On 2 July 2026, the Court of Justice, sitting as the Second Chamber, dismissed the appeal brought by Google LLC and Alphabet Inc. against the General Court’s 2022 judgment in the Android case (C‑738/22 P). The fine stays at 4.125 billion euros, the figure the General Court had already set after trimming the Commission’s original 4.34 billion. The result was widely expected. The reasoning is where the judgment earns attention, because the Court used the occasion to settle a methodological question that reaches well beyond Google.
The question concerns the as-efficient competitor test. For readers outside competition law, the test asks a single thing. If a rival were exactly as efficient as the dominant firm, meaning it had the same costs and the same capabilities, could that rival survive the conduct under scrutiny? If even an equally efficient rival would be driven out, the conduct looks exclusionary. If such a rival could cope, the conduct starts to look like ordinary competition. The test entered EU law through pricing cases, where it can be run with figures, and it became the reference point for abuse analysis after the Court’s 2017 ruling in Intel (C‑413/14 P).
Google built much of its appeal on that test. It argued that the Commission had condemned the pre-installation obligations in the Mobile Application Distribution Agreements, the contracts requiring device makers to pre-load Google Search and Chrome as the price of the Play Store, without ever showing that those obligations could foreclose an equally efficient rival. Google pressed the point by contrast. The General Court had applied the as-efficient competitor test to Google’s revenue share agreements, the payments made to manufacturers on condition that no rival search service was pre-installed, and had found the Commission’s version of the test flawed enough to annul that part of the decision. The same test, Google said, should have governed the pre-installation obligations.
The Court declined. At paragraph 272 it repeated the position it had taken in Google Shopping (C‑48/22 P, 2024), that Article 102 TFEU imposes no systematic obligation to prove a practice could exclude an as-efficient competitor. It then went further than any earlier ruling. At paragraphs 273 and 278 the Court held that in markets shaped by network effects and high barriers to entry, the digital ecosystems that Google itself illustrates, the test does not merely become optional. It stops making sense, because such markets can make the entry or survival of an equally efficient rival practically impossible whatever that rival does. The Court closed paragraph 278 with a sentence that will be quoted for years, stating that this holds all the more where the conduct does not lend itself to a quantitative, price-based analysis. Tying, the Court reasoned at paragraph 269, deviates from competition on the merits by its own nature, so no efficiency comparison is needed to identify it.
This is the doctrinal core, and it repays a close look at what the Court cited and what it passed over.
The support for the holding is real. Google Shopping said in terms, at its paragraphs 263 and 264, that a finding of abuse is not conditioned on proof that the conduct could exclude an equally efficient rival. The Court quoted those paragraphs almost verbatim. The line runs back through Post Danmark II (C‑23/14, 2015), where the Court had already called the test not indispensable, to Post Danmark I (C‑209/10, 2012). The precedent path is short and intact, which matters, because a holding this consequential resting on thin authority would invite reversal.
The result is not the whole story. The way the Court reached it leaves two seams. To arrive at its holding the Court leaned on part of its 2017 Intel ruling while stepping past the part that points the other way, and the framework it used for the tying half of the judgment had never been reviewed at this level until now. Neither seam is fatal to the outcome. Both are the kind of gap a future defendant reopens, and neither is visible unless you read the judgment against the line of cases behind it.
The practical consequence is straightforward. The Commission can establish tying and pre-installation abuses against platform firms by documenting how the conduct works and what it forecloses, without constructing the quantitative efficiency model that Intel had seemed to demand. For enforcement under the Digital Markets Act, which already dispenses with case-by-case effects analysis for designated gatekeepers, the judgment removes a line of resistance that gatekeepers might otherwise have imported from Article 102. Google Shopping opened this door in 2024. Android has now walked a platform tying case through it and named the market conditions, network effects and entry barriers, that justify leaving the efficiency test behind.
The exposure does not end at the fine. A confirmed abuse feeds the follow-on damages claims moving through national courts, and the Court has already shown in Heureka Group v Google (C‑605/21, 2024) that it will read limitation rules generously enough to keep those claims alive. Every manufacturer, rival, and complainant that can trace harm to the pre-installation obligations now litigates against a finding the Court of Justice has placed beyond further appeal.
The Court affirmed the General Court on every ground and changed the law while doing it. The affirmance was the headline. The retirement of the as-efficient competitor test from a whole class of digital cases is the part worth keeping. Which precedents the Court leaned on, and which it stepped past to reach that result, is the part worth watching.
**
For paid subscribers, the full audit follows. It splits the judgment into its seven distinct holdings, checks each cited paragraph against a corpus of over 1,000 competition rulings, traces the precedent path case by case, and names the authorities the Court did not cite, including the two paragraphs of Intel that decide the whole question and the tying precedents the Court passed over. It is built to drop into a brief.
