1. Introduction

Interconnected Energy AI Systems are networks in which artificial intelligence is used across electricity generation, transmission, distribution, storage, trading, demand response, forecasting, grid balancing, energy markets and consumer management.

Examples include AI systems that:

• forecast electricity demand;
• predict renewable generation;
• control battery storage;
• optimize power dispatch;
• manage electric-vehicle charging;
• detect grid faults;
• operate distributed energy resources;
• optimize transmission;
• trade electricity automatically;
• coordinate microgrids;
• manage demand-response programs.

The principal legal and competition concern is systemic dependency.

A power system can become dependent upon a small number of AI platforms, data providers, cloud providers, model providers or algorithmic coordinators. If one of those systems fails, behaves incorrectly, is manipulated, or becomes unavailable, the consequences may spread across the electricity ecosystem.

The central question is therefore:

When interconnected AI systems become critical infrastructure for energy markets, how should law manage the resulting dependency, concentration, interoperability and systemic-failure risks?

2. From Traditional Grid Dependency to AI Dependency

Traditional electricity systems already contain dependencies.

For example:

Generation → Transmission → Distribution → Consumer

AI adds another layer:

Data → AI model → automated decision → physical energy infrastructure

This creates a new type of dependency.

A grid operator may depend upon an AI forecasting platform to determine:

• expected demand;
• renewable output;
• reserve requirements;
• congestion;
• battery dispatch.

If many operators use the same model or provider, a common technological failure can affect multiple markets simultaneously.

3. What Makes AI Dependency Systemic?

Dependency becomes systemic when failure at one technological layer can propagate through multiple independent entities.

Consider:

Cloud provider

↓

AI model provider

↓

Grid-management platforms

↓

Transmission operators

↓

Distribution networks

↓

Industrial consumers

A failure at the cloud or AI layer can therefore potentially affect physical electricity infrastructure.

This is different from ordinary commercial dependency because the consequences may extend beyond individual contracting parties.

4. Major Sources of Systemic Dependency

A. Common AI Providers

Multiple utilities may rely on the same AI vendor.

This creates common-provider risk.

If the vendor's:

• model fails;
• service becomes unavailable;
• API changes;
• cybersecurity is compromised;
• pricing becomes prohibitive;

multiple utilities may simultaneously experience disruption.

B. Cloud Concentration

Energy AI systems frequently depend on cloud infrastructure.

The dependency chain can therefore become:

Cloud → data infrastructure → AI model → energy-management software → grid.

Cloud concentration can create a systemic single point of failure.

C. Data Dependency

AI energy systems depend upon:

• weather data;
• smart-meter data;
• grid telemetry;
• market data;
• consumption data;
• generation data;
• satellite information.

If a dominant data provider controls a critical dataset, competing AI systems may become dependent upon it.

5. Model Dependency

Energy companies may increasingly rely upon foundation models or specialized AI models.

Once an organization has:

• trained workflows;
• integrated APIs;
• automated decision systems;
• employee expertise;
• historical data pipelines;

switching to another AI provider may become difficult.

This creates AI vendor lock-in.

6. Feedback Loops

Interconnected AI systems can also create feedback loops.

For example:

AI forecasts demand

↓

AI controls energy prices

↓

Consumers change consumption

↓

New consumption data enters the model

↓

Model updates forecasts

↓

AI changes prices again

The system therefore becomes partially self-referential.

If many firms use similar models, their decisions may become correlated.

7. Algorithmic Correlation

Suppose ten electricity traders use substantially similar AI systems.

Each system:

• observes the same market;
• receives similar data;
• optimizes similar objectives;
• reacts rapidly to competitors.

Their algorithms could produce highly correlated decisions even without explicit communication.

This raises competition concerns involving:

• algorithmic coordination;
• tacit collusion;
• synchronized bidding;
• market manipulation;
• artificial scarcity.

8. Energy Markets Are Especially Sensitive

Electricity has unusual characteristics.

It is:

• difficult to store at massive scale;
• continuously balanced;
• highly time-sensitive;
• infrastructure dependent;
• essential for society;
• geographically constrained.

Therefore, an AI failure that would be merely inconvenient in another market can become a systemic infrastructure event in electricity.

9. AI-Controlled Distributed Energy Resources

The modern energy system contains:

• rooftop solar;
• batteries;
• electric vehicles;
• heat pumps;
• smart appliances;
• microgrids.

AI can coordinate millions of these assets.

This creates an enormous potential benefit.

