Hyperscaler Dominance in International Cloud Computing Markets

1. Introduction

Hyperscalers are very large cloud-computing providers capable of operating enormous global networks of data centres, computing infrastructure, storage, networking, databases, artificial-intelligence services and related digital platforms. The principal examples are Amazon Web Services (AWS), Microsoft Azure and Google Cloud, although other major providers operate significant regional or specialised infrastructure.

Hyperscaler dominance raises distinctive competition-law problems because cloud markets exhibit substantial economies of scale, high switching costs, network and ecosystem effects, data advantages, technical interoperability issues, and vertical integration. A hyperscaler may simultaneously operate:

• infrastructure-as-a-service (IaaS);
• platform-as-a-service (PaaS);
• software-as-a-service (SaaS);
• AI and machine-learning services;
• app marketplaces;
• databases;
• cybersecurity services;
• enterprise software;
• advertising or e-commerce platforms; and
• proprietary hardware and networking infrastructure.

The competition concern is therefore not merely that one provider has a large market share. The deeper question is whether hyperscaler power allows the provider to foreclose rivals, raise switching costs, discriminate against competing services, leverage dominance from one market into another, or control essential technological inputs.

2. Meaning of Hyperscaler Dominance

Hyperscaler dominance can be understood as a situation in which one or a small number of cloud providers possess substantial and durable market power over cloud-computing services.

Dominance may arise through several mutually reinforcing advantages:

A. Economies of scale

Cloud infrastructure requires enormous capital expenditure for:

• data centres;
• servers;
• GPUs and accelerators;
• fibre networks;
• storage systems;
• cooling;
• electricity;
• cybersecurity; and
• specialised networking equipment.

A very large provider can spread these fixed costs over millions of customers.

B. Economies of scope

A hyperscaler can offer hundreds of interconnected services.

For example:

cloud infrastructure → database → AI → cybersecurity → analytics → enterprise software → marketplace.

This can make the provider substantially more attractive than a smaller cloud-only competitor.

C. Switching costs

Customers may invest heavily in:

• proprietary cloud architecture;
• APIs;
• databases;
• applications;
• data pipelines;
• identity systems;
• security configurations; and
• cloud-specific development tools.

Moving to another provider can therefore be technically and financially expensive.

D. Data and technological advantages

Large providers receive enormous quantities of operational data concerning:

• workload performance;
• customer usage;
• network behaviour;
• resource demand;
• security threats;
• AI workloads; and
• application performance.

This can reinforce technological advantages.

E. Ecosystem effects

Cloud services increasingly function as ecosystems rather than individual products.

A customer purchasing cloud infrastructure may subsequently purchase the same provider's:

• AI model services;
• analytics;
• database;
• cybersecurity;
• productivity software;
• developer tools; and
• marketplace applications.

This can create ecosystem lock-in.

3. Relevant Competition-Law Markets

The relevant market cannot automatically be defined as simply "cloud computing."

Potentially distinct markets include:

1. IaaS;
2. PaaS;
3. SaaS;
4. public cloud;
5. private cloud;
6. hybrid cloud;
7. cloud storage;
8. cloud databases;
9. cloud cybersecurity;
10. AI cloud infrastructure;
11. GPU cloud computing;
12. cloud networking;
13. cloud marketplace services;
14. cloud-management tools; and
15. specialised cloud services for particular industries.

The market-definition exercise becomes particularly important where a hyperscaler argues that its competitors include every form of IT infrastructure.

A competition authority may instead ask whether a customer could realistically substitute:

on-premises infrastructure, private cloud, another public cloud, or specialised cloud services.

4. Sources of Hyperscaler Market Power

4.1 High capital requirements

Building hyperscale infrastructure requires billions in investment.

Consequently, a potential entrant cannot easily replicate an incumbent's global infrastructure.

This creates a structural barrier to entry.

4.2 Cloud switching costs

A customer may have thousands of applications and databases configured for a particular cloud.

