Competition Law and Future Regulation of Memory-Based Competition

1. Introduction

Memory-based competition describes competitive conditions in which firms gain an advantage from their ability to accumulate, retain, retrieve, combine and continuously learn from historical information. The relevant “memory” may consist of:

• historical search queries and clicks;
• consumer purchasing and browsing histories;
• seller and transaction data;
• pricing histories;
• advertising-response data;
• customer preferences and behavioural profiles;
• technical logs and usage data;
• AI training and interaction data;
• historical market intelligence; and
• institutional knowledge embedded in algorithms and models.

This creates a competition problem different from traditional control over physical assets. A firm may become stronger not merely because it has more users, but because each additional interaction improves its information stock, which improves its algorithm, which attracts more users, which produces still more information.

The European Commission has expressly recognised this dynamic in relation to Google Search: Google's large-scale access to search data can improve its search algorithm and thereby reinforce its position. In July 2026, the Commission adopted measures requiring Google to facilitate access to certain search data for eligible competing search engines.

Thus, future competition law is likely to treat organisational memory, data accumulation and algorithmic learning as potential sources of durable competitive advantage, while still distinguishing legitimate innovation from exclusionary conduct.

2. Meaning of Memory-Based Competition

Memory-based competition can be represented as:

Data collection → retention → learning → prediction → improved service → increased usage → additional data → stronger learning

This creates a possible data-learning feedback loop.

For example:

Platform A has 100 million historical transactions → its algorithm learns consumer preferences → recommendations become more accurate → consumers use Platform A more frequently → Platform A collects more transactions → the algorithm becomes still more accurate.

A competing entrant may possess equally good technology but lack the historical information necessary to reproduce the same level of performance.

The competitive concern therefore becomes:

Is the firm's advantage the product of competition on the merits, or has control over accumulated memory become an exclusionary barrier to competition?

3. Why Memory Is Different From Ordinary Data

Not every database creates a competition problem.

The important distinction is between:

Ordinary data

A dataset may be:

• readily available;
• commercially obtainable;
• easily reproduced;
• short-lived;
• of limited strategic value.

Strategic memory

Strategic memory may be:

• accumulated over many years;
• continuously updated;
• difficult for entrants to replicate;
• generated automatically by network activity;
• combined across different services;
• used to train algorithms or AI models;
• capable of predicting future behaviour.

The historical dimension is therefore critical.

A competitor may technically obtain today's data but still lack the longitudinal dataset necessary to reproduce the incumbent's predictive capabilities.

4. Competition-Law Issues Created by Memory-Based Markets

A. Data accumulation as a barrier to entry

Large historical datasets can create substantial entry barriers.

An entrant may need:

• millions of observations;
• years of behavioural history;
• transaction records;
• labelled data;
• user feedback;
• search histories;
• error and correction records.

Consequently, the relevant competitive asset may not simply be “data”, but data accumulated over time.

B. Feedback loops

The most important concern is the data-network-learning feedback loop.

A simplified model is:

More users

↓

More data

↓

Better algorithm

↓

Better product

↓

More users

↓

More data

This may produce increasing returns to scale.

Competition authorities may therefore need to ask whether an incumbent's advantage is:

1. temporary;
2. replicable;
3. innovation-based; or
4. structurally self-reinforcing.

C. Data combination across ecosystems

A firm operating several services may possess a particularly powerful form of memory.

For example:

Search + browser + maps + video + payments + advertising + cloud

may generate a much richer behavioural profile than any individual service could generate.

The competition issue arises when information collected in one market is used to strengthen market power in another.

The German Bundeskartellamt's Google proceeding specifically concerned combinations of personal data originating from different Google services and other sources. The authority treated data collection, processing and combination as important elements of digital market power.

5. Relevant Legal Framework

A. Article 102 TFEU

Article 102 can address memory-based competition where a dominant undertaking uses accumulated informational advantages to:

• exclude competitors;
• foreclose access;
• discriminate;
• self-preference;
• impose unfair conditions;
• leverage dominance into neighbouring markets.

The crucial point is that possession of valuable data is not itself unlawful.

The competition-law question is what the dominant undertaking does with that informational advantage.

B. EU Digital Markets Act

The DMA moves beyond traditional ex-post abuse analysis and establishes obligations for designated gatekeepers.

