Industrial AI Platform Competition Issues

1. Introduction

Industrial AI platforms are digital systems that combine artificial intelligence with industrial data, cloud infrastructure, sensors, industrial Internet of Things (IIoT), digital twins, predictive-maintenance systems, robotics, process-control software, and analytics. They may be used in manufacturing, energy, logistics, mining, automotive production, chemicals, utilities, and infrastructure.

Competition problems arise because an industrial AI platform may simultaneously control:

• access to industrial data;
• AI models and algorithms;
• cloud and computing infrastructure;
• industrial operating systems;
• machine-control interfaces;
• digital-twin environments;
• application programming interfaces (APIs);
• predictive-maintenance tools;
• software marketplaces;
• downstream industrial applications.

The central competition-law concern is therefore vertical and ecosystem power: a company controlling an important industrial platform can potentially use its position in one layer to restrict competition in adjacent layers.

2. Why Industrial AI Platforms Are Competition-Sensitive

Industrial AI markets have several characteristics that can facilitate market power.

A. High switching costs

Factories may integrate an AI platform into:

• machinery;
• enterprise resource planning;
• maintenance systems;
• sensors;
• production scheduling;
• robotics;
• cybersecurity;
• cloud infrastructure.

Once integrated, changing providers can require substantial expenditure and operational disruption.

B. Data advantages

Industrial AI improves through access to:

• machine telemetry;
• production histories;
• maintenance records;
• failure data;
• energy consumption;
• quality-control information;
• operational parameters.

A platform with access to a large installed industrial base can therefore develop better models.

This may create a data-feedback loop:

More machines → more data → better AI → better performance → more customers → still more data.

C. Network effects

The platform may become more valuable as more:

• manufacturers;
• equipment suppliers;
• developers;
• system integrators;
• maintenance providers

join the ecosystem.

D. Interoperability dependence

Industrial customers may depend on a platform's APIs and technical standards.

If interoperability is restricted, competing AI providers may be unable to operate effectively.

3. Relevant Competition-Law Theories

Industrial AI platform conduct may potentially implicate several doctrines.

3.1 Abuse of dominance

A dominant industrial AI platform may abuse its position through:

• exclusionary tying;
• refusal to provide interoperability;
• discriminatory access;
• self-preferencing;
• excessive contractual restrictions;
• discriminatory API access;
• loyalty rebates;
• exclusionary licensing.

3.2 Essential-facility-type theories

An industrial AI platform may become particularly important where competitors cannot reasonably replicate:

• unique industrial datasets;
• critical APIs;
• machine interfaces;
• proprietary protocols;
• platform infrastructure.

However, competition law generally does not require dominant firms to share every valuable asset. The legal threshold for compulsory access is normally demanding.

3.3 Leveraging

A platform dominant in industrial operating software might leverage that position into:

• AI analytics;
• predictive maintenance;
• robotics;
• industrial cloud;
• cybersecurity;
• digital twins.

3.4 Tying and bundling

A platform might require customers purchasing industrial-control software also to purchase:

the platform's proprietary AI analytics service.

This can foreclose independent AI suppliers.

3.5 Self-preferencing

Where the platform operates an industrial application marketplace, it might rank its own AI applications above competing applications.

3.6 Data foreclosure

A platform could potentially prevent competitors from accessing industrial data generated by machines connected to the platform.

This becomes especially important where the data are necessary for effective AI competition.

4. Industrial AI Platform as an Ecosystem

Industrial AI competition should not necessarily be analysed as a single software market.

A platform may contain multiple layers:

Industrial equipment

↓

Sensors / IIoT

↓

Industrial operating platform

↓

Cloud / compute infrastructure

↓

Industrial data layer

↓

AI models

↓

Applications

↓

Industrial customers

A firm may possess substantial market power at one layer and attempt to extend that power to another.

This produces an important analytical question:

Is the platform competing on the merits, or is it using control of one indispensable ecosystem layer to disadvantage rivals in another?

5. Important Case Laws

Case 1 — Microsoft Corp. v. Commission

Court: General Court of the European Union
Year: 2007

The Microsoft case is highly relevant to industrial AI platforms because it demonstrates how control over an important technological platform can create competition concerns when competitors require interoperability information.

Microsoft was found to have abused its dominant position by restricting interoperability information concerning work-group server products and by tying Windows Media Player to Windows.

Relevance to industrial AI

An industrial AI platform may similarly control:

• proprietary APIs;
• machine communication protocols;
• interoperability information;
• data-access interfaces.

