Competition Law and Future Competition Governance in Intelligence-Centric Economies

1. Introduction

An intelligence-centric economy is an economy in which competitive advantage increasingly depends on artificial intelligence (AI), machine learning, foundation models, algorithms, data, cloud computing, specialised chips, automated decision systems, and autonomous agents.

In such an economy, competition law cannot focus only on conventional questions such as price, market share and output. Market power may instead arise from control over:

• high-quality datasets;
• computing capacity and AI chips;
• cloud infrastructure;
• foundation models and model weights;
• proprietary algorithms;
• AI talent and specialised expertise;
• application programming interfaces (APIs);
• digital ecosystems;
• distribution channels;
• user-generated data;
• interoperability standards; and
• autonomous AI agents capable of making commercial decisions.

The OECD's recent work identifies compute, data, skills, vertical integration and first-mover advantages as important structural issues in AI markets, while also noting that some foundation-model markets have remained relatively dynamic.

The future therefore requires a shift from competition enforcement after market power has become entrenched toward a combination of ex-ante regulation, merger control, interoperability, access obligations, algorithmic monitoring and traditional antitrust enforcement.

2. Meaning of Intelligence-Centric Competition

Traditional industrial competition can be represented as:

Firm → Product → Consumer → Price

Intelligence-centric competition increasingly resembles:

Data → Compute → Model → Algorithm → Platform → Agent → Consumer/Business

The competitive significance of each layer can affect the others.

For example:

Control over cloud infrastructure → preferential access to compute → stronger AI model → more users → more data → better model → stronger downstream ecosystem.

This creates the possibility of self-reinforcing competitive advantages.

The OECD has specifically identified linkages across the AI value chain and barriers to access to quality data and computing power as potential competition risks.

3. Objectives of Future Competition Governance

Future competition governance should pursue several interconnected objectives.

A. Contestability

Markets should remain open to new entrants even where incumbents possess substantial technological advantages.

B. Innovation competition

Competition law should protect not merely today's prices but also:

• future innovation;
• alternative technologies;
• competing business models;
• technological diversity.

C. Access to essential inputs

Where particular inputs become indispensable, competition authorities may need to examine access to:

• compute;
• data;
• cloud infrastructure;
• APIs;
• model interfaces;
• interoperability standards.

D. Prevention of ecosystem foreclosure

A firm controlling several layers of an AI ecosystem may have incentives to disadvantage competitors at adjacent levels.

E. Consumer autonomy

AI systems can influence consumer choice through personalised recommendations, automated pricing and autonomous purchasing decisions.

4. New Sources of Market Power

4.1 Data concentration

Data may function as a competitive input comparable to capital or physical infrastructure.

A dominant firm may possess:

• proprietary behavioural data;
• transaction histories;
• search data;
• location data;
• industrial datasets;
• medical or scientific datasets;
• labelled training data.

The competition concern arises when rivals cannot reasonably reproduce the relevant dataset.

However, possession of large quantities of data does not automatically establish dominance. The relevant question is whether the data produces a durable competitive advantage that rivals cannot effectively replicate.

4.2 Compute concentration

Advanced AI requires substantial computing resources.

Competition may therefore depend upon access to:

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

The OECD has highlighted the structural concentration of cloud services and specialised AI chips as potential competition concerns.

This makes infrastructure competition an important part of future AI antitrust.

4.3 Algorithmic market power

Algorithms can:

• determine prices;
• rank products;
• allocate consumers;
• recommend content;
• determine credit;
• select suppliers;
• optimise advertising;
• control access to platforms.

An algorithm can therefore become the mechanism through which market power is exercised.

The OECD observes that algorithms can produce efficiency-enhancing effects but can also facilitate restrictions of competition.

5. Six Major Competition-Law Case Laws

Case 1: Google Search (Shopping) — European Commission

Facts

Google was found to have favoured its own comparison-shopping service within general search results while demoting competing comparison-shopping services.

Competition issue

The case concerned self-preferencing and leveraging of dominance.

Principle

A firm possessing substantial market power in one digital layer may potentially use that position to disadvantage competitors in an adjacent market.

Future relevance

The principle becomes particularly important where an AI platform controls:

Search → AI assistant → recommendation → transaction

If an AI assistant systematically favours its own commercial services, competition authorities may examine whether this amounts to leveraging or exclusionary conduct.

