastructure that may constitute the AGI value chain: advanced computing, chips, cloud infrastructure, training data, foundation models, model-development tools, AI agents, applications, distribution platforms and enterprise deployment.

Competition authorities have increasingly identified AI as an area where existing antitrust principles may become particularly important. In 2024, the U.S. Department of Justice, Federal Trade Commission, European Commission and UK Competition and Markets Authority jointly stated that competition concerns can arise in generative-AI foundation models and AI products. (Department of Justice) The OECD's 2026 assessment similarly identifies continuing structural risks in several AI layers, particularly foundation models, hardware and data. (OECD)

The central competition-law question is therefore not simply whether AGI will become powerful, but whether control over indispensable inputs, infrastructure, models or distribution channels could allow a small number of firms to restrict competition at other levels of the AI ecosystem.

2. Structure of an AGI Market

An AGI ecosystem can potentially be divided into several interconnected markets.

A. Hardware market

This includes:

GPUs and AI accelerators;

high-bandwidth memory;

networking equipment;

specialized AI processors;

data-centre infrastructure;

semiconductor manufacturing capacity.

Control over scarce computational hardware may create an upstream bottleneck.

B. Cloud-computing market

AGI development requires enormous computing resources. Major cloud providers can therefore function simultaneously as:

infrastructure suppliers;

investors in AI developers;

distributors of AI models; and

competitors to those same AI developers.

The FTC has specifically examined these relationships. Its 2025 study of Microsoft–OpenAI, Amazon–Anthropic and Alphabet–Anthropic identified potential concerns involving access to computing resources, engineering talent, switching costs and access to commercially sensitive information. (Federal Trade Commission)

C. Data markets

Advanced AI requires enormous datasets, potentially including:

books;

news;

scientific publications;

software;

images;

video;

consumer data;

proprietary enterprise data;

search-query information.

Control over high-quality or unique datasets can therefore constitute a significant competitive advantage.

D. Foundation-model market

This includes general-purpose models capable of supporting numerous downstream applications.

Potential competition concerns include:

model access restrictions;

exclusive licensing;

API restrictions;

preferential access to computing;

acquisitions of emerging competitors;

tying models to cloud infrastructure;

discriminatory model access.

E. AI-agent and application markets

AGI-like systems could perform:

research;

coding;

legal analysis;

financial services;

procurement;

logistics;

customer service;

healthcare administration;

autonomous commercial transactions.

The provider of the underlying AGI could therefore compete with companies whose services depend upon the AGI.

F. Distribution markets

Distribution could occur through:

operating systems;

search engines;

app stores;

browsers;

enterprise software;

smartphones;

cloud platforms.

This creates an important vertical-integration problem: a company may control both the AGI product and the route through which competing AGI products reach consumers.

3. Relevant-Market Definition

Traditional competition law normally begins by determining the relevant product and geographic market.

AGI creates unusual difficulties because the technology may simultaneously constitute:

a standalone product;

an input;

an infrastructure service;

a platform;

an application;

a distribution mechanism.

Possible markets could include:

Upstream

AI accelerators → cloud computing → training infrastructure → data

Middle layer

foundation models → model APIs → AI agents

Downstream

enterprise AI → consumer AI → specialized applications

A competition authority may therefore need to examine multiple interconnected markets rather than assuming that “AI” constitutes one market.

4. Market Power in AGI

Market power may arise from factors other than conventional market share.

Important indicators include:

4.1 Computational scale

If only a small number of firms can obtain sufficient computing capacity, computational access can become an entry barrier.

4.2 Data advantages

A firm possessing unique datasets may have an advantage that competitors cannot readily reproduce.

4.3 Network effects

More users can produce:

more interaction data;

better feedback;

greater developer participation;

more applications;

greater ecosystem value.

This can create positive feedback loops.

4.4 Switching costs

Enterprise customers may become dependent upon:

proprietary APIs;

model-specific applications;

proprietary embeddings;

fine-tuning;

cloud architecture;

data formats.

The FTC has specifically identified switching costs as a competition concern in major AI/cloud partnerships. (Federal Trade Commission)

4.5 Ecosystem integration

A firm controlling cloud infrastructure, an operating system, search, productivity software and an AI model may have opportunities to extend power across markets.