But it also creates a concentration problem:

Who controls the algorithm coordinating millions of decentralized energy assets?

If one platform controls those resources, it may acquire significant influence over energy markets.

10. Platform Power

An energy-AI platform could sit between:

Consumers → Distributed Energy Resources → Grid → Energy Markets

The platform could control:

• dispatch;
• pricing;
• participation;
• access;
• data;
• optimization.

This produces a potential energy platform monopoly.

11. Systemic Dependency vs Dominance

These concepts should be distinguished.

Dominance

A company possesses substantial market power.

Systemic dependency

Critical market participants cannot reasonably function without the company's technology.

A company could have relatively modest market share but still create systemic dependency if it supplies an indispensable technological layer.

12. Essential Infrastructure

The legal problem resembles the traditional essential-facilities debate.

Suppose a company controls an AI platform necessary for:

• real-time grid balancing;
• renewable forecasting;
• automated transmission management.

If no practical alternative exists, refusal to provide access could potentially raise competition concerns.

But the legal threshold should remain high.

Not every commercially valuable AI system is an essential facility.

13. Important Case Law

14. Case 1 — United States v Terminal Railroad Association

224 U.S. 383 (1912)

The U.S. Supreme Court addressed control over critical railroad infrastructure.

Relevance

The case is a foundational precedent for the essential-facilities concept.

The analogy to energy AI arises where a single platform controls infrastructure that competitors cannot reasonably duplicate.

If an AI platform becomes indispensable for critical energy coordination, questions of access and foreclosure may arise.

15. Case 2 — Aspen Skiing Co. v Aspen Highlands Skiing Corp.

472 U.S. 585 (1985)

The Supreme Court considered a refusal-to-deal situation involving a dominant enterprise.

Relevance

The case demonstrates that under particular circumstances, a dominant firm's decision to terminate an established cooperative relationship can raise antitrust concerns.

Applied to energy AI:

A dominant platform that suddenly withdraws previously available critical interoperability or access could potentially attract scrutiny.

16. Case 3 — Verizon Communications v Trinko

540 U.S. 398 (2004)

The Supreme Court substantially limited the circumstances in which antitrust law imposes duties to deal.

Relevance

This is an essential counterbalance.

An energy-AI company should not automatically be required to provide its technology to every competitor merely because the technology is important.

The case emphasizes the importance of preserving incentives to innovate.

Thus:

Systemic importance does not automatically create an antitrust duty to share.

17. Case 4 — Microsoft Corp. v Commission

Case T-201/04

The EU General Court upheld important findings concerning Microsoft's refusal to provide interoperability information.

Relevance

Interoperability is critical for energy AI.

A dominant platform could potentially use:

• proprietary APIs;
• incompatible data formats;
• technical restrictions;
• interoperability limitations

to prevent competing energy-management systems from functioning effectively.

Microsoft provides a powerful framework for analysing such conduct.

18. Case 5 — Commercial Solvents v Commission

Joined Cases 6/73 and 7/73

The European Court addressed refusal to supply an important input to downstream competitors.

Relevance

Consider:

AI infrastructure provider

↓

Energy-management software

↓

Utilities

If a vertically integrated provider supplies a critical input while also competing downstream, it may have an incentive to restrict rivals' access.

Commercial Solvents therefore provides an important precedent for vertical foreclosure.

19. Case 6 — Slovak Telekom v Commission

Joined Cases C-165/19 P and C-164/19 P

The case concerned exclusionary conduct involving access to telecommunications infrastructure.

Relevance

Telecommunications and energy are increasingly converging technologically.

Energy AI depends upon:

• communications networks;
• cloud services;
• APIs;
• telemetry;
• digital infrastructure.

Slovak Telekom demonstrates how infrastructure access can become an important competition-law issue.

20. Case 7 — Google Android

Case T-604/18

The EU General Court examined Google's conduct within the Android ecosystem.

Relevance

Energy AI platforms may similarly create ecosystems involving:

• operating systems;
• applications;
• data;
• cloud infrastructure;
• energy-management services.

Contractual or technical restrictions at one layer can reinforce dominance at another.

21. Case 8 — Google Shopping

Case T-612/17

The case concerned Google's treatment of its comparison-shopping service.

Relevance

The principle of self-preferencing can be applied by analogy to energy platforms.

Suppose an energy-AI platform operates:

• an energy-management marketplace; and
• its own energy-trading service.

It could potentially favour its own service through:

• ranking;
• data access;
• API functionality;
• recommendations.

