Wiki · Concept · Last reviewed August 12, 2026

Platform Monopoly Power

Platform monopoly power is significant and durable single-firm power in a defined market, expressed through control of a digital gateway, ecosystem, or infrastructure layer. Size, concentration, dependency, gatekeeper designation, unlawful monopolization, and remedy effectiveness are related but different claims; each needs its own market, conduct, procedural status, and evidence.

Category: Concept Published: June 25, 2026 Modified: August 12, 2026 Last reviewed: August 12, 2026 Tags: competition, gatekeepers, antitrust, platforms, AI infrastructure, cloud, app stores

Definition

Platform monopoly power is significant and durable single-firm market power exercised through a platform's intermediary or infrastructure role. In U.S. Section 2 guidance, monopoly power is shorthand for the long-term ability to raise price or exclude competitors in a relevant market. Possessing that power is not itself unlawful: a monopolization case also asks whether the firm acquired or maintained it through exclusionary conduct rather than competition on the merits. Other jurisdictions use different legal tests and regulatory tools.

Digital platforms connect groups that need one another: users and app developers, publishers and advertisers, merchants and buyers, cloud customers and software vendors, or models and downstream applications. Their power may therefore appear in non-price dimensions as well as fees: degraded quality or privacy, restrictive business terms, suppressed interoperability, discriminatory ranking, reduced innovation, or blocked access to a low-cost route to users. Analysis must still name the affected market or platform side, realistic substitutes, duration, and evidence; "the platform is powerful" is not a conclusion.

A market dominated by a few firms is concentrated or oligopolistic, not proof that any one firm has monopoly power. Likewise, platform governance, vendor dependence, EU gatekeeper status, UK Strategic Market Status, and an adjudicated competition-law violation are distinct categories. An institution can face serious operational lock-in without proving a legal monopoly, and a regulator can designate a gatekeeper without finding wrongdoing.

For AI, the relevant control point may be above or below the model: specialized compute, cloud commitments, training data, model access, enterprise identity, developer tooling, app distribution, assistant defaults, or procurement and compliance records. Vertical reach across several layers can strengthen leverage, but stack breadth alone does not establish monopoly power or unlawful conduct.

Boundary Tests

  1. Define the claim. Identify the product or service, geography, customer group, platform side, and time period. Separate a legal monopoly-power claim from concentration, gatekeeper status, bargaining imbalance, or a buyer's vendor-risk assessment.
  2. Test substitutes and durability. Ask where users or businesses would realistically turn after worse prices, quality, privacy, access, or terms. Examine shares over time, entry and expansion barriers, network effects, defaults, switching and multi-homing costs, scale economies, and direct evidence that customers or rivals cannot discipline the platform.
  3. Name the conduct. Exclusive defaults, tying or bundling, self-preferencing, discriminatory API or marketplace access, restrictions on steering, misuse of business-user data, interoperability limits, and strategic partnerships are different theories. Do not infer exclusion from size alone.
  4. Trace effects and the counterfactual. Specify who was harmed, through which control point, compared with what plausible competitive path. Measure price, output, quality, privacy, innovation, entry, visibility, creator or developer terms, and resilience where relevant.
  5. Evaluate justification and fit. Security, privacy, safety, fraud prevention, and product integrity can justify constraints. The evidence question is whether the rule addresses a defined risk, is applied consistently, and has a less restrictive alternative—not whether the platform invokes safety language.

Operational dependency deserves its own test: can the organization export data and records, move identity and permissions, substitute a model or provider, preserve service, and verify deletion at a known cost and time? A failed exit test is a governance risk even when no competition authority has defined a market or found liability.

Mechanisms

Network effects. A service becomes more valuable as more people, developers, businesses, or datasets join it. That can improve products, but it can also make rivals look empty or risky before they have a fair chance to grow.

Defaults and preinstallation. Search engines, browsers, app stores, assistants, wallets, identity providers, and productivity suites gain power when users meet them as the default path. The U.S. Google search case turned heavily on exclusionary agreements and default distribution, not simply on the quality of search.

Vertical integration. A platform may operate the marketplace and compete inside it. That can create incentives for self-preferencing, discriminatory access, preferential data use, or ranking rules that favor the platform's own products.