5. Exclusive AI Partnerships

One of the most important competition issues concerns agreements between AI developers and cloud providers.

An agreement may provide:

exclusive cloud hosting;

preferential computing capacity;

exclusive distribution;

equity participation;

revenue-sharing;

preferential model access;

restrictions on using competing infrastructure.

The FTC's AI partnership study specifically found that major cloud/AI relationships contained investment, revenue-sharing, consultation, control and exclusivity provisions and highlighted their possible competitive effects. (Federal Trade Commission)

Such arrangements are not automatically unlawful. Their legality would depend upon factors such as:

duration;

market power;

foreclosure percentage;

availability of alternatives;

efficiency justifications;

effect on entry;

effect on innovation.

6. Vertical Foreclosure

Vertical foreclosure could occur if a company controlling an upstream AGI input prevents rivals from obtaining access.

For example:

dominant cloud provider → restricts computing access → rival AGI developer → cannot scale model → downstream competition reduced.

Similarly:

dominant operating system → preferentially integrates its own AI assistant → competing assistants receive inferior functionality → competitors lose users.

The European Commission's 2026 DMA measures concerning Google illustrate the relevance of this problem: the Commission directed attention to ensuring competing AI services could access Android functionality on comparable terms. (Digital Markets Act (DMA))

7. Refusal to Supply Computing Resources

Suppose a dominant infrastructure provider refuses to provide sufficient computational capacity to an AGI competitor.

The relevant legal theories could include:

refusal to deal;

essential-facilities-type reasoning;

exclusionary conduct;

discriminatory access;

constructive refusal to supply.

However, competition law generally does not impose a universal duty on dominant firms to deal with competitors. The precise legal test differs among jurisdictions.

The classic U.S. Supreme Court decisions Verizon Communications Inc. v. Law Offices of Curtis V. Trinko, LLP and Aspen Skiing Co. v. Aspen Highlands Skiing Corp. are therefore particularly relevant.

8. Self-Preferencing by AGI Platforms

A vertically integrated company might operate:

a dominant platform;

an AGI model;

competing applications.

It could then give its own AGI preferential treatment through:

default placement;

superior API access;

better system permissions;

privileged data;

lower fees;

faster computing;

exclusive features.

This resembles the competition concerns examined in digital-platform cases involving Google and other vertically integrated firms.

The European Commission's 2026 Android/AI interoperability measures demonstrate that access to operating-system functionality can become an important competitive issue for AI services. (Digital Markets Act (DMA))

9. Tying and Bundling

AGI could be bundled with:

cloud services;

productivity software;

operating systems;

search;

browsers;

smartphones;

enterprise software.

For example:

Enterprise software + mandatory proprietary AGI

or:

Cloud computing + exclusive proprietary AI model.

The central competition-law question would be whether the conduct improperly leverages power in one market into another.

The leading cases include:

United States v. Microsoft Corp. (2001)

Microsoft's integration and contractual restrictions concerning Internet Explorer provided an important precedent concerning technological tying, exclusionary conduct and leveraging of platform power.

European Commission v. Google (Android)

The EU's Android case provides another important precedent concerning tying and leveraging involving a dominant digital ecosystem.

These principles could become relevant if an AGI provider uses control over a dominant platform to disadvantage rival AI systems.

10. AGI and Exclusive Data

Data exclusivity could become particularly important.

Suppose a dominant firm obtains exclusive access to:

scientific databases;

major publishing archives;

proprietary consumer datasets;

search-query data;

enterprise information.

If those datasets are indispensable to effective competition, exclusive arrangements could potentially produce foreclosure.

The analysis would consider:

whether substitutes exist;

whether the data are genuinely unique;

whether competitors can reproduce them;

whether access is technically feasible;

whether exclusivity produces efficiencies.

11. AI Training Data and Competition

Training data presents an unusual intersection between competition law and intellectual property law.

A firm may attempt to obtain exclusive rights to large quantities of training material.

Competition concerns could arise if a dominant AI company:

acquires exclusive datasets;

prevents rivals from obtaining equivalent data;

uses contractual restrictions to foreclose competing models;

combines proprietary data with dominant distribution infrastructure.

But ownership of intellectual property does not automatically establish competition-law liability.