22. Case 9 — European Commission v Gazprom

The EU's competition proceedings concerning Gazprom examined restrictions affecting European gas markets.

Relevance

Although gas differs from electricity, the case illustrates the competition-law importance of energy infrastructure and contractual dependence.

Energy-AI platforms could create a new digital layer of dependency analogous to traditional infrastructure dependence.

23. Case 10 — FERC v Electric Power Supply Association

577 U.S. 260 (2016)

The U.S. Supreme Court considered FERC's authority concerning demand-response participation in wholesale electricity markets.

Relevance

This case is particularly important for AI-controlled demand response.

AI can aggregate:

• electric vehicles;
• batteries;
• industrial loads;
• smart appliances.

The legal question becomes how algorithmically coordinated demand resources participate in electricity markets.

24. Competition-Law Risks

The most important competition risks include:

1. AI vendor foreclosure

A dominant provider prevents competitors from accessing essential AI infrastructure.

2. Data foreclosure

A dominant company restricts access to energy data.

3. Platform self-preferencing

The platform favours its own energy products.

4. Exclusive contracts

Utilities are locked into long-term AI arrangements.

5. Interoperability restrictions

Competitors cannot connect to the dominant system.

6. Algorithmic coordination

Similar AI systems produce coordinated market behaviour.

7. Bundling

AI services are tied to electricity or cloud services.

8. Predatory pricing

A large platform subsidizes AI services to eliminate smaller competitors.

25. Systemic Cybersecurity Risk

Energy AI systems also create cybersecurity concerns.

An attacker compromising:

AI platform → multiple utilities

could potentially cause simultaneous disruption.

This creates common-mode failure.

The competition issue becomes intertwined with:

• cybersecurity;
• resilience;
• critical infrastructure regulation;
• national security.

26. Model Manipulation

AI systems can also be attacked through their inputs.

For example, malicious actors might manipulate:

• weather information;
• smart-meter data;
• demand forecasts;
• grid telemetry.

The AI could then make incorrect decisions.

If numerous operators rely on the same model, the impact may become systemic.

27. AI Model Homogeneity

An underappreciated risk is model monoculture.

Suppose 70% of energy-management systems rely on essentially the same model architecture.

A previously unknown model error could then propagate throughout the sector.

This resembles biological monoculture:

Uniformity increases efficiency but reduces resilience to common failure.

Competition policy can therefore have a resilience dimension.

28. Interoperability as a Resilience Tool

Interoperability reduces dependency.

If an energy company can rapidly move from:

AI Provider A → AI Provider B

then the systemic importance of A is lower.

Therefore:

Interoperability is not merely a competition remedy; it can also be a resilience mechanism.

29. Data Portability

Data portability can similarly reduce dependency.

Energy operators should ideally be able to transfer:

• historical demand data;
• model inputs;
• operational records;
• asset information;
• forecasting histories.

Without portability, changing providers can become prohibitively expensive.

30. Switching Costs

Switching from one energy AI provider to another may require:

• new APIs;
• retraining;
• new staff;
• data conversion;
• validation;
• regulatory approval;
• operational testing.

These costs can entrench incumbents.

A dominant provider may therefore gain market power without formally excluding competitors.

31. Cloud Dependency

The dependency chain can become:

Energy company

→ AI software

→ AI model

→ cloud platform

→ data centre

→ semiconductor infrastructure.

This means a disruption at any one layer could potentially propagate downward.

Energy regulation must therefore increasingly consider digital supply-chain concentration.

32. Semiconductor Dependency

Advanced AI systems depend upon specialized processors.

If only a few firms supply critical AI chips, energy AI systems may indirectly depend upon those suppliers.

Thus, systemic dependency can exist at several layers:

chips → cloud → model → software → grid.

33. Energy-AI Mergers

Merger control becomes particularly important where transactions combine:

• AI providers;
• utilities;
• energy-data companies;
• cloud platforms;
• grid-management companies;
• battery aggregators.

Authorities should examine not only market shares but also control over strategic infrastructure.

34. Killer Acquisitions in Energy AI

A utility or technology incumbent might acquire a small AI company developing:

• superior renewable forecasting;
• advanced battery optimization;
• grid balancing technology;
• decentralized energy coordination.

The target may have minimal revenue but substantial future competitive significance.

Authorities should therefore assess potential competition and innovation.

35. Algorithmic Collusion

Energy markets are especially vulnerable because prices can change rapidly.

If multiple AI trading systems:

• observe competitors;
• react simultaneously;
• optimize profit;
• update continuously;

they could potentially learn strategies that reduce competitive pressure.