Data feedback loops. Usage data can improve ranking, targeting, recommendation, fraud detection, model quality, and product design. When a platform also controls the surface where behavior occurs, rivals may lack both users and the data needed to improve.

Switching and multi-homing costs. Contract terms, committed-spend discounts, data egress fees, proprietary APIs, app review processes, identity systems, enterprise integrations, cloud architecture, and user habits can make exit or simultaneous use of rivals expensive even when alternatives exist.

Bundling and ecosystem gravity. A platform can make an adjacent product appear cheaper, safer, or easier to adopt by bundling it with office software, search, mobile operating systems, cloud contracts, developer credits, identity, security tooling, or compliance workflows.

Strategic partnerships and partial integration. Equity stakes, cloud-spend commitments, exclusive access, revenue sharing, technical dependencies, governance rights, and access to sensitive information can affect entry and bargaining without a formal acquisition. Their competitive effect must be shown rather than inferred from the partnership label.

Control of safety or trust gates. Security reviews, abuse controls, app-store rules, model access policies, content moderation, age checks, fraud detection, and privacy justifications can be legitimate. They can also become opaque gatekeeping unless the rules are proportionate, appealable, and applied without self-preference.

Current Context

As of this August 12, 2026 review, court judgments, appeals, ex ante gatekeeper rules, Strategic Market Status designations, market investigations, conduct requirements, and commitments all touch platform power. The examples below are illustrative, not a global case inventory, and their procedural labels are part of the claim.

United States — search. The District Court for the District of Columbia found in August 2024 that Google unlawfully maintained monopolies in general search services and general search text advertising. Its December 5, 2025 final judgment restricts specified exclusive distribution arrangements involving Search, Chrome, Google Assistant, and Gemini, and provides for defined data-sharing and search-syndication remedies. By the review date, DOJ's case page recorded Technical Committee oversight, compliance and joint status reports through July 30, 2026, and a July 28 DOJ response brief and opening brief on cross-appeal. Compliance monitoring was active, while the appellate outcome and practical competitive effects remained unsettled.

DOJ's September 2025 remedy summary said the search remedies reach generative-AI technologies and companies so that the same distribution tactics are not transferred to GenAI products. That is the Department's characterization of the ruling; the final judgment and memorandum opinion are the operative sources. In the separate Eastern District of Virginia ad-tech case, the court found Google liable in April 2025 for monopolization in open-web display publisher-ad-server and ad-exchange markets and for unlawful tying. As of August 12, 2026, DOJ's case page listed remedy proposals and post-trial briefing but no final judgment.

European Union. The Digital Markets Act complements rather than replaces EU competition law and imposes ex ante duties on designated core platform services. The Commission's portal listed seven gatekeepers and 23 designated services. On July 23, 2026, the Commission adopted two Google non-compliance decisions and fines totaling €890 million: one concerning self-preferencing on Google Search and one concerning restrictions on steering from Google Play. These were DMA enforcement decisions, not findings that Google monopolized every service it operates.

The Commission's June 25 preliminary view was that AWS and Microsoft Azure should be designated for cloud computing even though the services did not meet the DMA's quantitative thresholds. It cited gateway importance, entrenched users, lock-in, switching costs, ecosystems, and AI tools and partnerships in cloud procurement. The firms retained an opportunity to respond, and no final cloud designation had been announced by the review date.

United Kingdom — search and mobile. The CMA designated Google with SMS in general search and search advertising and imposed three requirements in June 2026: publisher controls and information for search generative-AI features, fair ranking, and data portability. Its July 29 consultation outcome said a separate user-choice decision would follow. Apple and Google had also been designated with SMS for their mobile platforms in October 2025. The CMA published final app-review, ranking, data-use, and Apple interoperability commitments on April 1, 2026, then consulted on proposed mobile steering requirements from June 30 through July 28; those proposed requirements were not final at the review date.

United Kingdom — cloud and business software. The CMA's 2025 cloud investigation found Amazon and Microsoft held significant market power and identified egress, interoperability, and Microsoft software-licensing concerns. In March 2026 the CMA Board chose continued engagement on announced AWS and Microsoft cloud changes, with a six-month review, and opened a separate SMS investigation into Microsoft's business-software ecosystem on May 14. That investigation covered bundling, interoperability, defaults, cloud licensing, and AI integration; opening it did not designate Microsoft or find wrongdoing.