Switching may require:

• rewriting software;
• migrating databases;
• changing APIs;
• retraining employees;
• redesigning security systems;
• renegotiating contracts; and
• temporarily operating duplicate infrastructure.

Thus, even where several providers technically exist, contestability may be weak.

4.3 Data-egress charges

One major competition concern is the cost of transferring data from one cloud provider to another.

Suppose:

• Provider A charges little for bringing data into its ecosystem;
• Provider A charges substantially for transferring data out;
• Provider B offers a cheaper alternative.

A customer may nevertheless remain with A because transferring massive datasets is expensive.

The resulting effect can be:

low entry cost + high exit cost = customer lock-in.

5. Interoperability and Portability

Interoperability is central to cloud competition.

A dominant provider can potentially weaken competition by making it difficult for customers to use:

• competing clouds;
• third-party databases;
• rival AI systems;
• independent cybersecurity tools; or
• multi-cloud management platforms.

Competition authorities therefore increasingly examine:

• APIs;
• technical standards;
• data portability;
• interoperability;
• application portability;
• contractual restrictions; and
• switching mechanisms.

6. Self-Preferencing

A hyperscaler may compete with companies that depend upon its infrastructure.

This creates a potential vertical conflict of interest.

For example:

Hyperscaler operates cloud marketplace → independent software companies depend upon marketplace → hyperscaler offers competing software → hyperscaler possesses information about competitors.

The provider could potentially:

• favour its own products;
• rank its services preferentially;
• impose discriminatory terms;
• access competitively sensitive information;
• bundle its products; or
• disadvantage rival applications.

This resembles concerns previously examined in other digital-platform markets.

7. Bundling and Tying

A dominant cloud provider might condition access to one service upon purchasing another.

For example:

Cloud infrastructure + cybersecurity
Cloud infrastructure + database
Cloud infrastructure + productivity software
Cloud infrastructure + AI services

Bundling is not inherently unlawful.

The competition question is whether the practice:

1. involves a dominant undertaking;
2. concerns separate products or services;
3. coerces or materially incentivises customers to purchase the tied product;
4. forecloses competitors; and
5. lacks sufficient objective justification.

8. Exclusive-Use and Minimum-Spend Commitments

Enterprise cloud contracts may contain substantial:

• minimum-spend commitments;
• volume discounts;
• reserved-capacity arrangements;
• preferential pricing;
• loyalty incentives; or
• enterprise-wide purchasing commitments.

These arrangements may produce efficiencies, including predictable demand and infrastructure planning.

However, if a dominant hyperscaler uses them to capture most of a customer's future cloud expenditure, they can potentially produce foreclosure effects.

The critical issue is often the customer's contestable share of demand.

9. Case Laws

The following cases are particularly useful for understanding the legal principles applicable to hyperscaler dominance, even though several arose outside cloud computing specifically.

Case 1: Microsoft Corp. v. Commission (Microsoft I)

Court: General Court of the European Union
Year: 2007

Principle

The Microsoft case concerned Microsoft's dominance in PC operating systems and its conduct concerning interoperability and server software.

The General Court upheld important findings concerning Microsoft's refusal to provide interoperability information and the potential exclusionary consequences.

Relevance to hyperscalers

The case is highly relevant because modern cloud competition depends heavily upon interoperability.

A hyperscaler controlling an important infrastructure layer could potentially disadvantage rivals by:

• restricting technical information;
• limiting interoperability;
• making competing products technically incompatible; or
• degrading access to interfaces.

Competition-law significance

The case demonstrates that technical interoperability can itself become a competition-law issue where a dominant undertaking controls an important technological interface.

10. Case 2: Google Shopping

Case: Google and Alphabet v European Commission
Court: General Court of the European Union
Year: 2021

Principle

The EU courts upheld the central finding that Google had abused its dominant position by favouring its comparison-shopping service in search results.

The case is particularly significant for the doctrine of self-preferencing.

Relevance to hyperscalers

A hyperscaler may operate both:

infrastructure platform + competing downstream service.

For example, it may provide cloud infrastructure while simultaneously offering:

• database products;
• cybersecurity;
• AI services;
• analytics;
• developer tools.

