Competition Law and Future Governance of Intelligence-Based Economies

1. Introduction

An intelligence-based economy is an economic system in which artificial intelligence, machine learning, foundation models, autonomous agents, data, cloud computing, algorithms, robotics and automated decision-making become major inputs into production and commercial activity.

Competition law in such an economy cannot remain limited to traditional questions of price, output and market share. Competitive power may increasingly arise from control over:

• high-quality datasets;
• computing capacity and GPUs;
• foundation and frontier AI models;
• cloud infrastructure;
• AI application programming interfaces (APIs);
• algorithms and model weights;
• specialised AI talent;
• distribution platforms and app stores;
• user feedback and data-feedback loops;
• interoperability standards;
• AI-agent ecosystems; and
• strategic partnerships between model developers and infrastructure providers.

The emerging regulatory approach reflects this transformation. The UK CMA has specifically identified access, diversity, choice, flexibility, fair dealing, transparency and accountability as principles relevant to competitive AI foundation-model markets.

The European Union, United Kingdom and United States competition authorities have also recognised that AI markets may generate competition risks involving control of important inputs, lock-in, interoperability, partnerships and extension of market power.

2. Meaning of Intelligence-Based Economies

An intelligence-based economy differs from an ordinary digital economy because computational intelligence itself becomes an economic resource.

A simplified value chain is:

Data → Compute → Foundation Model → AI Infrastructure → AI Applications → AI Agents → Consumers/Businesses

For example:

Data provider → cloud/GPU provider → foundation-model developer → AI application → autonomous agent → business/customer

Competition concerns can arise at every layer.

Major characteristics

A. Data-driven production

Data may constitute a critical input into:

• training;
• fine-tuning;
• personalisation;
• recommendation;
• prediction;
• autonomous decision-making.

Large incumbents may possess extensive historical datasets unavailable to smaller rivals.

B. Compute concentration

Training sophisticated models requires substantial computing infrastructure.

This may create competition concerns where a small number of firms control:

• GPUs;
• specialised chips;
• cloud computing;
• data centres;
• networking infrastructure;
• AI acceleration technologies.

C. Network and feedback effects

AI systems can improve through user interaction.

More users → more data → better models → more users.

This can create a data-feedback loop capable of reinforcing market power.

D. Ecosystem integration

A single enterprise may operate:

• operating systems;
• search engines;
• cloud infrastructure;
• app stores;
• AI models;
• productivity software;
• advertising systems.

Vertical integration may produce efficiencies, but it can also create opportunities for exclusionary conduct.

3. Traditional Competition Law and the Intelligence Economy

The principal competition-law doctrines remain relevant.

A. Abuse of dominance

The central question remains whether a dominant undertaking uses market power to exclude competitors or exploit customers.

Potential AI examples include:

• tying an AI assistant to an operating system;
• exclusive cloud arrangements;
• discriminatory access to model APIs;
• self-preferencing an affiliated AI service;
• refusal to provide essential data;
• discriminatory access to computing resources;
• contractual restrictions preventing multi-homing.

B. Anti-competitive agreements

AI companies may exchange competitively sensitive information through:

• algorithms;
• common pricing systems;
• shared platforms;
• data pools;
• automated contracting systems.

This creates questions concerning algorithmic collusion.

An AI system could potentially learn that aggressive competition is disadvantageous and independently converge upon parallel pricing.

The legal question would remain whether the conduct results from independent decision-making or from an agreement, concerted practice or other legally cognisable coordination.

C. Merger control

Traditional merger control may be insufficient if an AI company acquires a small emerging competitor before the target generates significant revenue.

The target may nevertheless possess:

• valuable technology;
• highly specialised employees;
• datasets;
• model weights;
• intellectual property;
• strategic distribution channels.

Accordingly, killer acquisitions and acqui-hiring can become particularly important.

4. Six Major Case Laws and Their Relevance

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

Facts

Microsoft was found to have unlawfully maintained its operating-system monopoly through conduct involving Internet Explorer and the distribution of competing browsers.

The case concerned exclusionary strategies designed to prevent emerging technologies from becoming competitive threats to Microsoft's dominant platform.