AI partnerships. The FTC's January 2024 compulsory-information orders and January 2025 staff report examined relationships among Microsoft and OpenAI, Amazon and Anthropic, and Google and Anthropic. The report discussed cloud-spend commitments, switching costs, governance or consultation rights, access to sensitive information, and the competitive role of compute and cloud infrastructure. A Section 6(b) study gathers and analyzes information; it is not an adjudication that a named partnership violated antitrust law.

AI Relevance

AI can deepen platform power because models can be both products and infrastructure, but neither model capability nor investment size proves monopoly power. The same system can be a consumer assistant, developer API, office copilot, search surface, agent runtime, app marketplace, security filter, and procurement dependency. Once organizations build workflows around a model platform, switching may require new prompts, evaluations, security reviews, data connectors, logs, contracts, staff training, and compliance evidence.

The most important AI control points are not only model quality. They include compute supply, cloud discounts, chip availability, identity and permission systems, enterprise distribution, app stores, browser and operating-system defaults, training data licenses, retrieval indexes, developer tooling, safety policies, and the ability to certify or block agents. A platform that owns several of these layers can steer the market without openly banning competitors.

Agentic systems intensify the default problem. If an assistant, browser agent, enterprise copilot, or checkout agent chooses the search source, merchant, app, model, connector, or payment rail before the user sees alternatives, platform power moves from ranking pages to preselecting actions. A fair interface should make sponsored placement, unavailable alternatives, tool restrictions, and platform self-preference visible enough to contest.

This has safety implications. Concentrated AI platforms can impose useful safeguards, coordinate incident response, and keep dangerous tools away from high-risk users. They can also create single points of failure, suppress independent research, hide security problems, narrow model choice, or turn safety language into a justification for closed markets. Governance has to distinguish necessary risk controls from exclusionary gatekeeping.

AI also changes the meaning of platform dependency. A model endpoint is often embedded inside cloud identity, data warehouses, vector databases, monitoring, safety filters, app stores, procurement contracts, and agent tooling. A buyer may be able to change a model name in code while still being locked into the surrounding platform record, permissions, logs, workflows, and compliance evidence.

Open-weight models, interoperable standards, portable data, public procurement requirements, independent evaluations, cloud exit rights, and public option digital services can reduce particular dependencies. None automatically creates a competitive market: open weights may still rely on concentrated compute or distribution, portability may omit usable context, and procurement frameworks may reproduce incumbent qualification advantages.

Governance and Safety

The governance task is to preserve contestability without pretending that all interoperability, openness, or low-friction access is automatically safe. Some platform controls are necessary for security, privacy, child safety, fraud prevention, and misuse prevention. The test is whether the control is evidence-based, proportionate, transparent enough to contest, and separable from the platform's commercial preference for its own products.

Useful remedies and oversight tools include data portability, interoperability duties, anti-steering protections, non-discrimination rules, app-review transparency, appeal rights, cloud egress limits, procurement exit plans, merger reporting, audit access, source transparency, independent red-team access, and restrictions on using business-user data to compete against those users.

For public institutions, platform dependency should be treated as an operational risk. A city, school, court, hospital, newsroom, or agency that builds around one cloud, one app store, one search surface, one identity provider, or one AI model should know how it would switch providers, preserve records, keep services running, and audit decisions after a dispute, outage, policy change, price shock, model retirement, or acquisition.

Competition policy and safety policy can collide. A narrow safety gate may protect people from a real harm. A broad safety gate may protect an incumbent from competition. Good governance requires records that show the threat model, rule, affected parties, appeal path, and evidence for why a less restrictive measure would not work.

A practical governance record should connect competition remedies to AI system inventories, vendor registers, procurement files, AI bills of materials, audit trails, retention rules, and incident reporting. Without those records, an organization may know that a platform is legally regulated while still being unable to prove which model, cloud region, ranking rule, payment rail, or safety gate governed a specific case.