If the infrastructure provider gives its own downstream service preferential treatment, the Google Shopping reasoning provides an important analytical framework.

Competition concern

The fundamental concern is:

control over infrastructure can potentially be transformed into an advantage in adjacent competitive markets.

11. Case 3: Google Android

Case: Google and Alphabet v European Commission
Court: General Court of the European Union
Year: 2022

Principle

The case examined Google's contractual practices concerning Android and the relationship between different Google services.

The Court recognised the importance of examining contractual arrangements that can reinforce dominance and restrict competing platforms.

Relevance to cloud markets

Cloud ecosystems similarly involve multiple interconnected services.

A hyperscaler could potentially use contractual arrangements to reinforce its position across:

• infrastructure;
• operating environments;
• applications;
• search;
• AI;
• databases; and
• enterprise software.

Significance

The case demonstrates that competition analysis should consider ecosystem reinforcement, rather than examining every contractual restriction in isolation.

12. Case 4: Intel v Commission

Case: Intel Corp. v European Commission
Court: Court of Justice of the European Union
Year: 2017

Principle

The case concerned conditional rebates granted by a dominant undertaking.

The Court held that where the dominant undertaking submits evidence that its conduct was incapable of restricting competition, the authority must examine the circumstances and potential foreclosure effects.

Important considerations include:

• dominant position;
• market coverage;
• duration;
• size of rebates;
• conditions attached; and
• ability to foreclose an equally efficient competitor.

Relevance to hyperscalers

Cloud providers frequently use:

• volume discounts;
• committed-spend discounts;
• reserved capacity;
• enterprise pricing;
• credits; and
• preferential contractual terms.

The Intel framework is therefore highly relevant when determining whether cloud discounts are legitimate efficiencies or exclusionary loyalty mechanisms.

13. Case 5: Bronner v Mediaprint

Court: Court of Justice of the European Union
Year: 1998

Principle

Bronner established the strict conditions associated with refusal-to-deal and essential-facilities reasoning.

A refusal to provide access to infrastructure is not automatically abusive.

The classic requirements include circumstances involving:

1. indispensability;
2. inability to reproduce the facility realistically;
3. absence of a viable alternative; and
4. potential elimination of effective competition.

Relevance to cloud computing

A hyperscaler may control infrastructure that rivals or customers claim is indispensable.

Examples could include:

• specialised GPU capacity;
• unique cloud infrastructure;
• proprietary technical interfaces;
• critical data;
• specialised AI infrastructure.

However, Bronner warns against treating every commercially important infrastructure facility as an "essential facility."

14. Case 6: Slovak Telekom v Commission

Court: Court of Justice of the European Union
Year: 2021

Principle

The case concerned access to telecommunications infrastructure controlled by a dominant undertaking.

The Court examined exclusionary conduct involving access conditions and the relationship between the incumbent infrastructure provider and downstream competitors.

Relevance to hyperscalers

Cloud markets have a comparable structural characteristic:

infrastructure owner + downstream competitor.

A hyperscaler may control infrastructure upon which competing downstream providers depend.

This creates risks involving:

• discriminatory access;
• margin squeeze;
• foreclosure;
• discriminatory pricing;
• technical restrictions; and
• leveraging infrastructure dominance.

15. Case 7: Bronner and the Digital Infrastructure Analogy

Although Bronner itself is not a cloud case, its significance is especially important for hyperscaler disputes.

Cloud infrastructure differs from a traditional physical essential facility because:

• alternative cloud infrastructure may technically exist;
• customers can sometimes multi-cloud;
• private infrastructure can sometimes be built;
• specialised providers may exist.

Therefore, a claimant would need to demonstrate more than:

"The hyperscaler's infrastructure is commercially important."

It may need to establish that access is indispensable for effective competition.

16. Case 8: Deutsche Telekom v Commission

Court: Court of Justice of the European Union
Year: 2010

Principle

The case concerned a margin squeeze involving access pricing.