The DMA now contains important mechanisms concerning:

• data portability;
• business-user data access;
• search-data access;
• interoperability;
• restrictions on combining certain data;
• consumer profiling transparency.

The Commission states that the DMA's data-access provisions are intended to open digital markets and give businesses and users access to valuable data.

This is particularly important for memory-based competition because data-access remedies can attack the accumulation advantage directly.

C. German Competition Act — Section 19a GWB

Germany's Section 19a GWB provides enhanced abuse control for undertakings of paramount significance across markets.

The Bundeskartellamt has used this framework against major digital firms including:

• Google;
• Meta;
• Amazon;
• Apple; and
• Microsoft. 

This approach is significant because it permits intervention against certain digital-market practices before conventional market-definition analysis becomes sufficient to capture the competitive problem.

D. Indian Competition Law

In India, memory-based competition can potentially engage:

• Competition Act 2002, Section 3 — anti-competitive agreements;
• Section 4 — abuse of dominant position;
• Sections 5 and 6 — combinations;
• Competition Commission of India investigations involving digital platforms;
• data-driven leveraging and self-preferencing theories where the statutory requirements are satisfied.

India's future regulatory challenge will be determining when accumulated data constitutes a competitively significant asset without treating every successful data-driven business model as anticompetitive.

6. Six Major Case Laws

1. Meta Platforms Inc. v Bundeskartellamt — Case C-252/21

Court: Court of Justice of the European Union
Year: 2023

This is one of the most important cases for memory-based competition.

The case concerned Meta's combination and processing of personal data obtained through Facebook and other sources.

The CJEU confirmed that a competition authority can consider data-protection issues when assessing whether conduct by a dominant undertaking constitutes abuse, subject to the required institutional coordination with data-protection authorities.

Importance for memory-based competition

The case demonstrates that:

Data accumulation → behavioural knowledge → market power

cannot necessarily be separated into independent regulatory compartments.

It establishes an important foundation for future cases involving firms whose competitive advantage depends upon the continuous accumulation and combination of personal information.

2. Bundeskartellamt — Google Data Processing Proceeding

Case: B7-70/21
Authority: German Federal Cartel Office
Year: 2023

The Bundeskartellamt required Google to provide users with greater choice concerning the combination of personal data between Google's different services.

The authority specifically recognised that the collection, processing and combination of data form an important foundation of the market power of large digital companies.

Importance

This case is particularly relevant to the concept of institutional memory.

Google could potentially obtain a more comprehensive understanding of users by combining information from multiple services.

Competition law therefore becomes concerned with:

cross-service data combination + ecosystem power + accumulated informational advantage.

3. Google Shopping — Case C-48/22 P

Court: CJEU
Judgment: 10 September 2024

The case concerned Google's preferential treatment of its own comparison-shopping service in general search results.

The CJEU examined whether Google's conduct could disadvantage competing specialised search services and confirmed the relevance of foreclosure effects and causal analysis in assessing the abuse.

Importance for memory-based competition

Search engines accumulate enormous historical information concerning:

• queries;
• clicks;
• rankings;
• user behaviour;
• search results;
• interactions.

Preferential placement can therefore reinforce the ecosystem that generates the underlying information.

The case illustrates a broader principle:

Control over an information-generating gateway can reinforce the competitive advantage produced by accumulated information.

4. Amazon Marketplace — European Commission

Case: AT.40462 — Amazon Marketplace

The Commission investigated Amazon's use of non-public information obtained from independent sellers on its marketplace.

The Commission's preliminary concerns involved Amazon allegedly using large quantities of non-public seller data for its own retail business while simultaneously competing with those sellers.

The resulting commitments included restrictions on Amazon Retail's use of non-public seller data and measures concerning the Buy Box and logistics.

Importance

This case is central to competitive memory generated by an intermediary.

The platform can observe:

• product sales;
• quantities;
• prices;
• demand;
• customer behaviour;
• seller performance.

If the platform uses that accumulated knowledge to compete against the very businesses generating the information, the intermediary can potentially transform marketplace memory into competitive advantage.

5. Amazon — Bundeskartellamt, Section 19a GWB

The Bundeskartellamt established in 2022 that Amazon possessed paramount significance for competition across markets. Germany's Federal Court of Justice upheld that determination in April 2024.