Every remedy should state its theory of harm, covered service and conduct, beneficiaries, security or privacy constraints, implementation deadline, anti-circumvention rule, evidence-access path, and review or sunset trigger. Measure both compliance and outcomes. A choice screen can be displayed yet ineffective; an export can exist yet be unusable; an interoperability interface can be nominally open yet slower or less capable than the platform's private path.

Remedy and Evidence Records

Platform remedies are only governable if they produce evidence. A search default remedy should leave records about default placement, user choice screens, distribution agreements, syndication access, query data handling, and whether AI search or assistant surfaces receive equivalent treatment. An app-store remedy should leave records about review timelines, rejection reasons, ranking changes, developer-data use, payment steering, appeal outcomes, and interoperability requests.

A cloud or AI infrastructure remedy should document egress terms, committed spend, discount conditions, technical portability, identity dependencies, model endpoint substitution, data residency, audit logs, and transition support. This is why cloud competition is now linked to AI: the model may be portable in code while the surrounding data, permissions, logs, credits, and compliance evidence remain stuck.

Public buyers should maintain an exit record before adopting a strategic platform. The record should name the alternative provider path, export format, migration cost, retained logs, open standards used, contract termination rights, data deletion rules, and which human service would keep operating if an automated assistant, search surface, or identity provider became unavailable.

Evidence should preserve the counterfactual and baseline: what terms, ranking, latency, fees, rejection rates, switching rates, or interoperability capabilities existed before the remedy, and what changed afterward. A platform's compliance submission is evidence of what it says it implemented, not independent proof of contestability. Regulators, auditors, researchers, and affected businesses need reproducible tests, versioned interfaces, complaint outcomes, and enough protected data access to detect discrimination or circumvention without exposing personal data or security-sensitive details.

Source Discipline

Claims about platform monopoly power should name the evidence type. A court judgment is different from a complaint, proposed remedy, compliance report, regulator speech, company announcement, academic article, or press summary. A regulator's designation of gatekeeper or Strategic Market Status is not the same as a finding that the firm committed an antitrust violation.

Source discipline also means naming the market and the control point. "Big Tech monopoly" is too broad. Search defaults, publisher ad servers, app-store review, mobile operating systems, cloud egress fees, model APIs, data licenses, and assistant defaults are different markets with different evidence and remedies.

For current cases, dates matter. Litigation, appeals, compliance monitors, commitments, preliminary findings, final designations, conduct requirements, and implementation reports can change quickly. This page treats official court records, regulator case pages, statutory texts, and agency reports as stronger evidence than commentary. Use media coverage for reception and context, not for the legal status of a proceeding.

For AI-platform claims, separate four things that are often collapsed: market power, legal liability, safety control, and operational dependency. A cloud partnership can be lawful but lock-in producing; a safety rule can be legitimate but discriminatory in operation; an open model can reduce model dependence while leaving inference, app distribution, or procurement dependence intact.

For remedy claims, cite the operative document, not only the press release. A final judgment, conduct requirement, commitment decision, preliminary view, market-investigation report, staff report, and compliance filing each supports a different claim. If a source says "preliminary," "recommended," "proposed," or "under investigation," preserve that procedural label in the article.

Use verbs that preserve posture: a complaint alleges, an agency opening investigates, a preliminary view proposes or preliminarily finds, a court or regulator decision finds, and an appeal challenges. A press release can accurately report its issuing authority's position, but it should not silently substitute for the judgment, decision, notice, or report it summarizes. Current case pages should be checked for stays, appeals, later orders, compliance filings, and superseding remedies.

Spiralist Reading

For Spiralism, platform monopoly power becomes a civic problem when private defaults decide what can be seen, sold, remembered, searched, automated, or trusted. The question is not only price. It is reality power: who owns the interface through which knowledge, identity, agency, and memory become actionable.

A platform monopoly is not just a company with many users. It is a private chokepoint that can make some paths feel natural and others impractical. When assistants mediate work, search, commerce, or public services, the control point shifts from ranking visible options toward selecting tools and routes before a person sees the alternatives.

The Spiralist answer is not reflexive break-up rhetoric or reflexive trust in incumbent safety claims. It is source discipline, exit rights, interoperability, public-interest audits, visible defaults, appealable gates, and institutions that can tell the difference between stewardship and enclosure.

Open Questions

Competition and policy

AI stack

Platform systems

Sources

Return to Wiki