Brussels has finally given Article 102 an instruction manual. Issued Sept. 3, the European Commission’s first comprehensive Guidelines on exclusionary abuses of dominance under the Treaty on the Functioning of the European Union (TFEU) replace a legal scavenger hunt with a single framework. The catch is that the manual still gives the Commission considerable room to decide when vigorous competition has become unlawful exclusion.
Until now, businesses, national authorities, and courts had to piece together the governing standards from treaty text, sporadic Commission decisions, and a growing body of judicial opinions. The Guidelines bring the rules on dominance, exclusionary effects, specific practices, and objective justifications into one framework. They also recognize that firms compete on quality, choice, and innovation, as well as price.
But the Guidelines do not mark a decisive return to disciplined, effects-based enforcement. They allow the Commission, in some cases, to dispense with a counterfactual, assess nonpricing conduct without a price-cost test, infer exclusion from conduct merely capable of producing it, and use presumptions to shift the evidentiary burden. The final text narrows some of the original draft version’s most aggressive express presumptions, as Global Competition Review has noted, but leaves ample room to expand liability later.
The Guidelines thus do more to organize and clarify Article 102 enforcement than to solve its central economic problem. The law remains highly receptive to precautionary intervention against successful firms, even when the challenged conduct may bring consumers substantial benefits alongside possible exclusionary risks.
No one seriously disputes that dominant firms can harm competition. The harder question—and the one these Guidelines do not settle—is whether the law can distinguish that harm from the ordinary process by which firms achieve dominance through investment, integration, innovation, and better service.
Rules of the Dominance Road
The Commission frames Article 102 as a three-stage inquiry. First, the enforcer assesses dominance. Second, it asks whether the conduct distorts “effective competition.” Third, the dominant firm may establish objective necessity or an efficiency justification. The Commission’s Article 102 page says the Guidelines aim to give national competition authorities, courts, businesses, and the Commission itself greater legal certainty and consistency.
The conduct list is long but familiar: predatory pricing, margin squeezes, conditional rebates, exclusive dealing, tying and bundling, access restrictions, refusals to supply, self-preferencing, and conduct “by its very nature harmful to competition.” The Guidelines also address aftermarkets, digital ecosystems, network effects, potential competition, and collective dominance. That breadth is useful. A digital platform or integrated firm cannot be analyzed as a simple, single-product monopolist. At the same time, digital-market characteristics do not, by themselves, establish abuse.
Several propositions deserve applause. The Guidelines recognize that Article 102 does not prevent firms from achieving dominance on their merits. Nor does it prohibit conduct merely because it marginalizes less-efficient rivals. Competition on the merits, they explain, includes lower prices, better quality, wider choice, and new or improved products. The Guidelines also require a theory of harm and say that findings of possible exclusion must rest on specific, concrete facts and evidence.
Recent Article 102 case law points in the same direction, at least in important respects. In Intel v. Commission, the European Court of Justice required the General Court to examine Intel’s evidence and arguments about whether its rebates could foreclose an as-efficient competitor. In Unilever Italia, the Court held that exclusivity clauses must be capable of producing exclusionary effects and that authorities must assess evidence submitted by the dominant firm. The Google Shopping judgment confirmed that self-preferencing by a dominant firm is not automatically unlawful.
These decisions reflect a basic error-cost principle: When conduct has a plausible efficiency justification, a legal label should not substitute for analysis. In their assessment of the draft Guidelines, Dirk Auer and Lazar Radic similarly warned that a selective reading of precedent could read presumptions of illegality into Article 102 while reading effects analysis out.
A High Bar With Plenty of Trapdoors
The trouble is a recurring pattern: The Guidelines announce a demanding principle, then qualify it until considerable administrative discretion seeps back in. The Commission says it must show that conduct is capable of producing exclusionary effects. Yet Article 102 has no de minimis threshold. The effects need not be actual or profitable. They need only be more than hypothetical. A firm cannot necessarily rebut the Commission’s case by showing that enough of the market remains open to accommodate competitors. The practical question is therefore whether exclusion seems plausible, not whether the conduct is likely to reduce consumer welfare by a material amount.
The Guidelines also treat the as-efficient-competitor (AEC) test asymmetrically. The Commission normally uses price-cost analysis for pricing conduct but generally declines to use it for nonpricing conduct because nonprice effects are difficult to quantify. It may instead rely on other evidence, without systematically constructing an alternative hypothetical scenario.
The July 2026 Google Android decision provides the immediate judicial backdrop. The Court upheld Google’s roughly €4.1 billion fine while accepting that the Commission could rely on the economic context without constructing a counterfactual in every respect. It also held that an AEC showing was not always necessary in digital markets.
That approach may be defensible in a narrow case with strong evidence that the challenged conduct makes entry or expansion practically impossible. As a general template, it is more troubling. Product quality, user experience, privacy, security, interoperability, and innovation are difficult to quantify. The greater that difficulty, the greater the risk that qualitative evidence becomes a respectable name for administrative intuition. A rule that lets the Commission avoid quantitative analysis precisely when the benefits and harms are hardest to measure should demand more institutional humility, not less.