The Court recognised that a vertically integrated dominant undertaking can abuse its position where the relationship between upstream and downstream prices effectively prevents efficient downstream competition.

Cloud relevance

Imagine:

Hyperscaler controls cloud infrastructure upstream + competes in downstream cloud services.

If the provider structures internal or external access prices in a way that makes downstream competition commercially impossible, margin-squeeze principles could become relevant.

This is especially important where independent cloud service providers depend upon infrastructure controlled by a hyperscaler.

17. International Dimension

Hyperscaler dominance is inherently international.

A cloud provider can have:

• data centres in multiple countries;
• customers worldwide;
• globally negotiated contracts;
• centralised APIs;
• global pricing structures;
• cross-border data flows; and
• global developer ecosystems.

Consequently, competition authorities in different jurisdictions may investigate substantially similar conduct.

Potential jurisdictions include:

• European Union;
• United Kingdom;
• United States;
• India;
• Australia;
• Japan;
• South Korea; and
• other major digital economies.

This creates the possibility of parallel competition investigations.

18. Cross-Border Enforcement Problem

Suppose a hyperscaler adopts a global contractual policy.

The European Commission may consider it under EU competition law.

The UK's Competition and Markets Authority may consider its UK effects.

The US authorities may consider the same conduct under the Sherman Act or Clayton Act.

India's Competition Commission may assess its Indian market effects.

The same business practice may therefore generate different legal conclusions because jurisdictions use different approaches to:

• market definition;
• dominance;
• foreclosure;
• essential facilities;
• vertical restraints;
• consumer welfare;
• innovation;
• data;
• privacy; and
• digital ecosystems.

19. AI and Hyperscaler Dominance

The emergence of generative AI makes hyperscaler power even more significant.

AI development requires:

• GPUs;
• high-performance computing;
• data storage;
• networking;
• specialised chips;
• model-hosting infrastructure;
• inference capacity;
• electricity; and
• cloud orchestration.

The major cloud providers increasingly control access to these inputs.

This creates a potential cloud-AI dominance loop:

Hyperscale cloud
↓
Compute concentration
↓
AI infrastructure advantage
↓
More AI customers
↓
More cloud demand
↓
Greater economies of scale
↓
Stronger hyperscaler position.

Competition authorities may therefore increasingly investigate cloud markets together with AI markets.

20. Cloud Marketplace Concerns

Cloud marketplaces create another potential competition problem.

A hyperscaler may operate:

cloud infrastructure + application marketplace + competing applications.

Possible concerns include:

• preferential ranking;
• discriminatory commissions;
• self-preferencing;
• access restrictions;
• tying;
• discriminatory technical certification;
• use of marketplace data;
• restrictions on alternative payment mechanisms; and
• exclusion of competing cloud services.

This resembles broader digital-platform competition concerns.

21. Data Advantage

Cloud providers have access to enormous amounts of operational information.

Potentially valuable information includes:

• workload characteristics;
• resource utilisation;
• application performance;
• customer demand;
• security events;
• AI workloads;
• traffic patterns.

If a hyperscaler uses competitively sensitive information obtained from customers or marketplace participants to compete against them, competition concerns may arise.

The crucial distinction is between:

legitimate infrastructure optimisation

and

use of privileged information for competitive foreclosure.

22. Multi-Cloud Competition

Multi-cloud strategies can reduce hyperscaler dominance.

A customer may distribute workloads across:

• AWS;
• Azure;
• Google Cloud;
• regional cloud providers; and
• private infrastructure.

Multi-cloud competition can reduce dependence on one provider.

However, its effectiveness is limited where:

• applications are cloud-specific;
• migration is expensive;
• data-egress costs are high;
• proprietary APIs are used;
• enterprise discounts encourage concentration; or
• technical interoperability is weak.

Thus, the existence of several cloud providers does not automatically establish effective competition.

23. Competition-Law Theories of Harm

The major theories of harm can be summarised as follows:

24. Merger-Control Issues

Hyperscaler dominance also creates concerns concerning acquisitions.