12. Killer Acquisitions in AGI

Large technology companies may acquire:

AI startups;

model developers;

AI-agent companies;

robotics firms;

data companies;

semiconductor startups.

A transaction may be competitively significant even where the target has limited current revenue.

The concern is that the target may represent a future competitive constraint.

This makes traditional merger thresholds potentially problematic where innovative AI companies have:

high technological value;

low current turnover;

rapidly growing user bases;

significant future potential.

Competition authorities therefore need to examine:

pipeline competition;

innovation competition;

access to talent;

intellectual property;

data;

model architectures;

future market entry.

13. Acqui-Hiring and AI Talent

AGI markets may be unusually dependent upon scarce researchers and engineers.

A dominant company could potentially acquire a startup principally to obtain:

researchers;

engineers;

model architectures;

patents;

technical know-how.

Competition authorities could examine whether such transactions eliminate an important innovation competitor.

Labour-market competition principles may also become relevant where firms coordinate compensation or restrict employee mobility.

14. Algorithmic Collusion

AGI systems may eventually negotiate prices or commercial terms autonomously.

This creates an important distinction between:

parallel conduct, and

agreement or coordinated conduct.

If independent AI agents independently arrive at similar prices, traditional evidence of an agreement may be absent.

But if firms intentionally design their AI systems to:

exchange competitively sensitive information;

coordinate prices;

monitor rivals;

punish deviations;

maintain supra-competitive prices,

traditional cartel principles could potentially apply.

The fact that an algorithm executed the conduct would not necessarily eliminate the underlying competition-law issue.

15. Autonomous AGI Agents and Antitrust Liability

A more difficult question arises where an AGI agent independently makes commercial decisions.

For example, an agent could autonomously:

negotiate prices;

select suppliers;

allocate customers;

refuse transactions;

coordinate inventory;

respond to competitor pricing.

The legal question would be whether the conduct can be attributed to the firm deploying the system.

Competition law generally focuses on the economic conduct of undertakings rather than whether a human physically pressed the button.

Therefore, delegating commercial decision-making to an algorithm would not necessarily create an exemption from antitrust law.

16. Tacit Coordination

AGI may increase market transparency.

AI systems can rapidly:

monitor competitors;

detect price changes;

predict competitor behaviour;

adjust prices;

identify deviations.

This may make coordination easier even without direct communication.

The principal difficulty is proving the legal element of concerted conduct.

Competition authorities may therefore need to examine:

algorithm design;

training instructions;

communications between firms;

data exchanges;

common software providers;

pricing rules;

deliberate coordination mechanisms.

17. Platform Parity Clauses

An AGI platform might impose:

“You cannot offer your AI service at a lower price through another platform.”

Such parity clauses could restrict competition between AI distribution channels.

Potential effects include:

reduced price competition;

higher commissions;

reduced entry;

platform dependence;

exclusion of smaller distributors.

The legal treatment would depend upon the precise contractual structure and market power involved.

18. AGI Interoperability

Interoperability could become one of the most important competition questions.

If a dominant AGI provider refuses to permit competing services to interact with:

operating systems;

cloud infrastructure;

enterprise databases;

messaging platforms;

application ecosystems,

competition could be weakened.

Interoperability remedies may therefore include:

API access;

data portability;

functional access;

technical compatibility;

non-discriminatory interfaces.

The EU's 2026 DMA action concerning Google specifically illustrates how interoperability requirements can be used to address competitive concerns surrounding AI services. (Digital Markets Act (DMA))

19. Essential-Facilities Issues

Potential AGI bottlenecks include:

specialised computing capacity;

unique datasets;

AI chips;

cloud infrastructure;

operating-system access;

model interfaces.

However, not every scarce resource constitutes an essential facility.

Courts generally require demanding conditions before imposing compulsory access.

Two important authorities are:

Aspen Skiing Co. v. Aspen Highlands Skiing Corp.

Relevant to exceptional refusal-to-deal circumstances.

Verizon Communications Inc. v. Trinko

The Supreme Court cautioned against broadly imposing duties to deal with competitors.

These cases are important because competition law must balance access to essential inputs against incentives to invest and innovate.