The authority has also examined Amazon's price-control mechanisms and algorithms, including whether Amazon can influence prices offered by third-party sellers.

Importance

This case demonstrates the movement from:

static market share analysis

toward:

ecosystem + data + algorithm + accumulated information + behavioural control.

For memory-based competition, the key concern is that a platform can continuously learn from the activity of its ecosystem and subsequently use that knowledge to shape competitive conditions.

6. Napp Pharmaceutical Holdings Ltd v Director General of Fair Trading

Court: UK Court of Appeal
Year: 2002

Napp involved sustained-release morphine products and exclusionary pricing.

The court recognised significant barriers to entry, including Napp's established reputation and strong first-mover advantages. Napp had very high market shares in relevant segments.

Importance for memory-based competition

Although Napp was not a digital-data case, it provides an important conceptual precedent.

Competitive advantage can accumulate through:

• experience;
• reputation;
• established relationships;
• historical market presence;
• customer familiarity.

Modern AI and digital markets can transform this traditional first-mover advantage into algorithmic memory.

Thus:

traditional accumulated market knowledge → modern accumulated digital memory.

The case should therefore be understood as an analogy rather than a direct precedent concerning AI memory.

7. Comparative Significance of the Cases

8. Future Regulation of Memory-Based Competition

A. Data portability

One possible regulatory response is to enable users and businesses to transfer their historical information between platforms.

However, simple portability may be insufficient.

If the incumbent possesses ten years of processed behavioural information, giving an entrant a raw copy of today's data may not reproduce the incumbent's accumulated learning.

Future regulation may therefore distinguish:

raw data portability

from

meaningful functional portability.

9. Historical Search-Data Access

The EU has already moved in this direction.

In July 2026, the European Commission adopted binding measures concerning Google's sharing of search data with eligible competing search engines. The measures concern access to search information such as ranking, query, click and view data, subject to safeguards including anonymisation.

This is particularly significant for memory-based competition because search competitors may otherwise lack the historical information necessary to improve their own services.

The regulatory theory is essentially:

Incumbent's accumulated search memory

↓

Algorithmic improvement

↓

Entrenched position

↓

Controlled access to relevant information

↓

Potential competitive disadvantage for entrants

10. Data Interoperability

Future competition law may require dominant platforms to permit interoperability concerning:

• APIs;
• transaction histories;
• identity systems;
• social graphs;
• search information;
• recommendation signals;
• technical performance information.

The objective would not necessarily be to give competitors unrestricted access to proprietary information.

Instead, regulation may establish:

limited + secure + proportionate + non-discriminatory access.

11. AI Training Data

AI introduces a new dimension.

A foundation model can possess enormous institutional memory derived from:

• training datasets;
• user interactions;
• corrections;
• feedback;
• evaluations;
• retrieval systems;
• reinforcement signals.

A dominant AI firm may therefore have a competitive advantage not merely because of computing power but because its model has learned from a historically larger information base.

Future competition cases may ask:

Who owns the learning advantage?

Can competitors obtain sufficient training information?

Can users transfer interaction histories?

Can a platform use customer data to improve a competing downstream service?

Does exclusive access to high-value data create durable market power?

12. Algorithmic Memory and Tacit Coordination

Memory-based competition can also facilitate coordination.

Suppose competing algorithms retain:

• historical prices;
• competitor responses;
• inventory levels;
• demand patterns;
• promotional behaviour.

Each algorithm can learn how competitors react.

The result could be:

Past observation → algorithmic learning → prediction → strategic response → new observation → further learning.

Competition authorities may therefore have to examine whether algorithmic systems facilitate:

• tacit coordination;
• parallel pricing;
• personalised pricing;
• discriminatory pricing;
• market allocation;
• reduced competitive uncertainty.

The difficult legal question will be distinguishing independent intelligent adaptation from unlawful coordination.

13. Memory-Based Self-Preferencing

A platform can possess two roles:

1. information infrastructure, and
2. competitor.

This creates a structural conflict.

For example:

Marketplace → collects seller data → learns which products are successful → launches competing products → uses accumulated marketplace knowledge.

The Amazon Marketplace investigation demonstrates why this issue has become central to modern competition policy.

Future regulation may therefore impose information-use firewalls between platform infrastructure and downstream competitive operations.

14. Memory-Based Mergers

Traditional merger control often focuses on:

• market share;
• turnover;
• concentration;
• overlaps.