The European Commission's Google Shopping decision resulted in a €2.42 billion fine, and the OECD identifies the case as an important example of digital self-preferencing.

Case 2: United States v. Google — Search and Distribution

Facts

The United States challenged Google's agreements concerning distribution of its search engine.

Competition issue

The case concerned whether contractual arrangements could reinforce Google's position in general search by restricting effective distribution opportunities for rivals.

Principle

Competition analysis in digital markets must examine not merely the immediate contractual relationship but also how distribution arrangements affect market structure and entry.

Future AI significance

AI assistants may increasingly become distribution gateways.

For example:

Operating system → default AI assistant → search → advertising → commerce

A dominant company controlling the gateway could potentially use default arrangements or exclusive agreements to reinforce its position.

The U.S. Department of Justice continues to maintain the Google antitrust litigation and related remedies proceedings.

Case 3: Amazon Marketplace Pricing Algorithms — United States

Facts

The U.S. Department of Justice pursued an antitrust case involving online marketplace sellers using identical repricing software in circumstances raising concerns about coordinated pricing.

Competition issue

The important question was whether technology could facilitate coordination among competitors.

Principle

An algorithm does not immunise otherwise anticompetitive conduct from competition law.

Future significance

The issue becomes much more complicated with autonomous AI agents.

Suppose competing firms allow AI agents to independently determine prices. If those agents:

• observe competitors;
• exchange information;
• optimise toward common pricing outcomes; or
• use a common intermediary,

the resulting coordination could raise questions under cartel and concerted-practice rules.

The OECD identifies the U.S. Amazon Marketplace posters matter and the UK's Trod/GBE matter as examples of algorithm-assisted coordination.

Case 4: Trod Ltd / GB Eye Ltd — United Kingdom

Facts

Online sellers used automated pricing software in connection with selling posters and frames.

Competition issue

The competition authority examined the use of pricing algorithms to implement an arrangement that prevented effective price competition.

Principle

Technology may be the instrument of cartel implementation rather than the legal justification for the conduct.

Future significance

Future AI agents could potentially create:

• algorithmic collusion;
• autonomous price coordination;
• common-vendor coordination;
• market allocation;
• automated information exchange.

The legal challenge will be determining when apparently independent machine behaviour is attributable to the firms using the systems.

Case 5: Nvidia–Mellanox

Facts

Nvidia acquired Mellanox, a major provider of networking equipment used in data centres.

The transaction was examined by multiple competition authorities. The European Commission cleared it unconditionally, while China's SAMR imposed conditions.

Competition issue

The transaction illustrated the importance of vertical and conglomerate theories of harm in AI infrastructure.

Principle

Competition authorities increasingly need to examine whether control over complementary technologies can be used to disadvantage rivals.

Future significance

The same analysis may apply to:

AI chips + networking + cloud + foundation models + AI applications.

A merger need not eliminate a direct competitor to raise competition concerns. It may instead give the merged firm the ability to control a strategically important bottleneck.

Case 6: Illumina/GRAIL

Facts

Illumina acquired GRAIL, a company developing cancer-detection technology.

The case became particularly significant for European merger jurisdiction and the treatment of transactions involving firms that may have strategic importance despite not meeting traditional national turnover thresholds.

Competition issue

The broader significance concerns the ability of competition authorities to examine transactions involving innovative firms before a potentially important competitive constraint disappears.

Future AI significance

AI start-ups may have:

• low turnover;
• substantial intellectual property;
• important talent;
• valuable datasets;
• strategically significant algorithms.

Traditional turnover-based merger thresholds may therefore fail to identify some competitively important acquisitions.

The CJEU's 2024 judgment nevertheless placed limits on the European Commission's ability to rely on Article 22 EUMR referrals where referring Member States themselves lack jurisdiction under national law.

6. Additional Relevant Competition-Law Precedents

Several established cases also provide principles that can be adapted to intelligence-centric markets.

Microsoft

Important for:

• tying;
• interoperability;
• access to technical information;
• leveraging dominance.

Bronner

Important for the essential-facilities/access framework and the circumstances in which refusal of access by a dominant undertaking may raise Article 102 concerns.

Intel

Important for:

• exclusionary rebates;
• economic effects;
• assessment of foreclosure.