A large cloud provider may acquire:

• cloud-security firms;
• AI companies;
• database companies;
• observability platforms;
• cybersecurity providers;
• AI-model developers;
• GPU-cloud companies;
• cloud-management tools.

A transaction may appear small under traditional turnover thresholds while nevertheless eliminating a potentially important future competitor.

This raises the problem of nascent competition.

Competition authorities may therefore examine:

"What would this company have become if it had remained independent?"

rather than merely asking:

"What is its current market share?"

25. Regulatory Remedies

Possible remedies include:

Structural remedies

• divestiture;
• separation of business units;
• prohibition of acquisitions.

Behavioural remedies

• interoperability obligations;
• data portability;
• transparent pricing;
• restrictions on tying;
• non-discrimination obligations;
• restrictions on self-preferencing;
• access obligations.

Technical remedies

• interoperable APIs;
• common standards;
• portability mechanisms;
• standardised migration tools;
• cloud-neutral interfaces.

Contractual remedies

• restrictions on exclusivity;
• limits on minimum-spend commitments;
• transparent termination provisions;
• restrictions on anti-competitive switching charges.

26. Relationship with Data Protection and Digital Regulation

Cloud competition cannot always be analysed independently from other regulatory regimes.

Cloud services involve:

• personal data;
• cybersecurity;
• data localisation;
• confidentiality;
• cross-border transfers;
• critical infrastructure.

Consequently, competition authorities may need to coordinate with:

• data-protection authorities;
• cybersecurity regulators;
• telecommunications regulators;
• financial regulators; and
• digital-market regulators.

The central challenge is avoiding a situation in which regulatory compliance itself unintentionally reinforces hyperscaler concentration.

27. Key Legal Test

For a competition-law assessment of hyperscaler dominance, the following framework is useful:

Step 1 — Define the market

Determine whether the relevant market is:

cloud computing generally,

or a narrower market such as:

IaaS, AI compute, cloud databases, GPU cloud services, or cloud marketplaces.

Step 2 — Establish dominance

Consider:

• market shares;
• infrastructure;
• entry barriers;
• customer lock-in;
• switching costs;
• technological advantages;
• network effects;
• financial resources.

Step 3 — Identify conduct

Examine:

• tying;
• bundling;
• rebates;
• exclusivity;
• self-preferencing;
• discriminatory access;
• refusal to interoperate;
• data exploitation.

Step 4 — Assess foreclosure

Ask:

Could an equally efficient competitor realistically compete?

Step 5 — Consider efficiencies

The provider may demonstrate:

• security benefits;
• performance improvements;
• infrastructure efficiency;
• reduced costs;
• innovation;
• reliability.

Step 6 — Examine proportionality

Even legitimate technical restrictions may become problematic if a less restrictive method could achieve the same objective.

28. Core Competition-Law Principle

The central issue is not simply hyperscaler size.

Large scale can produce enormous consumer benefits:

• lower computing costs;
• greater reliability;
• global availability;
• faster innovation;
• cybersecurity;
• AI development;
• efficient infrastructure utilisation.

Competition law therefore should not penalise successful investment merely because a company becomes large.

The concern arises when:

scale becomes a mechanism for preventing customers and competitors from realistically switching to alternative suppliers.

29. Conclusion

Hyperscaler dominance in international cloud computing markets represents a particularly important modern competition-law problem because infrastructure, software, data, AI and enterprise ecosystems are increasingly converging.

The principal risks involve:

1. cloud customer lock-in;
2. high switching and migration costs;
3. data-egress barriers;
4. exclusive or minimum-spend contracts;
5. self-preferencing;
6. bundling and tying;
7. interoperability restrictions;
8. discriminatory access;
9. use of competitively sensitive customer data;
10. vertical leverage into AI and software markets;
11. foreclosure of smaller cloud providers; and
12. acquisition of emerging competitors.

The cases of Microsoft, Google Shopping, Google Android, Intel, Bronner, Slovak Telekom and Deutsche Telekom provide important doctrinal foundations for analysing these problems.