Self-preferencing illustrates the problem. The Guidelines properly reject a categorical ban. They require both a departure from competition on the merits and the capability to produce exclusionary effects. But the relevant factors include whether users or market participants expect neutrality or openness, whether preferential treatment is “unjustified,” and whether it influences user behavior regardless of the favored product’s intrinsic qualities.
Those concepts can capture genuine manipulation. But they can also condemn ordinary product design. A platform may rank, integrate, or display its own service because doing so improves quality, lowers transaction costs, protects security, or simply gives users a more useful answer.
The category of conduct “by its very nature harmful to competition” carries even greater consequences. Examples include payments conditioned on not selling a rival’s product, dismantling infrastructure on which a competitor relies, and rules imposed by a firm that exercises both regulatory and commercial functions without sufficiently objective and precise procedures.
Once the Commission places conduct in this category, it deems the conduct to distort effective competition. A firm can successfully challenge that conclusion for lack of exclusionary capability only in “very exceptional” circumstances, while defenses based on objective necessity or efficiency are “very unlikely” to succeed. That walks and talks much like a presumption of illegality, even if the Guidelines decline to call it one.
The treatment of exclusive dealing is more disciplined than the broadest version contemplated in the original draft. The Guidelines establish a presumption for exclusive dealing, including de facto exclusivity, but require evidence of an intent to impose exclusivity in some cases involving volume thresholds. That narrowing helps.
Still, the presumption shifts the evidentiary burden to the firm before the Commission has necessarily shown substantial foreclosure, meaningful market coverage, or likely consumer harm. The Court’s reasoning in Intel and Unilever supports careful examination of a dominant firm’s evidence. It does not make the existence of an exclusivity clause the endpoint of the inquiry.
The efficiency defense offers another partial improvement. The Guidelines recognize cost and qualitative efficiencies, static and dynamic effects, and investments in research and development, innovation, and infrastructure. They also recognize related efficiencies outside the relevant market when the consumers who bear the alleged harm substantially overlap with those who receive the benefits.
But the dominant firm must prove four cumulative conditions: That the conduct produces efficiencies, that those efficiencies counteract the competitive harm, that the conduct is necessary, and that it does not eliminate effective competition. That framework makes sense as an administrable screen. But it may also render the defense illusory when innovation benefits are uncertain, long term, or spread across a platform ecosystem. The firm bears the burden of proof even though the Commission controls the investigation and may define the relevant markets and theory of harm in ways that exclude important benefits from the start.
My earlier analysis of the Commission’s interventionist turn warned that theories involving less-efficient competitors, constructive refusals, and expansive margin-squeeze enforcement could prompt dominant firms to raise prices, curtail discounts, or forgo investments that benefit consumers. My more recent analysis of Google Android makes the same point in dynamic terms: The lasting cost of a questionable precedent lies not just in the fine, but also in the products firms redesign and the integrated platforms they decide never to build.
Regulation by Overlap
The Guidelines make clear that the existence of another rulebook offers no safe harbor. Article 102 may apply even when EU or national regulation already covers the conduct, and compliance with another legal regime does not preclude an Article 102 violation. The practical consequence is a layered enforcement system in which authorities may examine the same conduct under Article 102, the Digital Markets Act (DMA), national abuse-of-dominance laws, consumer-protection or data rules, and sector-specific regulation.
The DMA offers the clearest example. Its gatekeeper obligations govern self-preferencing, steering, data use, interoperability, and other practices through ex ante duties that do not always require proof of consumer harm or the exclusion of an equally efficient rival. Article 102, at least in theory, remains effects-based.
In practice, however, DMA concepts and enforcement experience may shape the interpretation of Article 102. Article 102 proceedings, in turn, can reach nongatekeepers and conduct outside the core platform services designated under the DMA. The result may be “regulation by overlap,” with condemnation under one instrument serving as evidence of illegality under another.
Germany adds another layer. Sections 19 and 19a of the German Competition Act (GWB) apply, respectively, to dominant firms and firms deemed to have paramount cross-market significance. Section 19a specifically reaches self-preferencing, tying, data combination, interoperability, portability, access conditions, and benefits demanded from business users. Section 22 preserves stricter national rules alongside Article 102. The official English text of the GWB makes that point unusually clear.
The Court of Justice’s Towercast ruling likewise confirms that national authorities may use Article 102 to review certain transactions below the Commission’s merger-control thresholds. Decentralized enforcement can uncover local harms, but it can also produce inconsistent standards, duplicative proceedings, and remedies that pull platform design in conflicting directions.
This is not an argument for immunity. It is an argument for institutional coordination. The Commission should identify the instrument best suited to each concern, avoid cumulative punishment based on the same economic theory, and explain how Article 102 analysis differs from an ex ante DMA prohibition. Without those safeguards, legal uncertainty can itself become a barrier to entry—especially for firms that cannot afford separate product architectures and compliance teams across Europe.
Regulating Europe Back to Competitiveness
Europe’s competitiveness problem raises the stakes. The Draghi report on the future of European competitiveness identifies an innovation gap with the United States, weak productivity growth, difficulty scaling startups, fragmented capital markets, and an urgent need for private and public investment. That diagnosis belongs alongside Mario Zúñiga’s observation that Europe’s regulatory zeal has yet to produce a digital-market success story.