Competition principle

The case demonstrates that competition law may intervene where a dominant platform uses its existing position to restrict the development or distribution of a potentially disruptive technology.

Relevance to intelligence economies

The Microsoft principle can be applied conceptually to AI ecosystems.

For example, competition authorities may examine whether a dominant platform:

• bundles an AI assistant with a dominant operating system;
• restricts rival AI assistants;
• limits access to distribution channels;
• imposes exclusionary contractual conditions;
• uses interoperability restrictions to disadvantage competitors.

The deeper lesson is that control of a platform can be leveraged into an emerging technological market.

Case 2: Google Android — Google and Alphabet v Commission, T-604/18

The EU General Court examined Google's conduct concerning Android devices, application distribution, search and browser applications.

The Court largely upheld the Commission's findings concerning restrictions imposed on manufacturers and mobile-network operators and reduced the Commission's fine to €4.125 billion.

Competition principle

The case demonstrates the importance of:

• tying;
• exclusivity;
• distribution restrictions;
• ecosystem leverage;
• reinforcing dominance across related markets.

AI relevance

An intelligence-based ecosystem could similarly involve:

Cloud → AI model → operating system → productivity software → AI assistant

A dominant undertaking could potentially use one layer to favour another.

For example:

dominant cloud provider + proprietary AI model + preferential access + restrictive contractual conditions

could create barriers for competing AI developers.

Thus, the Android case provides an important analytical framework for multi-layer AI ecosystems.

Case 3: Google Shopping — Google Search (Shopping)

The Google Shopping litigation concerned Google's treatment of its comparison-shopping service within its general search results.

The central competition concern was that a dominant search platform could use its position in one market to advantage its own related service.

Competition principle

The case is particularly relevant to:

• self-preferencing;
• ranking;
• platform neutrality;
• leveraging;
• discrimination between affiliated and independent services.

AI relevance

AI search and answer engines may become the principal gateway through which consumers obtain information.

An AI platform could potentially:

• favour its own products;
• prioritise affiliated AI applications;
• suppress competing model outputs;
• privilege its own shopping or advertising services;
• manipulate recommendation rankings.

The Google Shopping framework therefore has substantial relevance to AI-mediated markets, where an algorithm rather than a conventional search-results page determines commercial visibility.

Case 4: FTC v Facebook/Meta

The US Federal Trade Commission's monopolisation litigation concerning Facebook/Meta alleges that Facebook maintained monopoly power through a course of conduct including acquisitions of Instagram and WhatsApp and restrictive conditions imposed on developers.

The case remained subject to appellate proceedings after a 2025 district-court ruling in Meta's favour; the FTC filed an appeal in January 2026.

Competition principle

The case illustrates the importance of examining:

• acquisitions of emerging competitors;
• nascent competitive threats;
• network effects;
• platform ecosystems;
• cumulative exclusionary conduct.

AI relevance

AI markets develop extremely quickly.

An incumbent may acquire or establish close relationships with a smaller AI company possessing:

• superior technology;
• specialised talent;
• a valuable dataset;
• a new model architecture;
• an important application.

Therefore, merger review may need to consider future competitive significance, not merely current turnover.

Case 5: Microsoft / OpenAI Partnership — UK CMA

This is an especially important contemporary AI competition matter.

The CMA investigated Microsoft's partnership with OpenAI. In March 2025, the CMA decided that the partnership did not qualify for investigation under the UK merger provisions.

Competition significance

The investigation is important because it demonstrates that competition authorities are examining whether AI partnerships and investments themselves can produce merger-like competitive effects.

The concern is not necessarily traditional acquisition.

Instead, economic influence may arise through:

• investment;
• governance rights;
• exclusive arrangements;
• cloud dependence;
• computing access;
• technical integration;
• information rights;
• commercial agreements.

The FTC's separate study of major AI partnerships identified potential effects involving access to computing resources and engineering talent, switching costs and access to commercially sensitive information.

Future significance

This suggests that merger control in intelligence economies may need to examine economic control rather than formal ownership alone.

