Competition Law and Strategic Governance of AI-Driven Markets

1. Introduction

Artificial intelligence (“AI”) is transforming competition by changing how firms produce, price, distribute, advertise, innovate, and interact with consumers. AI-driven markets can generate substantial efficiencies, but they can also create new forms of market power through control over data, computing capacity, foundation models, algorithms, cloud infrastructure, application ecosystems, and distribution channels.

Competition law therefore increasingly has to address not merely traditional questions of price and market share, but also:

• control over strategic datasets;
• access to computing infrastructure;
• AI model interoperability;
• algorithmic pricing and coordination;
• self-preferencing by AI platforms;
• tying and bundling of AI services;
• exclusive arrangements involving cloud or AI infrastructure;
• acquisitions of AI startups;
• discrimination in access to APIs or model outputs;
• network effects and ecosystem expansion;
• switching costs and technological lock-in; and
• exploitation of innovation and quality dimensions of competition.

Strategic governance of AI-driven markets means designing competition policy before market structures become irreversibly concentrated, while preserving legitimate innovation incentives.

2. Meaning of AI-Driven Markets

An AI-driven market is a market in which AI substantially affects the competitive process.

Examples include:

1. generative-AI foundation models;
2. AI-enabled search engines;
3. AI assistants and agents;
4. autonomous vehicles;
5. algorithmic advertising;
6. AI-powered financial services;
7. AI healthcare platforms;
8. AI cloud-computing services;
9. semiconductor and GPU ecosystems;
10. AI cybersecurity;
11. recommendation systems;
12. AI-powered marketplaces.

The competitive structure can involve several interconnected layers:

Compute → Data → Foundation Model → API → Application → Distribution → Consumer

A firm that controls several layers may obtain an advantage extending beyond its traditional market.

3. Legal Framework

A. Abuse of Dominance

A dominant AI firm may abuse its position through:

• exclusionary contracts;
• refusal of interoperability;
• discriminatory access;
• tying;
• bundling;
• predatory pricing;
• excessive or exploitative terms;
• self-preferencing;
• exclusionary rebates;
• restrictions on data portability.

The precise legal test varies across jurisdictions.

India

The principal framework is the Competition Act, 2002, particularly:

• Section 4 — abuse of dominant position;
• Section 3 — anti-competitive agreements;
• Section 5 — combinations;
• Section 19 — investigation;
• Sections 26 onward — investigative/enforcement procedure.

The Competition Commission of India (“CCI”) can therefore examine AI conduct where the relevant market and statutory jurisdiction are established.

4. Market Definition in AI Markets

Traditional market definition becomes complicated because AI products may be:

• free to consumers;
• monetised through advertising;
• supplied through APIs;
• bundled with other software;
• offered as cloud services; or
• subsidised by another product.

Consequently, price-based SSNIP analysis alone may be insufficient.

Relevant dimensions can include:

Product market

For example:

• general-purpose AI assistants;
• enterprise AI platforms;
• AI cloud infrastructure;
• AI-powered search;
• AI advertising tools.

Geographic market

AI services can potentially operate globally, although:

• regulation;
• language;
• data localisation;
• cloud infrastructure;
• cybersecurity requirements; and
• consumer preferences

may create narrower geographic markets.

Non-price competition

Competition authorities may examine:

• accuracy;
• reliability;
• privacy;
• latency;
• model quality;
• interoperability;
• innovation;
• safety;
• customization.

5. Data as a Strategic Competitive Asset

AI models require enormous quantities of data.

Data can create competitive advantages through:

Data acquisition → model training → improved performance → greater adoption → additional data → further model improvement

This can produce a feedback loop.

A dominant platform may therefore obtain competitive advantages by controlling:

• consumer behavioural data;
• search data;
• transaction data;
• proprietary datasets;
• training datasets;
• user feedback;
• location information.

However, possession of large quantities of data does not automatically establish dominance. Competition analysis must consider whether the data is:

• unique;
• difficult to reproduce;
• commercially valuable;
• necessary for effective competition;
• obtainable from alternative sources.

6. Compute and GPU Bottlenecks

AI development depends heavily on computing infrastructure.

Important inputs include:

• GPUs;
• AI accelerators;
• cloud computing;
• high-bandwidth networking;
• data centres;
• model-training infrastructure.

This creates potential competition concerns where a firm controlling an essential technological input also competes downstream.

Possible theories include:

Vertical foreclosure

A cloud provider may restrict access to computing resources for competing AI developers.

Discriminatory access

An infrastructure provider may provide materially better terms to its affiliated AI service.

