Competition Law and Strategic Competition Policy for Intelligence-Based Ecosystems

1. Introduction

Intelligence-based ecosystems are markets in which artificial intelligence (AI), machine learning, data analytics, automated decision-making, foundation models, cloud computing, algorithms, sensors, and digital platforms interact to produce products or services.

Examples include:

• generative-AI ecosystems;
• autonomous-vehicle platforms;
• AI-powered healthcare;
• algorithmic financial services;
• smart manufacturing;
• intelligent logistics;
• AI advertising;
• recommendation platforms;
• autonomous purchasing systems;
• smart-grid systems; and
• AI-enabled cloud ecosystems.

These ecosystems create a distinctive competition-law problem because market power may no longer arise from control over a single product. It may arise from control over several interconnected layers:

Data → Compute → AI Model → Platform → Application → Distribution → Consumer

A firm controlling several layers can potentially use advantages at one level to reinforce its position at another.

The strategic competition-policy objective is therefore to preserve contestability, innovation, interoperability, access and consumer choice without unnecessarily restricting technological development.

2. Meaning of Intelligence-Based Ecosystems

An intelligence-based ecosystem is a network of complementary markets in which AI or other computational intelligence is central to the creation, distribution or consumption of products.

A simplified ecosystem is:

Data

↓

Compute / Chips / Cloud

↓

Foundation Model

↓

AI Application

↓

Platform / Marketplace

↓

Consumer or Business User

Each layer can create a potential bottleneck.

For example:

• data may be difficult to replicate;
• compute may be controlled by a small number of suppliers;
• foundation models may require enormous investment;
• cloud providers may control infrastructure;
• platforms may control distribution;
• applications may be dependent upon APIs.

3. Why Competition Policy Needs a Strategic Approach

Traditional competition law often focuses on a particular relevant market.

Intelligence ecosystems require a broader analysis because competitive advantages can migrate between connected markets.

For example:

A cloud provider controls computing infrastructure.

↓

It develops a leading AI model.

↓

The model is integrated into its productivity software.

↓

The software is distributed through its operating system.

↓

Users become dependent on the ecosystem.

The competitive issue is therefore not simply:

"Does the company dominate cloud computing?"

It may involve ecosystem leverage across multiple markets.

4. Core Strategic Competition Principles

A strategic framework should focus on:

1. Contestability

Can new firms enter and expand?

2. Interoperability

Can competing systems communicate?

3. Data access and portability

Can competitors obtain necessary data without undermining privacy?

4. Compute access

Can emerging AI developers obtain adequate computing resources?

5. Model neutrality

Can competing AI models reach users on reasonable terms?

6. Distribution neutrality

Can platforms avoid unjustifiably favouring their own AI products?

7. Innovation

Does competition remain capable of producing technological breakthroughs?

8. Competitive neutrality

Do state-supported or vertically integrated companies compete under comparable conditions?

5. Strategic Competition Concern 1 — Data Concentration

Data can be a major source of competitive advantage.

A leading intelligence-based ecosystem may accumulate:

• consumer data;
• search data;
• transaction data;
• location data;
• behavioural data;
• industrial data;
• medical data;
• financial information.

The competitive concern is not merely the quantity of data.

It is whether:

The data advantage creates a durable barrier that competitors cannot reasonably overcome.

Relevant questions include:

• Is the data unique?
• Can competitors reproduce it?
• Is the data essential?
• Is access technically feasible?
• Would sharing violate privacy?
• Does the incumbent combine data across markets?

6. Strategic Competition Concern 2 — Compute Concentration

Advanced AI requires substantial computational resources.

Important inputs include:

• GPUs;
• specialised AI accelerators;
• cloud computing;
• data centres;
• networking;
• energy.

If a small number of firms control these inputs, downstream AI competition may become dependent upon them.

Potential concerns include:

• discriminatory access;
• exclusive arrangements;
• capacity reservation;
• tying;
• preferential cloud access;
• refusal to supply;
• discriminatory pricing.

7. Strategic Competition Concern 3 — Foundation Models

Foundation models can become central infrastructure for numerous applications.

A dominant model provider could potentially control:

Model → API → Application ecosystem → Users

Potential competition concerns include:

• exclusive distribution;
• discriminatory API access;
• bundling;
• self-preferencing;
• restrictions on competing models;
• excessive switching costs;
• preferential access to data.

However, economies of scale in model development can also generate genuine efficiencies.

Competition policy must therefore distinguish efficient scale from strategic foreclosure.

8. Strategic Competition Concern 4 — Cloud and AI Integration

Cloud providers may simultaneously supply:

• compute;
• storage;
• AI infrastructure;
• foundation models;
• applications.