If competing AI developers cannot effectively interoperate with industrial equipment because the dominant platform withholds necessary technical information, the Microsoft principles become highly relevant.

Competition lesson

Control over interoperability can become a source of exclusionary power.

6. Case 2 — IMS Health v. NDC Health

Court: Court of Justice of the European Union
Year: 2004

The case concerned access to a commercially important database and the circumstances under which refusal to license intellectual property may constitute abuse of dominance.

The Court established demanding conditions for compulsory access, including circumstances where the requested input is indispensable and refusal risks eliminating effective competition.

Relevance to industrial AI

Consider an industrial AI platform possessing a unique dataset containing:

• decades of machine-failure information;
• highly granular industrial telemetry;
• proprietary maintenance histories.

If competitors cannot realistically reproduce the dataset, they may argue that access is indispensable.

But mere usefulness is insufficient.

Competition lesson

Industrial data can be strategically important without automatically becoming an essential facility.

7. Case 3 — Bronner v. Mediaprint

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

Bronner established a strict framework for determining when refusal to provide access to infrastructure constitutes an abuse of dominance.

The Court emphasised the importance of indispensability and the absence of a realistic alternative.

Relevance to industrial AI

Suppose an industrial AI platform controls the only commercially viable interface connecting a particular class of industrial machines to third-party AI applications.

A competitor seeking mandatory API access would potentially have to demonstrate:

1. the interface is indispensable;
2. there is no realistic substitute;
3. replication is not reasonably possible;
4. refusal risks eliminating effective competition;
5. there is no objective justification for the refusal.

Competition lesson

Not every proprietary industrial interface constitutes an essential facility.

8. Case 4 — Google Shopping

Case: Google Search (Shopping)
Court: General Court of the European Union
Year: 2021

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

The case is important for understanding self-preferencing and leveraging of platform infrastructure.

Relevance to industrial AI

Imagine an industrial AI platform operating an application marketplace.

It could potentially:

• place its own predictive-maintenance application first;
• reduce visibility of competing AI applications;
• favour its own digital-twin software;
• impose discriminatory ranking conditions.

The competition concern would not simply be that the platform competes with application providers.

It would be that the platform controls the gateway through which competing applications reach customers while simultaneously competing in that downstream market.

Competition lesson

Platform neutrality can become a competition issue where control of an infrastructure layer is used to disadvantage downstream competitors.

9. Case 5 — Android / Google

Case: Google Android
Court: General Court of the European Union
Year: 2022

The Google Android litigation concerned practices involving Android, Google Search, Chrome, and application distribution.

The case illustrates how contractual and technological restrictions imposed by a powerful platform can influence competition across connected markets.

Relevance to industrial AI

Industrial AI platforms may similarly combine:

• operating systems;
• application stores;
• cloud services;
• AI services;
• industrial hardware.

A platform could potentially make access to one component conditional upon adoption of another.

For example:

Industrial operating software + mandatory proprietary AI service + mandatory cloud hosting.

This could create ecosystem foreclosure.

Competition lesson

Competition authorities may examine the combined effects of contractual restrictions across interconnected platform markets.

10. Case 6 — Intel v. Commission

Court: Court of Justice of the European Union
Year: 2017

Intel concerned loyalty rebates and the assessment of exclusionary effects.

The Court clarified the importance of examining whether rebate practices are capable of foreclosing an equally efficient competitor where that analysis is relevant.

Relevance to industrial AI

An industrial AI platform could provide customers with incentives such as:

• discounts for exclusive use;
• rebates based on percentage of AI workloads;
• preferential pricing for customers using the platform's complete ecosystem;
• discounts conditional on purchasing its cloud and AI products together.

Such arrangements could make it economically difficult for independent AI providers to obtain sufficient scale.

Competition lesson

Commercial discounts can become exclusionary when they materially restrict competitors' access to demand.

11. Case 7 — Qualcomm

Case: Qualcomm
Court: General Court of the European Union
Year: 2022

The Qualcomm litigation concerned exclusionary payments and competition in the semiconductor sector.

The case is particularly useful for industrial AI because AI platforms depend heavily upon hardware components such as:

• GPUs;
• AI accelerators;
• industrial edge processors;
• networking components.

Relevance

An industrial AI platform controlling critical hardware and software could potentially use rebates or payments to discourage equipment manufacturers from adopting rival AI technologies.

Competition lesson

Vertical relationships between technology suppliers and downstream platform operators can create foreclosure concerns even where the conduct is formally contractual.