20. AI Infrastructure and Bottleneck Power

The AGI ecosystem may create several bottlenecks simultaneously:

The OECD's 2026 analysis specifically identifies hardware and data as areas where structural concentration can potentially entrench incumbent advantages. (OECD)

21. Six Important Case Laws

Because there is not yet a mature body of reported judicial decisions specifically concerning AGI, existing digital-platform, refusal-to-deal, tying, essential-facilities and innovation cases provide the principal legal analogies.

1. United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

Principle: exclusionary conduct involving a dominant platform.

The case concerned Microsoft's operating-system dominance and practices affecting browser competition.

AGI relevance:
A dominant operating-system or cloud provider integrating its own AGI while restricting competing AI systems could raise analogous questions concerning platform foreclosure and exclusionary conduct.

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

Principle: exceptional refusal to deal.

The Supreme Court found liability in circumstances involving termination of a previously profitable cooperative arrangement and conduct inconsistent with ordinary economic interests.

AGI relevance:
Potentially relevant where an infrastructure provider historically supplied an AI competitor and then deliberately terminates access in a manner designed to eliminate competition.

3. Verizon Communications Inc. v. Trinko, 540 U.S. 398 (2004)

Principle: limits on compulsory dealing.

The Supreme Court emphasized that competition law ordinarily does not impose a general obligation on monopolists to share their resources with rivals.

AGI relevance:
Highly important for disputes involving demands that dominant cloud or computing providers supply infrastructure to competing AGI developers.

4. Magill TV Guide/Joined Cases C-241/91 P and C-242/91 P, RTE and ITP v Commission

Principle: exceptional circumstances concerning refusal to license intellectual property.

The European Court of Justice developed important conditions for competition-law intervention in refusal-to-license situations involving intellectual property.

AGI relevance:
Potentially relevant where a dominant AI company controls indispensable model-related intellectual property and refuses access to rivals.

5. IMS Health GmbH & Co. OHG v NDC Health GmbH & Co. KG, Case C-418/01

Principle: refusal to license intellectual property and indispensable infrastructure.

The Court established demanding conditions for compulsory licensing under EU competition law.

AGI relevance:
Could become relevant to proprietary AI architectures, datasets, interoperability interfaces or other protected resources where competitors claim that access is indispensable.

6. European Commission v. Google (Android), Case AT.40099

Principle: tying and leveraging within a digital ecosystem.

The European Commission examined Google's contractual and technological arrangements concerning Android and associated services.

AGI relevance:
An integrated AI ecosystem could similarly involve:

operating system → app distribution → search → AI assistant → cloud.

If a dominant firm uses power in one layer to advantage its AI service in another, Android-type reasoning may become relevant.

22. Additional Relevant Authorities

United States v. Google LLC — Search

The Google search litigation provides a modern example of competition law examining exclusionary arrangements and distribution channels surrounding a digital platform. The case remains subject to continuing remedies proceedings in 2026. (Department of Justice)

AGI relevance: distribution agreements and defaults could determine which AI assistant users encounter first.

European Commission v. Google Shopping

The case is relevant to self-preferencing and the use of dominance in one digital market to advantage a related service.

AGI relevance: a dominant platform could potentially rank or display its own AGI service more prominently than competing AI products.

Qualcomm Inc. v. FTC

Relevant to exclusionary licensing practices, technological markets and competition involving important inputs.

AGI relevance: analogous questions could arise around licensing of AI technologies and essential technological components.

United States v. Google LLC — AdTech

The U.S. government's ad-tech litigation provides another contemporary example of competition concerns arising where a company occupies multiple vertically related positions. The 2026 remedies proceedings remain ongoing. (Reuters)

23. Competition Concerns in Cloud–AI Partnerships

Cloud/AI partnerships deserve special attention because they combine several competitive relationships.

A cloud company may simultaneously be:

Investor + supplier + distributor + competitor + data holder.

This creates potential conflicts.

For example:

Cloud provider invests in AGI developer
↓
Developer becomes dependent upon cloud infrastructure
↓
Rival AI developers face less favourable infrastructure access
↓
Cloud provider gains downstream AI advantage.

The FTC's 2025 report specifically identified access to computing, engineering talent, switching costs and sensitive information as potential competition issues in such partnerships. (Federal Trade Commission)

24. Killer Acquisitions and Innovation Competition

Traditional market-share analysis may underestimate competition in AGI.