Bundling

Cloud services may be bundled with proprietary AI models in a manner that disadvantages competing models.

Switching costs

Long-term cloud commitments can make migration to another infrastructure provider expensive.

7. Foundation Models and Platform Power

Foundation models can function as technological platforms.

A major AI provider may simultaneously operate:

• a foundation model;
• API infrastructure;
• cloud services;
• applications;
• an app store;
• search;
• advertising;
• productivity software.

This creates the possibility of ecosystem leverage.

The central competition question becomes:

Can market power in one AI layer be leveraged to protect or extend power into another layer?

8. Algorithmic Pricing and Collusion

AI systems can independently analyse competitors' prices and rapidly adjust prices.

This creates difficult questions under cartel law.

Traditional cartel enforcement generally looks for:

• agreement;
• concerted practice;
• communication;
• coordination.

But AI systems may produce parallel pricing without direct human communication.

Three situations should be distinguished:

1. Explicit coordination

Humans instruct algorithms to implement an unlawful agreement.

This can constitute ordinary cartel conduct.

2. Algorithm-mediated coordination

Competitors exchange information or use a common algorithm capable of facilitating coordinated behaviour.

Competition authorities may examine whether the technology merely implements an existing coordination mechanism.

3. Autonomous algorithmic parallelism

Algorithms independently learn that following competitors' prices maximises profits.

This creates a more difficult question because parallel conduct alone does not necessarily establish an unlawful agreement.

9. AI and Tacit Coordination

AI can increase:

• speed of price adjustments;
• market transparency;
• ability to monitor competitors;
• ability to punish deviations;
• frequency of interaction.

These characteristics may facilitate tacit coordination.

Competition authorities therefore need sophisticated economic evidence rather than treating every instance of algorithmic parallel pricing as a cartel.

Relevant evidence may include:

• model instructions;
• pricing algorithms;
• training data;
• communication between competitors;
• internal documents;
• pricing patterns;
• monitoring systems;
• automated retaliation mechanisms.

10. Self-Preferencing by AI Platforms

An AI platform may operate both:

Platform + competing application

For example, an AI marketplace might rank its own AI application more prominently than rival applications.

Potential concerns include:

• preferential ranking;
• preferential API access;
• preferential computing resources;
• preferential training-data access;
• preferential visibility;
• discriminatory interoperability.

The relevant question is whether the conduct harms the competitive process rather than merely disadvantaging an individual competitor.

11. Tying and Bundling

AI can be bundled with:

• operating systems;
• search engines;
• productivity suites;
• cloud services;
• smartphones;
• browsers;
• enterprise software.

Competition concerns may arise if customers effectively have to obtain Product A to access Product B.

The legal assessment may consider:

1. whether the products are distinct;
2. whether the firm has market power in the tying product;
3. whether customers are coerced or commercially induced;
4. whether rivals are foreclosed;
5. whether legitimate efficiencies justify the arrangement.

12. AI Mergers and Acquisitions

AI creates particularly important merger-control questions.

A startup may have:

• relatively low revenue;
• highly valuable technology;
• important engineers;
• proprietary datasets;
• strategic partnerships;
• an important AI model.

A conventional revenue-based merger threshold can therefore fail to capture strategically significant transactions.

Competition authorities increasingly examine:

• innovation pipelines;
• potential competition;
• access to data;
• computing resources;
• AI talent;
• intellectual property;
• interoperability;
• ecosystem effects.

13. Killer Acquisitions and Nascent Competition

An incumbent may acquire a small AI company before it becomes a significant competitor.

The transaction may eliminate:

• future competition;
• innovative technology;
• alternative AI architecture;
• a potential source of disruptive innovation.

This is particularly significant where the target has little current revenue but substantial technological potential.

14. AI Ecosystems and Network Effects

AI markets can exhibit strong network effects.

A simplified cycle is:

More users → more data → better model → better product → more users

Additional network effects can arise from:

• developers;
• APIs;
• plugins;
• third-party applications;
• enterprise integrations.

This can create high barriers to entry.

Competition policy therefore needs to consider dynamic competition, not merely current market shares.

15. Interoperability and Data Portability

Interoperability can reduce switching costs.

Possible mechanisms include:

• API access;
• model portability;
• data portability;
• open technical standards;
• compatibility between AI agents;
• interoperable identity systems.

A dominant AI platform that deliberately prevents interoperability may potentially protect its market position.

However, mandatory interoperability can also create:

• cybersecurity risks;
• intellectual-property concerns;
• privacy risks;
• reduced incentives to innovate.