12. Case 8 — Slovak Telekom

Case: Slovak Telekom v Commission
Court: Court of Justice of the European Union
Year: 2021

The case concerned exclusionary conduct involving access to telecommunications infrastructure.

The judgment is useful for analysing situations in which a dominant undertaking controls infrastructure that competitors need to reach customers.

Industrial AI relevance

An industrial AI platform can function as a technological gateway similar to infrastructure.

Potential examples include:

• industrial APIs;
• machine operating environments;
• proprietary data buses;
• industrial cloud interfaces;
• digital-twin infrastructure.

Competition lesson

Where a dominant platform controls an infrastructure layer, access restrictions can have downstream exclusionary consequences.

13. Industrial AI Data Foreclosure

Data may become one of the most important competition parameters.

Consider two AI providers:

The incumbent may therefore benefit from a data advantage that compounds over time.

However, competition law must distinguish between:

legitimate competitive advantage

and

strategic exclusion of rivals.

The mere possession of superior data is not necessarily unlawful.

14. Industrial AI and Interoperability

Interoperability is particularly important because industrial environments often contain equipment from multiple generations and manufacturers.

A dominant platform could potentially restrict interoperability through:

• closed APIs;
• proprietary protocols;
• certification restrictions;
• technical incompatibility;
• licensing restrictions;
• API rate limits;
• discriminatory authentication;
• artificial latency;
• restricted software-development kits.

The competition question becomes:

Does the technical architecture reflect legitimate security and performance requirements, or has interoperability been restricted primarily to disadvantage competitors?

15. AI Model Lock-In

Industrial AI models may become embedded into production processes.

For example:

Factory → Platform AI → Predictive Maintenance → Automated Ordering → Production Scheduling

Once multiple processes depend upon one AI system, customers may face substantial switching costs.

A platform could strengthen lock-in through:

• proprietary model formats;
• non-portable embeddings;
• incompatible APIs;
• restricted historical-data export;
• contractual termination charges;
• proprietary digital-twin formats.

This can create technological path dependence.

16. Self-Preferencing in Industrial AI Marketplaces

Industrial AI platforms increasingly resemble app ecosystems.

A platform could operate a marketplace containing:

• predictive-maintenance applications;
• quality-control AI;
• robotics applications;
• energy-optimisation systems;
• industrial cybersecurity tools.

If the platform simultaneously sells its own competing applications, it may have an incentive to manipulate:

• ranking;
• recommendation;
• certification;
• search visibility;
• API access;
• pricing;
• technical compatibility.

This raises a classic gatekeeper-versus-competitor conflict.

17. Bundling and Tying

Suppose a dominant industrial platform requires:

Industrial cloud + proprietary AI model + proprietary analytics + proprietary data-storage service.

Customers may technically have a choice, but the commercial architecture may make independent alternatives economically unattractive.

Competition authorities could examine:

1. dominance in the tying product;
2. distinctness of the tied product;
3. coercion or practical conditionality;
4. foreclosure;
5. objective justification;
6. effects on competitors and consumers.

18. Algorithmic Discrimination Between Competitors

Industrial AI platforms can potentially use algorithms to differentiate between:

• their own applications;
• preferred partners;
• independent developers;
• competing equipment manufacturers.

For example, an algorithm might provide the platform's own maintenance software with:

• faster API response;
• higher data quotas;
• privileged telemetry;
• greater compute allocation;
• earlier access to new features.

The discrimination could be difficult to detect because the discriminatory rule may be embedded within automated systems.

This creates a new evidentiary problem:

Competition authorities may need to audit algorithms rather than merely review contracts.

19. Industrial AI and Mergers

Industrial AI markets may also raise merger-control concerns.

A large industrial company acquiring an AI startup may obtain:

• valuable datasets;
• specialised algorithms;
• key engineers;
• industrial customers;
• proprietary interfaces.

The traditional turnover-based approach can sometimes understate the competitive significance of an emerging AI company.

A transaction involving a relatively small AI firm may nevertheless eliminate a future competitive threat.

Relevant theories include:

• killer acquisitions;
• nascent competition;
• data consolidation;
• vertical foreclosure;
• ecosystem expansion;
• innovation competition.

20. Industrial AI and Vertical Integration

Vertical integration can produce both efficiencies and competition risks.

Potential efficiencies

Integration may provide:

• better cybersecurity;
• faster data processing;
• lower latency;
• improved predictive maintenance;
• better interoperability;
• reduced transaction costs.