This creates potential vertical foreclosure.

For example:

Cloud provider

↓

AI infrastructure

↓

Foundation model

↓

Enterprise application

A vertically integrated undertaking may have incentives to disadvantage competing model providers.

The relevant competition question is:

Does integration create efficiencies that benefit users, or does the integrated structure materially restrict rival access?

9. Case Law 1 — Microsoft v Commission

Facts

Microsoft was found to have abused its dominant position through practices involving interoperability information and tying.

Competition principle

A dominant technology provider can infringe competition law when it uses control over a technological platform to restrict competition in adjacent markets.

Intelligence-ecosystem relevance

The case provides an important conceptual foundation for AI ecosystems.

A modern equivalent might involve:

dominant operating system/cloud platform → AI service → competing AI providers.

If competitors cannot obtain necessary interoperability or access on reasonable terms, ecosystem control can potentially become a mechanism of foreclosure.

10. Case Law 2 — Google Shopping

Facts

Google was found to have systematically favoured its own comparison-shopping service in general search results.

The General Court largely upheld the Commission's decision.

Principle

A dominant digital gateway can potentially be used to favour an undertaking's own downstream service.

AI relevance

An AI-powered search or assistant could similarly become a major gateway.

For example, an AI assistant might:

• recommend its owner's products;
• rank affiliated services more prominently;
• suppress competing applications;
• steer users toward affiliated marketplaces.

Thus, AI-mediated self-preferencing is a potential extension of established digital competition principles.

11. Case Law 3 — Google Android

Facts

The European Commission examined Google's conduct involving Android, including tying and contractual restrictions concerning search and applications.

Principle

Dominance at one technological layer can potentially be leveraged into neighbouring markets.

AI relevance

The same ecosystem logic can apply to:

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

If access to an important platform is conditioned upon adoption of affiliated AI services, competition concerns may arise.

12. Case Law 4 — Qualcomm

Facts

The European Commission examined Qualcomm's conduct concerning baseband chipsets and alleged exclusionary payments.

The case involved the strategic importance of key technological components.

Principle

Control over an important technological input can have downstream competitive consequences where commercial conduct forecloses rivals.

Intelligence-ecosystem relevance

AI ecosystems depend on specialised hardware.

Potentially analogous concerns could arise where an important AI-chip supplier:

• ties hardware to software;
• restricts access to competing platforms;
• enters exclusive arrangements;
• uses rebates to exclude competing chip suppliers.

13. Case Law 5 — Intel v Commission

Facts

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

The CJEU held that where a dominant undertaking contests an infringement finding by arguing that its conduct was not capable of foreclosing equally efficient competitors, the Commission must properly consider the relevant economic circumstances.

Principle

Effects-based analysis is important in exclusionary-abuse cases.

Intelligence-ecosystem relevance

AI companies may provide:

• cloud credits;
• model discounts;
• developer subsidies;
• preferential API pricing;
• infrastructure rebates.

The fact that such incentives exist does not by itself establish illegality.

The key question is their capability and effect in foreclosing competition.

14. Case Law 6 — Bronner v Mediaprint

Facts

A newspaper publisher sought access to Mediaprint's distribution infrastructure.

CJEU approach

The Court adopted a demanding standard for refusal to provide access to an alleged essential facility.

Intelligence-ecosystem relevance

Potential bottlenecks in AI include:

• compute infrastructure;
• proprietary datasets;
• APIs;
• cloud services;
• specialised chips.

But strategic importance alone does not establish an essential facility.

A competition authority would need to examine:

• indispensability;
• alternatives;
• feasibility of duplication;
• exclusionary effects.

15. Case Law 7 — IMS Health

Facts

IMS Health controlled a system for pharmaceutical sales data.

A competitor sought access to the system.

Principle

Refusal to license intellectual property can constitute an abuse in exceptional circumstances where the stringent legal requirements are satisfied.

AI relevance

The case is important for:

• proprietary AI datasets;
• specialised databases;
• software interfaces;
• model inputs;
• technical interoperability.

The basic policy tension remains:

Innovation incentives versus competitive access.

16. Case Law 8 — Deutsche Telekom

Facts

Deutsche Telekom was found to have engaged in a margin squeeze concerning telecommunications access.

Principle

A dominant vertically integrated infrastructure provider may not structure access and downstream pricing in a way that excludes effective competitors.

AI relevance

The modern analogy could involve:

Cloud infrastructure → AI service → downstream application.

If a dominant cloud/AI provider controls an essential input and makes downstream competition commercially unviable, competition law may become relevant.