Qualcomm

Important for:

• technology licensing;
• exclusionary arrangements;
• innovation-intensive markets.

Meta/Facebook

Important for considering how data-related practices can interact with market power and competition.

These precedents demonstrate that AI competition law does not necessarily require an entirely new legal philosophy. Existing doctrines can often be adapted to new technological environments.

7. Intelligence Ecosystems and the Leveraging Problem

The most significant future issue may be ecosystem leverage.

Consider:

Cloud → Compute → Foundation Model → AI Assistant → App Store → Payments → Advertising → Consumer Data

A company controlling several layers may possess incentives to favour its own downstream products.

Possible practices include:

• preferential API access;
• discriminatory cloud pricing;
• exclusive AI partnerships;
• self-preferencing;
• tying;
• bundling;
• interoperability restrictions;
• discriminatory model access;
• data combination;
• refusal to supply;
• loyalty arrangements.

The competition authority therefore needs to examine the whole ecosystem, rather than treating every layer as completely independent.

8. AI Partnerships and Competition

AI development increasingly involves partnerships between:

• model developers;
• cloud providers;
• chip manufacturers;
• platform operators;
• telecommunications companies;
• financial institutions.

Such arrangements may generate enormous efficiencies.

But competition analysis should ask:

1. Does the agreement exclude competing AI developers?
2. Does it provide preferential access to scarce compute?
3. Does it create exclusive distribution?
4. Does it prevent switching?
5. Does it facilitate information exchange?
6. Does it increase entry barriers?
7. Does it create a de facto standard?

The appropriate approach should distinguish pro-competitive technological cooperation from arrangements that substantially reduce contestability.

9. Merger Control in Intelligence-Centric Economies

Traditional merger thresholds based primarily on turnover may become inadequate.

An AI start-up may have:

• minimal current revenue;
• valuable patents;
• leading researchers;
• a powerful model;
• strategic training data;
• an important technology.

Consequently, future merger policy may require greater attention to:

A. Transaction value

Large acquisition prices may signal competitive significance.

B. Innovation assets

Authorities may examine technology and research pipelines.

C. Talent acquisitions

Acquiring an entire team may eliminate a potential competitor.

D. Data acquisitions

Control of datasets can create substantial competitive advantages.

E. Compute-related acquisitions

Control over AI infrastructure may have strategic implications beyond current revenues.

The Illumina/GRAIL litigation demonstrates the importance—and limits—of attempting to bring competitively significant transactions within merger scrutiny.

10. Ex-Ante Competition Regulation

Traditional antitrust generally operates after potentially harmful conduct occurs.

Intelligence-centric markets may require some ex-ante obligations.

The European Union's Digital Markets Act is a prominent example of this approach, imposing obligations on designated gatekeepers to protect contestability and fair conditions.

Potential future obligations could include:

Interoperability

Dominant AI ecosystems may be required to provide technically meaningful interoperability.

Data portability

Users and businesses could move relevant data between competing services.

Non-discrimination

Infrastructure providers could be restricted from discriminating against downstream competitors.

Transparency

Certain automated ranking or access decisions could require explanation or auditability.

Switching rights

Technical and contractual barriers to changing providers could be reduced.

11. Cloud and Compute as Competition Infrastructure

Cloud infrastructure may become analogous to critical infrastructure in the intelligence economy.

A particularly important future concern is the combination:

Cloud + AI models + chips + data + distribution

The European Commission announced in June 2026 a preliminary view that Amazon Web Services and Microsoft Azure should be designated as DMA gatekeepers for cloud computing services, citing their gateway role, entrenched user bases, switching costs and AI-related ecosystems.

This illustrates how competition governance is increasingly moving beyond conventional software markets toward infrastructure governance.

12. Essential-Facilities Doctrine and AI

The essential-facilities concept could become relevant where a particular resource is indispensable and cannot reasonably be replicated.

Potential examples could include:

• uniquely valuable datasets;
• specialised computing infrastructure;
• proprietary interoperability standards;
• critical APIs;
• certain AI infrastructure interfaces.

However, not every valuable AI resource should be classified as an essential facility.

Competition law must balance:

Access for competitors

against

incentives to invest and innovate.

Overly broad access obligations could reduce incentives for firms to develop new infrastructure.