Potential risks

The same integration may enable:

• input foreclosure;
• customer foreclosure;
• self-preferencing;
• data foreclosure;
• interoperability restrictions;
• discriminatory access.

Therefore, vertical integration should not automatically be treated as anticompetitive.

21. Industrial AI and Killer Acquisitions

An incumbent industrial platform may acquire a startup developing:

• autonomous factory AI;
• advanced robotics;
• predictive-maintenance technology;
• industrial foundation models;
• machine-vision systems.

Even where the startup has little current revenue, it may represent a future competitive constraint.

The competition analysis should therefore examine:

innovation potential, technological trajectory, customer adoption, intellectual property, engineering capability and data assets—not merely current turnover.

22. Industrial AI and Collective Dominance

A further possibility arises where several large industrial AI platforms become mutually dependent.

For example:

• Platform A controls industrial cloud;
• Platform B controls industrial equipment;
• Platform C controls AI accelerators.

They may remain nominally independent but form a highly concentrated ecosystem.

Competition analysis may therefore need to examine:

• common ownership;
• contractual dependencies;
• interoperability arrangements;
• data-sharing arrangements;
• common standards;
• algorithmic coordination.

23. Algorithmic Coordination

Industrial AI systems may independently optimise:

• prices;
• production;
• inventory;
• capacity;
• procurement.

If competing firms deploy similar optimisation systems, algorithms could potentially produce coordinated market outcomes without traditional human communication.

The legal challenge is distinguishing:

parallel optimisation

from

anticompetitive coordination.

Evidence may include:

• common software providers;
• shared algorithms;
• common data inputs;
• coordinated pricing outputs;
• communications between firms;
• deliberate algorithmic design.

24. Remedies

Competition authorities could employ several remedies.

Structural remedies

• divestiture;
• separation of platform and downstream AI operations;
• prohibition of certain acquisitions.

Behavioural remedies

• interoperability obligations;
• API access;
• data portability;
• non-discrimination rules;
• ranking transparency;
• access commitments.

Technical remedies

• open standards;
• API documentation;
• model portability;
• machine-readable data export;
• interoperability testing.

Monitoring remedies

• independent algorithm audits;
• compliance officers;
• periodic reporting;
• technical monitoring;
• access logs.

25. Key Legal Principles Emerging From the Case Law

26. Overall Competition-Law Test

A useful analytical framework for industrial AI platforms is:

Step 1 — Define the relevant market

Possible markets include:

• industrial AI software;
• predictive maintenance;
• industrial cloud;
• digital twins;
• industrial operating systems;
• AI-enabled robotics;
• industrial data services.

Step 2 — Determine market power

Consider:

• market share;
• switching costs;
• network effects;
• data advantages;
• interoperability;
• ecosystem control;
• entry barriers.

Step 3 — Identify the bottleneck

Ask whether the platform controls:

data + infrastructure + interfaces + customers + AI models.

Step 4 — Identify the conduct

Examples:

• tying;
• bundling;
• self-preferencing;
• refusal to interoperate;
• discriminatory access;
• loyalty rebates;
• data foreclosure;
• exclusionary acquisition.

Step 5 — Analyse foreclosure

Determine whether rivals are actually or potentially prevented from:

• entering;
• scaling;
• innovating;
• accessing customers;
• obtaining necessary data.

Step 6 — Consider efficiencies

Possible justifications include:

• cybersecurity;
• reliability;
• safety;
• privacy;
• latency;
• interoperability;
• quality control.

Step 7 — Select proportionate remedies

The remedy should preserve legitimate technological integration while preventing exclusionary leverage.

27. Conclusion

Industrial AI platforms create a new form of competition-law risk because they can combine control over industrial data, software, infrastructure, AI models and customer access within one ecosystem.

The principal concern is not simply that an industrial AI company becomes large. The deeper concern is ecosystem leverage:

control of industrial infrastructure → control of data → superior AI → customer lock-in → stronger ecosystem → reduced competitive access.

The most important case-law foundations are Microsoft, IMS Health, Bronner, Google Shopping, Google Android, Intel, Qualcomm and Slovak Telekom. Together, they provide principles for analysing interoperability, essential inputs, tying, self-preferencing, loyalty incentives, vertical foreclosure and platform access.

For future industrial AI competition law, the decisive issues are likely to be data portability, API interoperability, algorithmic discrimination, AI-model lock-in, industrial-data foreclosure, ecosystem acquisitions, and the ability of independent AI developers to compete on an equal technological footing.