17. Strategic Competition Concern 5 — AI Self-Preferencing

Self-preferencing occurs where a platform gives its own service preferential treatment.

In an intelligence ecosystem, this could involve:

• AI search;
• AI shopping;
• AI travel;
• AI financial services;
• AI advertising;
• AI productivity tools.

Example:

AI assistant receives ten competing products.

↓

It systematically recommends its owner's product.

↓

Competing products receive less visibility.

The competition concern is strongest where:

• the platform is dominant;
• users rely heavily upon its recommendations;
• switching is difficult;
• competitors cannot access equivalent distribution.

18. Strategic Competition Concern 6 — AI Agent Gatekeepers

AI agents may increasingly act as intermediaries between users and businesses.

A consumer might instruct:

"Find and purchase the cheapest suitable flight."

The AI agent could:

1. search airlines;
2. compare prices;
3. rank options;
4. choose a provider;
5. complete payment.

The agent could therefore become a new digital gatekeeper.

Competition concerns could include:

• biased recommendations;
• exclusive commercial relationships;
• discriminatory ranking;
• hidden commissions;
• self-preferencing;
• suppression of rival suppliers.

19. Strategic Competition Concern 7 — Algorithmic Coordination

AI systems can observe competitors continuously.

Suppose several firms deploy autonomous pricing systems.

The systems may:

• observe competitors;
• adjust prices;
• learn from market responses;
• predict competitor behaviour.

The systems might converge on high prices even without explicit human communication.

Competition law must distinguish:

Independent adaptation

from

Coordinated conduct.

The mere existence of parallel algorithmic pricing is not automatically proof of an unlawful agreement.

Evidence concerning system design, communication, instructions and market conduct remains critical.

20. Strategic Competition Concern 8 — Common AI Providers

Several competitors may purchase AI services from the same provider.

For example:

Retailer A → AI Pricing Provider
Retailer B → AI Pricing Provider
Retailer C → AI Pricing Provider

If the provider receives sensitive information from all three customers and uses that information to optimise pricing across the market, competition concerns may arise.

Potential risks include:

• exchange of commercially sensitive information;
• coordination;
• uniform pricing;
• strategic information leakage.

This creates a hub-and-spoke risk.

21. Strategic Competition Concern 9 — Interoperability

Interoperability becomes increasingly important as AI ecosystems develop.

A dominant ecosystem might restrict:

• API access;
• model portability;
• agent communication;
• data portability;
• application compatibility.

The strategic concern is:

Can competing intelligence systems communicate sufficiently to preserve effective competition?

Interoperability may reduce:

• switching costs;
• lock-in;
• network effects;
• entry barriers.

But mandatory access should remain sensitive to:

• cybersecurity;
• privacy;
• intellectual property;
• technical feasibility.

22. Strategic Competition Concern 10 — AI Mergers

AI markets create difficult merger-control questions.

A startup may have:

• little turnover;
• valuable researchers;
• proprietary data;
• an innovative model;
• rapidly increasing users.

A large platform may acquire it before it becomes a significant competitor.

The strategic competition framework should therefore consider:

Innovation competition

Could the target become an important innovator?

Data competition

Does the target possess unique data?

Technology competition

Does the target have an alternative architecture?

Ecosystem competition

Could it challenge the incumbent ecosystem?

Talent competition

Would the acquisition eliminate an independent research centre?

23. Strategic Competition Concern 11 — Killer Acquisitions

A killer acquisition occurs where a dominant or powerful company acquires an emerging competitor partly or substantially to eliminate future competition.

In intelligence-based markets, this may be particularly significant because technological trajectories can change rapidly.

Today's small AI application could become tomorrow's:

• platform;
• model;
• distribution channel;
• data ecosystem.

Merger review therefore needs to consider potential competition, not merely current market share.

24. Strategic Competition Concern 12 — Data Portability

Data portability can reduce ecosystem lock-in.

Example:

Platform A

↓

Consumer data

↓

Portable format

↓

Platform B

This can make switching easier.

However, data portability should be designed consistently with:

• privacy law;
• data security;
• intellectual property;
• confidential business information.

Competition policy should therefore distinguish between legitimate portability and compulsory access to commercially sensitive information without sufficient justification.

25. Strategic Competition Concern 13 — Vertical Integration

An intelligence-based company could control:

Chip → Cloud → Model → Agent → Marketplace.

Vertical integration can produce efficiencies:

• lower transaction costs;
• improved performance;
• better security;
• faster innovation.

But it can also enable:

• input foreclosure;
• customer foreclosure;
• tying;
• bundling;
• self-preferencing;
• discriminatory access.

The analysis should therefore be effects-based rather than assuming that vertical integration is inherently harmful.