Case 6: Microsoft / Inflection AI — UK CMA

The CMA examined Microsoft's hiring of former Inflection employees and associated arrangements.

The CMA determined that the transaction constituted a relevant merger situation but found no realistic prospect of a substantial lessening of competition resulting from horizontal unilateral effects.

Competition principle

The case is significant because it demonstrates that competition authorities may scrutinise:

• acqui-hiring;
• acquisition of human capital;
• transfer of intellectual property;
• licensing arrangements;
• technology partnerships.

AI relevance

AI companies depend heavily upon specialised technical employees.

Consequently, acquiring an AI firm's principal team may have competitive implications even where conventional asset acquisition does not occur.

The case therefore illustrates the emerging importance of talent concentration as a competition issue.

5. Emerging AI Competition Principles

The CMA's work provides a useful framework.

Its updated AI principles include access, diversity and choice, together with other principles designed to support competitive AI markets.

The international 2024 statement by the EU, UK and US competition authorities similarly emphasised:

1. fair dealing;
2. interoperability; and
3. choice. 

These principles can be translated into future competition governance.

6. Access to Critical AI Inputs

One of the most important future questions is:

Who controls the inputs necessary to build intelligence?

These inputs include:

• datasets;
• compute;
• chips;
• cloud infrastructure;
• engineering talent;
• energy;
• specialised knowledge;
• model architectures.

If access becomes concentrated, competition authorities may consider:

Refusal to supply

A dominant infrastructure provider may potentially refuse access to critical infrastructure.

Discriminatory access

Access may be technically available but offered to rivals on worse conditions.

Exclusive dealing

A cloud provider might impose arrangements preventing an AI developer from using competing infrastructure.

Foreclosure

Vertical integration may allow an infrastructure provider to disadvantage independent AI developers.

7. Data as a Competition Asset

Data creates several novel competition questions.

A. Data accumulation

A dominant AI platform may accumulate enormous quantities of:

• behavioural data;
• search data;
• transaction data;
• conversational data;
• location data;
• purchasing data.

B. Data feedback loops

More users can generate more information.

More information can improve the AI model.

A better model attracts more users.

This creates:

Users → Data → Better AI → More Users

Such feedback effects can raise barriers to entry.

The EU has already required Google, under the Digital Markets Act, to share anonymised search data with eligible competing search engines on fair, reasonable and non-discriminatory terms; the Commission's 2026 specification proceedings specifically address effective data sharing, including for AI chatbots offering search functionality.

8. Compute as a New Essential Input

Compute may become analogous to infrastructure.

Large-scale AI development can require:

• specialised processors;
• enormous computing clusters;
• data centres;
• high-speed networking;
• electricity;
• cloud infrastructure.

Consequently, future competition law may increasingly ask:

Can a new AI firm realistically compete if it cannot obtain sufficient computing capacity?

This could lead to competition concerns involving:

• exclusive cloud agreements;
• capacity reservation;
• discriminatory cloud pricing;
• preferential access;
• interoperability barriers;
• switching costs.

The issue is particularly important because AI partnerships may simultaneously connect a model developer with a major cloud provider.

9. AI Partnerships and Merger Control

Traditional merger control asks whether two undertakings combine.

Intelligence economies increasingly require examination of quasi-integration.

Examples include:

Investment

A cloud company invests in an AI developer.

Partnership

The AI developer becomes commercially dependent upon the cloud company.

Exclusive supply

The model is supplied exclusively through one cloud.

Talent transfer

The infrastructure company hires the AI company's principal research team.

Licensing

Critical AI technology is licensed to one platform.

These arrangements may not always constitute mergers, but they can have merger-like competitive effects.

The CMA's investigations of Microsoft/OpenAI, Microsoft/Inflection and Microsoft/Mistral AI demonstrate how competition authorities are beginning to examine these structures.

10. Algorithmic Collusion

Artificial intelligence creates a new form of competition risk.

Suppose several competing firms employ autonomous pricing systems.

Each system may:

1. observe competitors;
2. analyse market conditions;
3. adjust prices;
4. learn from the resulting market;
5. modify its strategy.