Future merger analysis may additionally examine:

Historical-data concentration

Will the transaction combine two unique datasets?

Learning concentration

Will one company obtain substantially superior training information?

Feedback-loop effects

Will combined data improve the merged firm's algorithm?

Entry barriers

Could rivals reproduce the merged firm's informational advantage?

Future markets

Could a seemingly small data acquisition facilitate dominance in an emerging AI or digital market?

The FTC has already examined previously unreported acquisitions by major technology companies, demonstrating the importance of scrutinising transactions involving large technology firms beyond conventional headline merger thresholds.

15. Remedies for Memory-Based Competition

Future competition authorities could use several remedies.

1. Data-access remedies

Permit competitors to obtain specified datasets under controlled conditions.

2. Data portability

Allow users and businesses to transfer historical information.

3. Interoperability

Require technical systems to communicate with competing services.

4. Data-use restrictions

Prevent a platform from using confidential information obtained from competitors.

5. Information firewalls

Separate marketplace infrastructure data from the platform's own competing business.

6. Algorithmic auditing

Examine whether historical information systematically produces exclusionary outcomes.

7. Data-combination restrictions

Prevent unrestricted combination of datasets across services where that combination reinforces market power.

8. Structural remedies

In exceptional circumstances, regulators could consider separation of infrastructure and competitive businesses.

16. Key Challenges for Regulators

A. Data is not automatically an essential facility

A dataset may be valuable without being legally indispensable.

Therefore, authorities should avoid assuming:

valuable data = essential facility = mandatory access.

The traditional legal requirements for intervention remain important.

B. Privacy and competition may conflict

Opening data to competitors can improve competition but potentially increase:

• privacy risks;
• surveillance;
• cybersecurity risks;
• re-identification risks.

Therefore, competition remedies should incorporate:

data minimisation + anonymisation + security + purpose limitation.

C. Dynamic efficiency

A company may invest heavily in collecting and processing information.

If competitors receive unrestricted access to the resulting knowledge, investment incentives could theoretically decline.

Regulators therefore face a balance between:

contestability

and

innovation incentives.

17. Future Legal Test for Memory-Based Competition

A useful analytical framework can be developed around M-E-M-O-R-Y:

M — Market Power

Does the undertaking possess substantial and durable market power?

E — Exclusivity

Is the relevant historical information uniquely or disproportionately controlled?

M — Memory Accumulation

Has the advantage resulted from long-term accumulation of information?

O — Opportunity for Replication

Can competitors realistically reproduce the informational advantage?

R — Reinforcement

Does accumulated information reinforce the incumbent's market position?

Y — Yielded Competitive Harm

Has the use or withholding of the accumulated information produced exclusionary effects?

This framework can assist regulators without treating data accumulation itself as unlawful.

18. Emerging Concept: The Right to Competitive Memory

A potentially important future regulatory concept is a right to competitive memory.

This would not mean that competitors automatically own another company's information.

Rather, it could mean that in designated digital ecosystems:

• users can transfer historical information;
• businesses can access information generated through their own activity;
• competitors can receive specified non-personal datasets;
• dominant platforms cannot indefinitely monopolise competitively indispensable information;
• interoperability prevents historical information from becoming a permanent switching barrier.

The EU's developing search-data-access regime provides an important contemporary example of this regulatory direction.

19. Relationship Between Memory and Market Power

The traditional model was:

Market share → market power → possible abuse

The emerging digital model is increasingly:

Data → learning → better service → users → more data → stronger learning → market power

Therefore, competition law may increasingly need to examine how market power is reproduced over time, rather than merely measuring market share at a single point.

This is the central significance of memory-based competition.

20. Conclusion

Memory-based competition represents a shift from competition based principally on current assets toward competition based on accumulated informational experience.

The most important legal concern is not that companies remember information. It is whether accumulated memory becomes a self-reinforcing competitive barrier that competitors cannot reasonably reproduce.

The cases involving Meta, Google and Amazon, together with the older first-mover principles illustrated by Napp, show the developing legal trajectory: competition authorities increasingly examine how information, ecosystem integration, algorithms and accumulated advantages affect market power.

Future regulation is therefore likely to focus on:

data accumulation + interoperability + portability + algorithmic learning + information-use restrictions + merger control + access remedies.