26. Strategic Competition Concern 14 — AI and Small Businesses

AI can actually increase competition by lowering barriers to entry.

A small company can use AI for:

• marketing;
• accounting;
• customer service;
• logistics;
• product development;
• coding.

This can reduce the minimum scale necessary to compete.

However, if SMEs become dependent upon a small number of AI infrastructure providers, a new form of dependency can emerge:

SME → cloud provider → model provider → AI platform.

Strategic competition policy should therefore preserve multiple suppliers.

27. Strategic Competition Concern 15 — Consumer Welfare

Consumer welfare in intelligence-based ecosystems includes more than price.

It includes:

Price

Are AI-enabled services affordable?

Quality

Does AI improve the product?

Choice

Can consumers choose among competing systems?

Privacy

How is consumer information used?

Innovation

Can new AI products emerge?

Reliability

Are consumers exposed to systemic technological failures?

Competition policy should therefore evaluate the full competitive effects of AI ecosystems.

28. Strategic Competition Framework

A practical framework can be expressed as:

Stage 1 — Map the ecosystem

Identify:

Data → Compute → Model → Application → Distribution → Consumer

Stage 2 — Identify bottlenecks

Ask whether any undertaking controls:

• data;
• compute;
• infrastructure;
• APIs;
• distribution.

Stage 3 — Assess market power

Consider:

• market shares;
• network effects;
• switching costs;
• data advantages;
• scale;
• ecosystem integration.

Stage 4 — Identify conduct

Look for:

• tying;
• bundling;
• self-preferencing;
• discriminatory access;
• exclusivity;
• refusal to deal;
• predatory strategies.

Stage 5 — Evaluate effects

Consider:

• entry;
• innovation;
• consumer choice;
• prices;
• quality;
• resilience.

Stage 6 — Consider efficiencies

Determine whether integration creates:

• genuine technological efficiencies;
• security benefits;
• lower costs;
• better quality;
• innovation.

Stage 7 — Select proportionate remedies

Potential remedies include:

• interoperability;
• data portability;
• non-discrimination;
• access obligations;
• behavioural commitments;
• structural separation where justified.

29. Strategic Competition Policy Matrix

30. Competition Compliance for Intelligence-Based Businesses

Companies operating intelligence-based ecosystems should adopt an AI competition-compliance programme.

A. Algorithmic audits

Review whether algorithms:

• coordinate prices;
• discriminate against competitors;
• favour affiliated products.

B. Data governance

Prevent inappropriate use of competitor-sensitive information.

C. Access policies

Use transparent criteria for API and platform access.

D. Interoperability

Assess whether technical restrictions are objectively justified.

E. Merger controls

Review acquisitions of emerging AI competitors carefully.

F. Human oversight

Maintain responsibility for competition-sensitive decisions delegated to AI.

31. Key Case-Law Principles

32. Central Distinction

Competition policy should distinguish:

Pro-competitive intelligence ecosystem

Data

↓

Innovation

↓

Efficient AI model

↓

Better products

↓

Lower costs

↓

More consumer choice

from:

Anti-competitive intelligence ecosystem

Data control

↓

Infrastructure control

↓

Model control

↓

Distribution control

↓

Self-preferencing

↓

Rival foreclosure

↓

Entrenched ecosystem dominance

The existence of an integrated ecosystem does not itself establish an infringement. The crucial question is whether market power is being used in a manner capable of substantially restricting effective competition.

33. Conclusion

Strategic competition policy for intelligence-based ecosystems requires competition law to move beyond isolated product markets and examine the interconnected architecture through which modern AI markets operate.

The central strategic risks include:

• concentration of data;
• concentration of compute;
• foundation-model dominance;
• cloud dependency;
• interoperability restrictions;
• AI-agent gatekeeping;
• self-preferencing;
• algorithmic coordination;
• vertical foreclosure;
• exclusionary access conditions; and
• acquisitions of emerging AI competitors.

The leading authorities—Microsoft, Google Shopping, Google Android, Qualcomm, Intel, Bronner, IMS Health and Deutsche Telekom—provide established competition-law principles that can be applied to these emerging structures.

The fundamental principle is:

Competition policy should permit intelligence-based ecosystems to obtain legitimate efficiencies from scale, integration, data and technological innovation, while preventing control over critical ecosystem layers from being converted into durable barriers to entry, interoperability, innovation or consumer choice.

Thus, the appropriate strategic approach is neither automatic regulation of large AI ecosystems nor unrestricted technological consolidation, but a competition framework centred on contestability, interoperability, non-discriminatory access, innovation and evidence-based assessment of actual or likely competitive effects.