The systems might converge upon stable prices without conventional human communication.

The central legal distinction would be between:

Independent algorithmic adaptation

and

coordination resulting from an agreement or concerted practice.

Competition law therefore needs sophisticated rules for determining:

• who designed the algorithm;
• what objectives were programmed;
• what information was shared;
• whether firms communicated;
• whether coordination was foreseeable;
• whether the conduct was autonomous;
• whether human operators intervened.

11. Autonomous AI Agents

The next stage beyond generative AI is the AI agent.

An AI agent may independently:

• negotiate;
• purchase products;
• select suppliers;
• change prices;
• execute contracts;
• allocate resources;
• manage inventories.

This creates an important question:

Who bears competition-law responsibility when an autonomous system makes an anti-competitive decision?

Potentially relevant actors include:

• the deploying enterprise;
• the model developer;
• the software provider;
• the data provider;
• the human operator.

Competition law will therefore increasingly need algorithmic accountability.

12. Interoperability

Interoperability may become one of the most important future competition tools.

AI ecosystems could become locked into:

• proprietary model formats;
• proprietary APIs;
• proprietary agent protocols;
• proprietary datasets;
• proprietary cloud environments.

A consumer or business may therefore face high switching costs.

Competition authorities may consider whether firms should provide:

• API portability;
• data portability;
• model interoperability;
• technical compatibility;
• switching mechanisms.

The EU/UK/US joint statement specifically identifies interoperability as a principle supporting competition and innovation.

13. Self-Preferencing in AI

Traditional search engines rank websites.

AI systems increasingly generate answers.

This changes the competitive problem.

An AI platform controlling the interface could theoretically favour:

its own model → its own application → its own marketplace → its own advertising service.

This may make self-preferencing more difficult to detect because the output is generated dynamically rather than displayed through a conventional ranking mechanism.

Competition authorities may therefore need to examine:

• model-output allocation;
• recommendation logic;
• training incentives;
• ranking criteria;
• affiliated-service treatment;
• disclosure of commercial relationships.

14. AI and Essential Facilities

The traditional essential-facilities concept may acquire new significance.

Potential AI essential inputs could include:

• critical datasets;
• cloud capacity;
• AI infrastructure;
• interoperability interfaces;
• technical standards.

However, not every valuable AI input should automatically be treated as an essential facility.

Authorities would need to consider established requirements such as:

• indispensability;
• lack of realistic alternatives;
• competitive foreclosure;
• duplication feasibility;
• justification for refusal.

15. Digital Markets Regulation and Intelligence Economies

Traditional competition law is generally ex post.

It intervenes after potentially harmful conduct occurs.

Modern digital regulation increasingly introduces ex ante obligations.

The UK's Digital Markets, Competition and Consumers Act regime is particularly relevant. The CMA states that the new digital-markets regime is intended to provide flexible intervention in rapidly developing digital markets, including emerging technologies such as AI.

The EU's Digital Markets Act similarly creates obligations concerning access, interoperability and data.

This produces a developing regulatory model:

Antitrust + merger control + ex-ante digital regulation + sector regulation

rather than relying upon conventional antitrust alone.

16. Cloud and AI Infrastructure

Cloud computing is becoming closely connected with AI competition.

In June 2026, the European Commission announced preliminary findings that Amazon Web Services and Microsoft Azure should be designated as DMA gatekeepers for their cloud services, citing their scale, entrenched positions, switching costs, ecosystem effects and the increasing importance of AI tools and partnerships in cloud procurement.

This demonstrates the increasingly close relationship between:

Cloud power → AI infrastructure → AI models → downstream applications.

Future competition analysis may therefore need to examine the entire technological stack rather than isolated product markets.

17. Future Governance Model

A comprehensive competition framework for intelligence-based economies could contain the following components.

18. Role of Competition Authorities

Future competition authorities will increasingly need technical capabilities in:

A. Algorithmic auditing

Authorities may need to understand how algorithms make commercial decisions.

B. Data analysis

Large datasets may need to be examined to identify discriminatory or exclusionary effects.