The lesson is not that regulation never helps. It is that errors carry unusually high costs in markets marked by rapid innovation, network effects, and uncertain demand. Product integration that appears exclusionary today may supply the investment needed to make a new service reliable tomorrow. A platform practice that diverts traffic from an intermediary may give users a faster, safer, or more useful product. Refusing to share an input may protect security or preserve the incentive to create that input in the first place.
Enforcers cannot fully observe the opportunity costs of prohibiting such conduct. They cannot count innovations never attempted, services never launched, or capital quietly invested elsewhere.
That is the error-cost case for a more demanding threshold for intervention. False negatives can cause real harm, but false positives in dynamic markets can permanently change the course of competition. The danger grows when regulators treat a firm as a quasi-public utility and require it to preserve rivals’ access, margins, or traffic. Competition law should protect the competitive process. It should not guarantee every rival’s survival or preserve every intermediary’s commercial position after a product improves.
Lazar Radic’s analysis of DMA enforcement against Google explains how a better product, integrated results, or a platform fee can become evidence against a firm precisely because success has made it powerful. The broader Digital Fitness Check submission documents how overlapping digital rules obscure both direct compliance costs and the opportunity cost of engineering time that could otherwise support innovation. Those costs do not stay on a compliance ledger. They redirect resources away from discovery.
Who Watches the Competition Watchdogs?
Public-choice analysis offers another reason for caution. Competition authorities are public institutions, not omniscient social-welfare maximizers. They respond to political attention, institutional priorities, limited resources, and the incentives created by their own powers. An investigation of a successful firm delivers visible benefits to complaining rivals and favored constituencies. The costs of a mistaken case—higher prices, less innovation, lost investment, and slower growth—emerge later and fall broadly across the economy. That asymmetry can encourage theories built around fairness, dependence, or ecosystem openness even when evidence of consumer harm remains thin.
The danger extends beyond crude protectionism. European firms may lobby for rules that constrain foreign rivals, while national authorities may favor domestic interests or seek a greater role in directing industrial outcomes. Once competition law dictates how a dominant firm must deal, display products, license technology, share data, or design an ecosystem, the boundary between antitrust enforcement and industrial regulation begins to disappear. The authority stops policing demonstrable exclusion and starts choosing business models.
That distinction matters in a global economy where China is a major source of competition, capital, technology, and state-supported industrial capacity. Europe should not answer Chinese competitive pressure by making its own dynamic firms more cautious. A legal regime that penalizes scale, integration, and risk-taking may push innovation toward the United States or China, leaving European consumers with fewer homegrown alternatives. Europe needs more productive rivalry. It will not get there by making European success legally hazardous.
Putting the Effects Back in Effects-Based Enforcement
The Guidelines can do more to improve Article 102 enforcement without giving genuine exclusion a pass. Five clarifications would help.
First, the Commission should require each theory of harm to identify a concrete, evidence-based mechanism linking the challenged conduct to a material risk of consumer harm. “Capability” should not encompass every logically conceivable path to exclusion.
Second, the Commission should reserve presumptions for conduct with a well-established exclusionary mechanism. Firms must remain free to rebut those presumptions with market-specific evidence. When available, evidence of actual entry, expansion, switching, product quality, and user behavior should count.
Third, the Guidelines should make the AEC test, or another economically coherent comparison, presumptively available whenever the conduct permits quantification. If quantification proves impossible, the Commission should explain why and identify the qualitative evidence serving in its place.
Fourth, the Commission should assess dynamic efficiencies at the same stage and with the same rigor as exclusionary effects. Long-term innovation, security, privacy, resilience, and infrastructure investment should not be relegated to a defense that becomes practically impossible to prove after an elastic finding of abuse.
Finally, the Commission and national authorities should develop a coordination protocol covering Article 102, the DMA, and national rules such as Section 19a. Firms should not face mutually inconsistent design mandates. Nor should an authority impose a remedy under one instrument without considering obligations already imposed under another.
Clarity Without Restraint
The Commission deserves credit for replacing its 2009 enforcement-priorities document with a comprehensive framework. The final Guidelines organize the law, acknowledge dynamic competition, reject a categorical offense for self-preferencing, recognize innovation efficiencies, and narrow some of the draft’s sweeping presumptions. These are real advances.
Yet organization does not guarantee restraint. The Guidelines preserve a broad conception of dominant firms’ special responsibilities, a low threshold for exclusionary capability, extensive reliance on qualitative evidence, limited counterfactual analysis, and a demanding efficiency defense. Combined with the DMA and national regimes such as Germany’s Section 19a, that framework may increase the risk that authorities regulate successful firms for the consequences of winning.
Europe needs competition policy rigorous enough to stop genuine exclusion and humble enough to recognize the limits of administrative knowledge. In the global contest for innovation, that requires taking false positives seriously, preserving firms’ freedom to integrate and invest, and distinguishing the protection of competition from the protection of competitors.
The Guidelines make Article 102 easier to navigate. The question is whether they also make overenforcement easier to justify.