A startup may have:

no substantial revenue;

few customers;

significant technological potential.

Yet it may be developing a fundamentally different architecture capable of challenging an incumbent.

Therefore, merger analysis should examine:

current competition;

potential competition;

innovation competition;

future market entry;

access to researchers;

intellectual property;

datasets;

computing resources.

25. Consumer Welfare and AGI

Competition authorities may examine effects on:

prices;

quality;

innovation;

privacy;

choice;

interoperability;

service reliability.

AGI introduces an additional issue: quality may not be adequately captured by price.

Many AI services are offered at low or zero monetary prices.

Competitive harm may instead manifest through:

reduced model quality;

reduced privacy;

fewer choices;

reduced innovation;

restrictive data practices;

lower interoperability.

26. Innovation Competition

Innovation is likely to be particularly important in AGI.

A dominant firm could potentially harm competition without immediately increasing prices by:

acquiring emerging competitors;

suppressing rival research;

restricting API access;

controlling scarce computing;

limiting interoperability;

preventing employees from joining competitors.

Consequently, authorities may need to consider dynamic competition, not merely static market shares.

27. Regulatory Interaction

Competition law will coexist with:

AI safety regulation;

data-protection law;

copyright law;

semiconductor controls;

cybersecurity regulation;

consumer protection;

sector-specific regulation.

An important issue is whether AI firms could justify restrictive cooperation on the ground that it is necessary for safety.

Competition authorities are already considering this issue. In September 2026, a senior DOJ antitrust official stated that AI-safety coordination did not appear inherently anticompetitive, while noting the continuing competition-law questions surrounding such cooperation. (Reuters)

This demonstrates the need to distinguish:

legitimate technical/safety cooperation

from

commercially coordinated restrictions on competition.

28. Possible Antitrust Remedies

If unlawful conduct is established, possible remedies could include:

Structural remedies

divestiture;

separation of cloud and AI businesses;

prohibition of certain acquisitions.

Behavioural remedies

non-discrimination obligations;

interoperability;

API access;

data portability;

prohibition of exclusive contracts.

Merger remedies

divestiture of technology;

licensing;

access commitments;

restrictions on information exchange.

Monitoring

Independent monitoring could be required where a dominant AI platform has substantial continuing market power.

29. Challenges for Competition Authorities

AGI creates several enforcement difficulties.

29.1 Rapid technological change

A market definition can become obsolete quickly.

29.2 Lack of transparent pricing

Many AI services are free or subscription-based.

29.3 Multisided markets

The same company may serve:

consumers;

developers;

advertisers;

enterprises;

cloud customers.

29.4 Intangible inputs

Data, algorithms and talent are difficult to measure.

29.5 Innovation uncertainty

It can be difficult to determine whether a startup would actually become a competitive threat.

29.6 Global markets

AI firms operate across jurisdictions, requiring cooperation between competition authorities.

30. Overall Legal Framework

A useful competition-law framework for AGI can therefore be summarized as follows:

31. Conclusion

The principal antitrust significance of AGI lies in the possibility that market power may accumulate across several vertically connected layers rather than solely at the model level.

The most important areas for competition-law scrutiny are likely to be:

control of AI computing infrastructure;

exclusive cloud–AI partnerships;

access to unique training data;

foundation-model dominance;

self-preferencing by integrated platforms;

tying and bundling of AI with other digital services;

acquisitions of emerging AI competitors;

interoperability and API restrictions;

algorithmic coordination;

control over AI distribution channels.

The existing authorities—particularly Microsoft, Aspen Skiing, Trinko, Magill, IMS Health and Google Android—provide the principal doctrinal foundations for analysing these issues. There is not yet a settled body of jurisprudence specifically defining an “AGI market,” so courts and competition authorities will likely adapt established principles concerning digital platforms, essential inputs, vertical foreclosure, intellectual property, mergers and innovation competition to the evolving AI ecosystem.

The contemporary regulatory direction confirms the importance of this analysis: the FTC has already investigated major cloud–AI partnerships, while European authorities have moved toward addressing AI interoperability and access issues within digital ecosystems. (Federal Trade Commission)

Thus, the central competition-law challenge for AGI is to preserve contestability and innovation without treating technological scale, integration or legitimate AI collaboration as unlawful in themselves.