C. Computational economics

Traditional market-share analysis may need to be supplemented by:

• network effects;
• switching costs;
• multi-homing;
• data advantages;
• feedback loops;
• ecosystem effects.

D. AI expertise

Competition agencies may require:

• AI engineers;
• data scientists;
• cybersecurity specialists;
• economists;
• technology lawyers.

E. International cooperation

AI markets operate across national boundaries.

The EU, UK and US competition authorities have already recognised the importance of cooperation in AI competition enforcement.

19. Challenges for Future Competition Law

1. Defining the relevant market

Is the relevant market:

• AI models?
• foundation models?
• AI assistants?
• cloud AI?
• search?
• productivity software?

The answer may change rapidly.

2. Measuring market power

Market share alone may not capture AI power.

Other indicators include:

• computing capacity;
• data access;
• model performance;
• ecosystem control;
• user dependence;
• switching costs;
• developer dependence.

3. Innovation versus exclusion

AI companies often need scale to innovate.

Competition law must therefore distinguish between:

legitimate technological integration

and

exclusionary leveraging of market power.

4. Speed of technological change

Traditional litigation may take years while AI markets can change within months.

This creates pressure for faster regulatory intervention.

5. Global enforcement

An AI model may be trained in one country, hosted in another, incorporated into software in a third and sold globally.

Jurisdictional cooperation will consequently become increasingly important.

20. Future Competition-Law Tests

A useful analytical framework for intelligence-based economies is:

Step 1 — Identify the intelligence layer

Is the conduct occurring in:

Data → Compute → Model → Cloud → Application → Agent → Distribution?

Step 2 — Identify the source of market power

Ask whether power derives from:

• data;
• compute;
• technology;
• network effects;
• ecosystem integration;
• distribution;
• intellectual property;
• switching costs.

Step 3 — Identify the conduct

Possible conduct includes:

• tying;
• bundling;
• exclusivity;
• self-preferencing;
• refusal to supply;
• discriminatory access;
• predatory conduct;
• acquisition;
• algorithmic coordination.

Step 4 — Assess competitive effects

Examine:

• foreclosure;
• entry barriers;
• innovation;
• consumer choice;
• prices;
• quality;
• privacy;
• switching costs.

Step 5 — Examine efficiencies

AI integration can generate legitimate benefits through:

• economies of scale;
• improved model performance;
• security;
• lower costs;
• innovation;
• interoperability.

Step 6 — Select proportionate remedies

Potential remedies include:

• access obligations;
• interoperability;
• data portability;
• non-discrimination;
• behavioural commitments;
• structural remedies;
• merger prohibition;
• divestiture where legally justified.

21. Overall Legal Significance

The central transformation is from competition between products to competition between intelligence ecosystems.

The future competitive environment may involve:

Data + Compute + Models + Cloud + Platforms + Agents + Users

A company controlling several of these layers can potentially possess a form of economic power substantially different from traditional industrial dominance.

The Microsoft, Google Android, Google Shopping and Meta cases provide established legal principles concerning platform leverage, tying, exclusion, ecosystem effects and acquisitions. The newer Microsoft/OpenAI, Microsoft/Inflection and related AI investigations demonstrate that competition authorities are increasingly applying these concerns to AI-specific commercial structures.

The CMA's foundation-model work is particularly significant because it expressly recognises that competitive AI markets require continuing access to inputs, diversity among models, meaningful choice and protection against exclusionary conduct.

Conclusion

Competition law in intelligence-based economies will evolve from a primarily market-based framework into an ecosystem-governance framework.

The principal future concerns will be:

1. control of AI inputs;
2. concentration of compute and cloud infrastructure;
3. data-feedback loops;
4. foundation-model concentration;
5. AI partnerships and quasi-mergers;
6. acqui-hiring and talent concentration;
7. self-preferencing by AI platforms;
8. algorithmic coordination;
9. AI-agent autonomy;
10. interoperability and switching costs;
11. access to essential AI infrastructure; and
12. cross-border cooperation between competition authorities.

Accordingly, the future of competition law will not be concerned merely with who sells the product, but increasingly with who controls the intelligence, infrastructure, data and interfaces through which economic decisions are made.