The remedy must therefore be proportionate.

16. AI and Essential Facilities

Traditional essential-facility reasoning may become relevant to:

• AI computing infrastructure;
• critical datasets;
• cloud infrastructure;
• model-access interfaces;
• technical standards.

The strongest cases generally require more than simply showing that a resource is useful.

Questions include:

1. Is the facility genuinely indispensable?
2. Can competitors reasonably reproduce it?
3. Is access technically feasible?
4. Would refusal eliminate effective competition?
5. Is there an objective justification for refusal?

17. Six Important Case Laws

1. United States v. Microsoft Corp. (2001)

Facts

Microsoft was found to have used its operating-system dominance to restrict competition from rival browser technologies.

Competition principle

The case demonstrates how a dominant technology platform can leverage power from one technological layer into another.

Relevance to AI

AI platforms may similarly possess:

• operating-system power;
• search power;
• cloud power;
• distribution power;

and potentially use that position to advantage their own AI products.

The case is therefore highly relevant to ecosystem leverage, tying and technological foreclosure.

18. Google Search (Shopping) — European Commission

Facts

The European Commission found that Google had systematically favoured its own comparison-shopping service in search results over competing services.

Principle

The case illustrates competition concerns surrounding self-preferencing by a dominant platform.

AI relevance

An AI search platform could potentially:

• rank its own services preferentially;
• favour affiliated AI applications;
• prioritise proprietary content;
• disadvantage competing AI agents.

The competition concern would depend on the applicable legal framework and evidence of exclusionary effects.

19. Google Android — European Commission

Facts

The European Commission examined Google's contractual practices involving Android devices, including tying arrangements involving Google applications and restrictions affecting competing search services.

Principle

The case demonstrates how dominance in one technological ecosystem can potentially be leveraged into adjacent markets.

AI relevance

The same analytical problem may arise when AI is embedded into:

• smartphones;
• operating systems;
• browsers;
• search;
• cloud services.

AI therefore makes the traditional concept of technological tying particularly important.

20. FTC v. Facebook, Inc. / Meta Platforms Litigation

Facts

The U.S. Federal Trade Commission challenged Meta's acquisitions of Instagram and WhatsApp, alleging that the transactions contributed to the preservation of monopoly power.

Principle

The proceedings illustrate the importance of examining acquisitions involving nascent or potential competitors, rather than focusing exclusively on current revenue.

AI relevance

An established AI platform acquiring an emerging AI developer could raise similar questions where the target represents:

• potential competition;
• technological innovation;
• an alternative ecosystem;
• a future disruptive technology.

The ultimate legal assessment depends on the evidence and applicable merger law.

21. United States v. Google LLC — Search and Advertising Litigation

Facts

U.S. antitrust proceedings against Google have examined practices concerning search distribution and digital advertising.

Principle

The proceedings demonstrate the importance of analysing distribution agreements, defaults, network effects and exclusionary conduct in digital markets.

AI relevance

AI assistants may become new gateways to information.

If an AI assistant becomes a major information-distribution channel, agreements governing:

• default placement;
• search integration;
• device distribution;
• advertising;
• data access

could become important competition-law issues.

22. Epic Games, Inc. v. Apple Inc.

Facts

Epic challenged Apple's App Store rules, particularly Apple's control over distribution and payment arrangements.

Principle

The case illustrates the competition significance of platform governance, access restrictions and control over digital distribution.

AI relevance

AI platforms may develop comparable ecosystems involving:

• AI applications;
• agents;
• plugins;
• model marketplaces;
• APIs.

Control over access to such ecosystems can become an important source of market power.

23. Qualcomm Antitrust Litigation

Facts

Qualcomm faced extensive antitrust litigation concerning its licensing practices and relationships with device manufacturers.

Principle

The case demonstrates the importance of analysing vertical relationships, intellectual property, licensing and exclusionary strategies in technologically intensive markets.

AI relevance

Similar questions may arise concerning:

• AI model licensing;
• semiconductor IP;
• AI accelerators;
• model APIs;
• proprietary technical standards.

24. Key Competition Issues in AI Markets

25. Strategic Governance Model

A modern AI competition regime can be structured around seven governance pillars.

Pillar 1 — Market Monitoring

Competition authorities should continuously monitor:

• AI concentration;
• model providers;
• cloud infrastructure;
• GPU supply;
• acquisitions;
• exclusive agreements.