13. Algorithmic Collusion

AI creates a particularly difficult cartel problem.

Traditional cartel:

Human executives → agreement → coordinated prices.

AI cartel possibility:

Firms → autonomous agents → common information → algorithmic coordination.

Potential mechanisms include:

• common pricing algorithms;
• common data providers;
• common optimisation systems;
• AI agents monitoring competitors;
• automated responses to competitor prices.

The OECD has specifically identified common pricing software and common model providers as potential sources of increased algorithmic-collusion risks.

Future competition authorities may therefore need algorithmic auditing capabilities.

14. Agentic AI and Competition Law

The next generation of AI may involve agentic systems capable of:

• negotiating contracts;
• purchasing goods;
• selecting suppliers;
• changing prices;
• allocating advertising budgets;
• managing inventories.

This creates a new attribution problem.

Question

If an AI agent independently makes an anticompetitive decision, who is legally responsible?

Potentially relevant parties include:

• the company deploying the agent;
• the AI developer;
• the platform provider;
• the data provider;
• the human decision-maker.

Future competition law will probably need clearer principles concerning human control, foreseeable algorithmic conduct and corporate responsibility.

The OECD has identified attribution of liability and agentic AI as areas requiring further research.

15. Competition Between AI Models

Competition should not be measured only through price.

Relevant dimensions include:

• accuracy;
• reliability;
• speed;
• safety;
• privacy;
• interoperability;
• customisation;
• energy efficiency;
• model openness;
• developer ecosystem.

A model may be offered at zero monetary price while still generating competitive advantages through:

• data collection;
• ecosystem lock-in;
• advertising;
• complementary services.

Therefore, non-price competition becomes central.

16. Innovation Competition

In intelligence-centric markets, the most important competitive harm may occur before consumers experience higher prices.

A dominant firm may reduce:

• technological experimentation;
• independent research;
• alternative models;
• start-up formation;
• innovation incentives.

Accordingly, competition authorities should consider innovation foreclosure.

The OECD's 2026 empirical work reports that concentration in AI innovation is associated with higher sales concentration and that AI start-ups are frequently acquired by large incumbents, while also finding substantial dynamism in the start-up ecosystem.

17. Competition and AI Start-Ups

AI start-ups often depend upon incumbent infrastructure.

For example:

Start-up → cloud provider → GPU provider → foundation model → application market

This can create dependency relationships.

Potential competition problems include:

• discriminatory cloud pricing;
• preferential treatment of affiliated models;
• exclusive contracts;
• restrictions on model portability;
• acquisition of emerging competitors;
• access restrictions to essential data.

Competition governance must therefore protect potential competition, not merely existing market shares.

18. Data Sharing and Competition

Data sharing can have two opposite effects.

Pro-competitive

Data sharing may:

• reduce entry barriers;
• improve interoperability;
• promote innovation;
• enable smaller competitors.

Anti-competitive

Data sharing can also:

• facilitate cartel coordination;
• expose commercially sensitive information;
• strengthen dominant platforms;
• create discriminatory access.

Therefore:

Data sharing is neither automatically pro-competitive nor automatically anticompetitive.

Its competitive effect depends on the nature of the data, participants, purpose, timing and market structure.

19. Competition Governance Model for the Future

A comprehensive framework can be represented as:

Market Monitoring
↓
Identification of AI Bottlenecks
↓
Market Definition + Ecosystem Analysis
↓
Merger Screening
↓
Dominance/Monopolisation Analysis
↓
Algorithmic Conduct Investigation
↓
Interoperability / Access Assessment
↓
Ex-Ante Regulation Where Necessary
↓
Remedies + Continuous Monitoring

This represents a shift from a purely reactive model to continuous competition governance.

20. Role of Competition Authorities

Future competition authorities will require capabilities beyond traditional economic analysis.

They will need:

Technical expertise

• machine learning;
• AI architecture;
• cloud infrastructure;
• data engineering.

Algorithmic auditing

Authorities should be capable of examining:

• training data;
• model outputs;
• ranking systems;
• pricing algorithms;
• recommendation systems.

Economic modelling

They should evaluate:

• network effects;
• switching costs;
• multi-sided markets;
• dynamic competition;
• innovation incentives.

International cooperation

AI markets are inherently cross-border.