This does not necessarily establish illegal collusion.

However, competition authorities should examine:

• common software providers;
• common data feeds;
• communication mechanisms;
• algorithmic instructions;
• coordinated outcomes.

36. Regulatory Coordination

Competition law cannot address every systemic risk alone.

Effective governance may require coordination between:

• competition authorities;
• energy regulators;
• cybersecurity authorities;
• data-protection authorities;
• financial regulators;
• critical-infrastructure agencies.

This is particularly important because an AI system can simultaneously be:

a commercial service + critical infrastructure + data processor + market intermediary.

37. Proposed Systemic Dependency Test

A regulator could examine seven factors.

Factor 1 — Criticality

How essential is the AI system to energy operations?

Factor 2 — Concentration

How many alternative providers exist?

Factor 3 — Substitutability

Can the service be replaced?

Factor 4 — Interoperability

Can customers switch systems easily?

Factor 5 — Common exposure

How many market participants rely on the same provider?

Factor 6 — Failure propagation

Could one failure affect multiple markets?

Factor 7 — Competitive foreclosure

Can the provider use dependency to exclude competitors?

38. Potential Remedies

A. Interoperability requirements

Require standardized APIs and data formats.

B. Data portability

Allow energy companies to transfer operational data.

C. Multi-provider requirements

Critical operators could maintain alternative suppliers.

D. Operational redundancy

Critical AI systems should have fallback mechanisms.

E. Non-discrimination

Dominant providers should provide comparable access where legally required.

F. Algorithmic auditing

Critical models should undergo appropriate independent testing.

G. Transparency

Energy operators should know the material operational limitations of AI systems.

H. Merger remedies

Authorities may impose:

• divestiture;
• licensing;
• interoperability;
• data-access;
• non-exclusivity obligations.

39. The Tension Between Competition and Security

An important difficulty is that competition and resilience do not always point in exactly the same direction.

Centralizing an AI system may produce:

• economies of scale;
• better cybersecurity;
• better models;
• lower costs.

But centralization also produces:

• concentration;
• single points of failure;
• dependency;
• reduced contestability.

Therefore, regulators must determine whether the efficiency gained through centralization justifies the systemic dependency created.

40. Energy AI as Critical Digital Infrastructure

The future electricity system may increasingly depend upon:

digital infrastructure as much as physical infrastructure.

Historically, energy resilience focused on:

• power plants;
• substations;
• transmission lines;
• fuel supplies.

Increasingly, resilience must also consider:

• AI models;
• cloud services;
• data centres;
• software platforms;
• algorithms;
• APIs;
• communications networks.

This creates the concept of digital energy infrastructure.

41. Competition and Systemic Risk Are Interconnected

Market concentration can increase systemic risk.

If one provider supplies AI services to most of an electricity market:

Market concentration ↑

→ common dependency ↑

→ substitutability ↓

→ failure impact ↑

Therefore, competition policy can indirectly promote resilience by preventing excessive technological concentration.

42. But Fragmentation Can Also Create Risks

Complete decentralization is not automatically optimal.

Too many disconnected AI systems can create:

• incompatible standards;
• coordination failures;
• cybersecurity weaknesses;
• inconsistent data;
• inefficient grid management.

The goal should therefore not be:

"Maximum fragmentation."

Instead, it should be:

Contestable interconnection with resilient interoperability.

43. Key Legal Principles from the Case Law

The cases collectively provide several principles.

44. Conclusion

Interconnected Energy AI Systems create a new category of systemic dependency in which market power and infrastructure risk can become inseparable.

The problem is not simply that utilities use AI. The deeper concern arises when:

many critical energy actors depend upon the same AI model, cloud provider, data infrastructure, platform or algorithmic coordinator.

At that point, a private technological service may acquire characteristics of systemically important digital infrastructure.

The most important competition-law questions concern:

• essential facilities;
• refusal to deal;
• interoperability;
• vertical foreclosure;
• platform dominance;
• self-preferencing;
• exclusive contracts;
• data concentration;
• algorithmic coordination;
• merger control.

The cases of Terminal Railroad, Aspen Skiing, Trinko, Microsoft, Commercial Solvents, Slovak Telekom, Google Android, Google Shopping and FERC v EPSA demonstrate that existing competition law already contains important principles for addressing these risks.

However, future regulation will require a broader framework combining competition law, energy regulation, cybersecurity, data governance and critical-infrastructure resilience.