Pillar 2 — Merger Surveillance

Authorities should scrutinise acquisitions based not only on revenue but also:

• innovation significance;
• potential competition;
• data;
• talent;
• technology;
• ecosystem position.

Pillar 3 — Infrastructure Access

Where appropriate, competition authorities can examine discriminatory access to:

• compute;
• cloud;
• APIs;
• technical infrastructure.

Pillar 4 — Algorithmic Conduct

Authorities should develop technical capacity to investigate:

• algorithmic pricing;
• automated coordination;
• ranking algorithms;
• recommendation systems.

Pillar 5 — Interoperability

Appropriate interoperability requirements can reduce:

• switching costs;
• ecosystem lock-in;
• artificial entry barriers.

Pillar 6 — Data Governance

Competition policy should coordinate with:

• privacy law;
• data-protection law;
• intellectual-property law;
• consumer protection.

Pillar 7 — Innovation Protection

Competition enforcement should preserve incentives to invest in:

• model development;
• computing infrastructure;
• datasets;
• AI safety;
• new applications.

26. Strategic Foresight in AI Antitrust

AI competition law should not simply respond after dominance has become entrenched.

A strategic-foresight approach asks:

Stage 1 — Identify emerging bottlenecks

Examples:

GPU → cloud → foundation model → API → application → distribution

Stage 2 — Identify potential gatekeepers

Determine which firms control critical inputs or distribution channels.

Stage 3 — Monitor exclusionary strategies

Look for:

• exclusive agreements;
• acquisitions;
• discriminatory access;
• bundling;
• self-preferencing;
• interoperability restrictions.

Stage 4 — Measure dynamic effects

Consider:

• innovation;
• entry;
• investment;
• consumer choice;
• quality;
• privacy.

Stage 5 — Intervene proportionately

Possible remedies include:

• behavioural commitments;
• interoperability;
• non-discrimination;
• divestiture;
• access remedies;
• merger prohibition in appropriate cases.

27. AI-Specific Evidence

Competition authorities increasingly need technical evidence.

Important evidence can include:

• source code;
• model architecture;
• API documentation;
• training-data arrangements;
• cloud contracts;
• GPU allocation;
• internal communications;
• algorithmic logs;
• pricing outputs;
• ranking algorithms;
• acquisition documents.

Traditional documentary evidence should therefore be supplemented by technical and computational evidence.

28. Relationship Between Competition Law and AI Regulation

Competition law should not operate in isolation.

AI governance can involve:

Competition Law + Data Protection + Consumer Protection + AI Safety + Intellectual Property + Sector Regulation

For example, mandatory data sharing may increase competition but simultaneously create privacy concerns.

Similarly, transparency requirements may improve accountability but expose proprietary technology.

Therefore, regulatory coordination is essential.

29. Challenges for Competition Authorities

1. Rapid technological change

AI markets evolve faster than traditional regulatory processes.

2. Technical complexity

Authorities require expertise in:

• machine learning;
• cloud architecture;
• semiconductors;
• algorithms;
• data science.

3. Difficult market definition

Many AI products are multi-sided or offered at zero monetary price.

4. Dynamic competition

Today's small AI firm could become tomorrow's major competitor.

5. International markets

AI supply chains frequently cross multiple jurisdictions.

6. Innovation trade-offs

Intervention that is too weak may permit durable concentration; intervention that is poorly designed may reduce innovation.

30. Conclusion

Competition law in AI-driven markets is moving from a traditional focus on price, market share and conventional cartels toward a broader examination of data, computing power, algorithms, ecosystems, interoperability, innovation and technological gatekeeping.

The central strategic governance challenge is to ensure that AI innovation remains contestable.

The most important areas for competition authorities are:

1. AI merger control;
2. control over compute and cloud infrastructure;
3. data-related market power;
4. foundation-model concentration;
5. self-preferencing and ecosystem leverage;
6. algorithmic coordination;
7. AI platform interoperability;
8. exclusive arrangements;
9. AI marketplace governance; and
10. protection of potential and nascent competition.

The Microsoft, Google Shopping, Google Android, Meta/Facebook, Epic Games–Apple and Qualcomm lines of cases demonstrate that many apparently new AI competition problems have recognizable precedents in technology-platform, vertical-conduct, merger and ecosystem competition law. At the same time, AI introduces genuinely new questions because algorithms can autonomously learn, adapt and interact at a speed and scale that traditional competition frameworks were not designed to address.

Thus, strategic governance of AI-driven markets requires competition authorities to combine conventional antitrust principles with continuous technological monitoring, economic analysis, merger foresight, interoperability assessment and technically informed enforcement.