On 23 July, the European Commission fined Google €890 million across two decisions, the first non-compliance findings against the company under the Digital Markets Act. The Search decision imposes €460 million under Article 6(5) for treating Google’s own shopping, hotels, transport and sports results more favorably than rival services. The Play decision imposes €430 million under Article 5(4) for restricting developers who want to steer users toward cheaper purchase channels. Google has sixty days to comply, after which periodic penalties of up to five percent of average daily worldwide turnover become available.
The press release runs to a page. The decisions themselves are not published. What follows is a list of arguments to check rather than a final assessment.
What the Court refused to say in 2024
The Court of Justice closed the Google Shopping litigation in September 2024. The case had run for just under fifteen years, from Foundem’s complaint of 3 November 2009 through the Commission’s 2017 decision to the final judgment. In closing it, the Court refused to hold that a dominant firm favoring its own services departs, as a general rule, from competition on the merits. Google lost, but it lost on the specific circumstances of its conduct, the characteristics of the general search market, the traffic dynamics, the demotion algorithms. The Court took care to say the condemnation did not travel automatically.
Article 6(5) of the DMA is that general rule. It prohibits gatekeepers from self-preferencing in ranking, per se, no circumstances required. On 23 July, the Commission applied it to the same company, for the same conduct family, extended to three verticals the 2017 decision never reached. The proceedings, opened on 25 March 2024, took twenty-eight months. The press release does not mention the 2017 Google Shopping decision or the litigation that followed it, cites no Article 102 precedent, and contains no statement of continuity with the case law. Whether any of that appears in the decisions themselves is the first thing to check.
The rule the Court declined to write in 2024 was already in the Regulation. On 23 July the Commission enforced it against the company whose own litigation had produced the refusal.
The obvious defense is that the DMA is a different instrument with a different legal basis, and that a legislature may regulate what the case law will not condemn. The consequence holds anyway. On self-preferencing, the DMA and EU competition law point in opposite directions for the same firm and the same conduct. Under Article 102, Google’s favoring of its own results is unlawful only in circumstances a court must examine. Under Article 6(5), it is unlawful always. A company designing a search results page in Europe must now satisfy two legal standards that diverge at the level of principle, administered by the same institution, with the stricter one adopted without any stated position on the case law it outruns. Whatever one thinks of self-preferencing as a theory of harm, this is a recipe for legal uncertainty, and nobody should expect it to settle as the two regimes bed in, because the divergence sits in the design of the DMA rather than in the transition to it.
Two prices, one morning
Self-preferencing in search cost Google €2.42 billion under Article 102 in 2017. It costs €460 million under the DMA in 2026, for a wider set of verticals. A fifth of the earlier figure, for more conduct.
Steering runs the other way. The comparison needs care. The Commission fined Apple €1.84 billion in March 2024, but only €40 million of that came from the 2006 Guidelines methodology. The rest was a lump sum added for deterrence and for non-monetary harm. Measured against the basic amount, the €430 million imposed on Google is more than ten times the Article 102 price of the same conduct. Measured against the headline, it is a quarter. The honest comparison depends on what the lump sum was for, so if it priced the conduct the headline figure is the right one, and if it priced Apple’s size the €40 million is.
Now, some basic arithmetic. A DMA fine sanctions a compliance failure measured in months, an Article 102 fine an infringement measured in years. The 2017 decision found the abuse beginning in January 2008 in Germany and the United Kingdom, October 2010 in France, May 2011 in Italy, the Netherlands and Spain, and November 2013 in six further countries, and running to the decision in all of them. The duration multiplier therefore differs by market, and the €2.42 billion carries a weighted average somewhere between the 5.4-year unweighted mean and the 9.5-year maximum. The DMA fine covers non-compliance since the obligations bound in March 2024, a little over two years. Annualized, €460 million comes to between two fifths and three quarters of the 2017 rate.
The same calculation on steering runs the other way. The €40 million basic amount against Apple covered a continuous infringement from June 2015 to March 2024, nearly nine years. The €430 million against Google covers a little over two years. Annualized, the DMA rate is roughly forty times the Article 102 rate for the same conduct.
Under the 2006 Guidelines the basic amount is the value of sales to which the infringement relates, multiplied by a gravity percentage of up to thirty percent, multiplied by duration. Apple’s €40 million was small because its App Store commission revenue on European music streaming subscriptions is a small base. The €1.8 billion came from point 30 of the same Guidelines, which permits an increase for undertakings whose total turnover far exceeds the sales concerned. Article 30 of the DMA has no value-of-sales step at all. It caps the fine at ten percent of total worldwide turnover and instructs the Commission to weigh gravity, duration and recurrence. No fining guidelines exist under the DMA.
What changed is the denominator. Under Article 102 the fine is anchored to the revenue the conduct touched, under the DMA to the size of the firm. On steering that swap alone accounts for the factor of forty, and I want to see which anchor the published decision uses.
The nine-year paradox
After the 2017 decision, Google introduced its CSS auction in September 2017. Article 4 required a first compliance report on the day the infringement effectively ended, then one every four months for five years, sixteen in all. The Commission monitored on that basis until 2022 and never adopted a non-compliance decision under Regulation 1/2003. It did not stay silent either. Margrethe Vestager said publicly in November 2019 that the mechanism was not bringing traffic to comparison shopping rivals, and the complainants pressed throughout for non-compliance proceedings. So the Commission said the remedy was failing and declined to find that it was unlawful. The mechanism has now run for close to nine years, its first five under formal monitoring.
Under the DMA, shopping is fined again. Shopping deserves to be taken on its own here, because hotels, transport and sports sat outside the 2017 remedy and the Commission never examined them there. Either Article 6(5) imposes a substantively stricter standard than the mechanism the Commission monitored for five years and declined to condemn, or the Commission is now fining under the DMA what it left standing under Article 102.
There is a third answer available. Google rebuilt the European results page for the March 2024 deadline, so the Commission may say it is fining an architecture built after the monitoring stopped. That answer holds only if the favoring mechanism changed and not merely the interface. The press release confronts none of this, and which of the three branches the published decision takes will tell us more about the DMA’s real content than any compliance workshop.
The Commission is setting the fee, not banning it
There is a fourth thread, quieter but consequential. The Play decision concedes that Google may charge fees when it first brings a customer to a developer, and objects only that the level of the fee and the length of the charging period went too far. The Commission is not prohibiting the fee. It is fixing a view about what level and what duration would be acceptable, which is something the competition case law spent decades saying an administrative authority should avoid doing.
What follows for paid subscribers is the audit. I have run the press release against two corpora, the full set of published Commission antitrust decisions and the Court’s antitrust judgments, and set out six exhibits pinned to the operative text. The first two put the rule the Court refused beside the rule the Regulation enacted, then reproduce the 2017 operative part that describes Thursday’s finding almost word for word. The last returns to the daily penalty clause that has sat in the 2017 decision, available against Google the whole time, never once applied.