Cooperation among:

• European Commission;
• U.S. agencies;
• UK CMA;
• China's SAMR;
• India's CCI;
• OECD and other international bodies

can reduce inconsistent enforcement and improve technical expertise.

The OECD itself emphasises cross-border cooperation and sustained monitoring as important for AI competition governance.

21. Future Remedies

Competition remedies may evolve beyond fines.

Structural remedies

• divestiture;
• separation of business units.

Behavioural remedies

• non-discrimination;
• fair access;
• interoperability;
• prohibition of tying.

Technical remedies

• API access;
• data portability;
• interoperability standards;
• switching mechanisms.

Merger remedies

• licensing;
• access commitments;
• supply obligations;
• firewall arrangements.

Monitoring remedies

• independent compliance monitors;
• algorithmic audits;
• reporting requirements.

The remedy must correspond to the theory of harm rather than automatically imposing structural separation.

22. Challenges for Competition Law

22.1 Rapid technological change

A market investigation can take years while AI technology may change within months.

22.2 Difficult market definition

AI services may overlap across:

• search;
• software;
• cloud;
• advertising;
• productivity;
• consumer services.

22.3 Dynamic competition

Today's dominant model may not remain technologically superior.

22.4 False positives

Aggressive intervention could prevent efficient integration and innovation.

22.5 False negatives

Delayed enforcement could allow a temporary advantage to become an entrenched monopoly.

This creates the central regulatory dilemma:

Intervene early enough to preserve competition, but not so early that legitimate technological innovation is suppressed.

23. India-Specific Perspective

For India, intelligence-centric competition governance would operate principally through the Competition Act, 2002, administered by the Competition Commission of India, alongside emerging digital and data regulation.

Important issues include:

• AI-platform dominance;
• cloud concentration;
• digital public infrastructure;
• interoperability;
• data access;
• algorithmic pricing;
• digital advertising;
• AI-enabled financial services;
• app ecosystems;
• merger control involving technology start-ups.

India's particular challenge is to preserve both:

innovation + digital inclusion + competitive markets.

The Competition Commission may increasingly need technical capacity to investigate AI-enabled conduct rather than relying solely on conventional market-share analysis.

24. Core Legal Principles Emerging

The future framework can be condensed into ten principles:

1. AI does not displace conventional competition law.
2. Technology-neutral principles should remain the foundation.
3. Data can constitute an important competitive input.
4. Compute can become a strategic bottleneck.
5. Algorithmic conduct remains attributable to competition-law principles.
6. AI partnerships require examination for foreclosure and coordination.
7. Merger control must account for innovation and potential competition.
8. Interoperability may become an important competitive remedy.
9. Ex-ante regulation may complement—not replace—antitrust enforcement.
10. Continuous monitoring is essential because AI markets evolve rapidly.

25. Conclusion

Competition law in intelligence-centric economies will increasingly become a system of continuous market governance rather than merely a mechanism for punishing completed anticompetitive conduct.

The central competition question will move from:

“Does this firm have a large market share?”

toward a broader inquiry:

“Does this firm control a critical intelligence, infrastructure, data, compute or distribution bottleneck in a manner that prevents effective competition, innovation or market entry?”

The most important future competition issues are therefore likely to concern AI infrastructure, compute, data, foundation models, cloud services, algorithms, autonomous agents, interoperability, ecosystem leverage and acquisitions of potential competitors.

At the same time, intervention must remain sensitive to the fact that AI markets are still technologically dynamic. The OECD's 2026 research presents a mixed picture: foundation-model competition has shown substantial dynamism, while structural risks around compute, data, skills, vertical integration and concentration remain significant.

Thus, the future of competition governance is likely to be a hybrid model:

Traditional Antitrust + Ex-Ante Digital Regulation + AI/Algorithmic Auditing + Infrastructure Access Rules + Dynamic Merger Control + International Cooperation.

Key Case Laws for Examination

1. Google Shopping — European Commission
2. United States v. Google LLC (Search)
3. Amazon Marketplace Pricing Algorithms — United States
4. Trod Ltd / GB Eye Ltd — United Kingdom
5. Nvidia/Mellanox — European Commission & SAMR
6. Illumina/GRAIL — European Union
7. Microsoft — European Commission
8. Bronner — CJEU
9. Intel — CJEU
10. Qualcomm — European Commission