Google's Android antitrust defeat and a 37% power surge: the cost of winning AI

A 37% jump in contracted electricity capacity, booked in a single quarter by the operator of the world's dominant mobile operating system, is the kind of figure regulators usually meet with a phone call. On 1 July 2026, the European Commission delivered something rarer: a conviction. The Commission found that Google had abused its dominant position in Android app distribution, imposing remedies that re-open a chapter of competition policy the company thought it had closed in 2018. The energy disclosure landed the same week. Read together, the two events describe a single platform under two different kinds of pressure at once.

The headline is the legal one, and it deserves to be set out plainly. The Commission's decision, building on its 2018 Android finding and the subsequent 2022 General Court judgment that largely upheld it, treats app-store tying and default-bundling practices as ongoing abuses rather than historical ones. Remedies under discussion in Brussels include the unbundling of Google Play from the broader Google Mobile Services suite, the mandatory offering of alternative app stores and billing systems on Android devices sold in the EU, and a possible fine calculated as a percentage of group turnover. None of these levers is theoretical; all of them have been used, in some form, against other gatekeepers under the Digital Markets Act. The Commission is now signalling that Android, long treated as a special case because of its open-source code, will be measured by the same yardstick.

The capacity that has to be built

The other number in the air belongs to a different ledger. According to industry analyses of hyperscaler power contracts, Alphabet disclosed in its most recent sustainability and infrastructure filings that its contracted electricity capacity has risen by roughly 37% year-on-year, driven almost entirely by data-centre build-out to support AI training and inference workloads. That figure is not a Google press release. It is an aggregate reading of disclosed power-purchase agreements, grid interconnection queue positions, and state utility filings, cross-checked against the company's own environmental disclosures. It is also consistent with the direction of travel across the industry: Microsoft's, Amazon's and Meta's disclosed capacity has grown at comparable rates over the same window.

The two stories collide at the data centre. A regulatory order to unbundle Google Play from Google Mobile Services does not directly change the cost of training a frontier model. It does, however, change the cash-flow profile of the business that has to fund the build-out. Android, Search, and the Play Store have been treated inside Alphabet as the cash engine that lets the company absorb the capital intensity of AI infrastructure. If that engine is forced to accept lower take-rates, more competitive defaults, and the loss of exclusive pre-installation deals with handset makers, the operating margin that subsidises the compute arms race is narrower than the financial statements have implied.

The open-source wrinkle

Critics of the original 2018 case always pointed to Android's open-source licensing as a reason the remedy should be light-touch. The Commission's response, in the 2026 decision, is to treat the licence as a contractual artefact rather than a structural one. Open-source code can still be distributed through a closed channel. The relevant market is not the kernel; it is the route by which a typical European consumer actually installs software on a typical European handset. On that route, Google Play remains dominant, and the tying of the Play Store to Google's billing, search, and browser defaults remains the conduct under attack.

The Commission's theory of harm is straightforward, and it borrows from the Digital Markets Act's self-preferencing logic. A phone that ships with Google Play pre-installed, with Google's payment system as the only friction-free option, and with Google's search box placed prominently on the home screen, is a phone on which the gatekeeper's downstream businesses face a structurally lower cost of distribution than any rival. The remedy package now on the table attempts to flatten that structural advantage: alternative app stores must be installable without technical friction, alternative billing must be permitted for in-app purchases, and the default search engine auction must be re-opened under conditions that the Commission, not the company, sets.

What the remedies actually do

A fine, even a large one, is a one-off. The structural remedies are what change the platform. Three of them matter most.

First, mandatory third-party app stores. If implemented as proposed, handset makers shipping Android in the EU will be required to make at least one certified alternative store discoverable during device setup, and to refrain from contractual or technical measures that steer users back to Google Play. The Commission is borrowing directly from the remedies imposed on Microsoft Windows a decade earlier, with the important difference that the mobile ecosystem is, if anything, more concentrated than the desktop one was in 2009.

Second, billing interoperability. Developers selling digital goods and services through an Android app in the EU will gain the right to use a payment processor of their choice, with Google permitted to charge a reduced fee on the transactions it processes, and prohibited from charging any fee on the transactions it does not. The revenue model of the Play Store, which depends on the spread between the developer's price and Google's 15-to-30% commission, narrows.

Third, the search default. The Commission is moving to dismantle the current default-search contracts with handset manufacturers, on the grounds that the revenue-share terms have foreclosed competing general search services. A re-opened auction, with visibility and revenue split mandated by the Commission rather than negotiated by Google, is the likely outcome. Combined with the ongoing DMA proceedings on self-preferencing in Search itself, the cumulative effect is to convert Google's distribution advantage in general search from a settled fact into a contestable one.

The energy bill nobody is reading

The 37% capacity figure belongs in this story for a reason that has nothing to do with antitrust law and everything to do with capital allocation. AI infrastructure is a power-infrastructure story. New training clusters are sited where gigawatts of firm capacity can be contracted years in advance, and where the local grid operator will not throttle the load during a heatwave. Hyperscalers have, over the past 18 months, become some of the largest single corporate buyers of long-dated power in the United States, and increasingly in Northern Europe. The contracts are public, the grid filings are public, and the disclosed capacity is public.

What is less public is the cost. Power-purchase agreements signed at 2024 and 2025 prices are now being delivered into a market where the marginal cost of compute is rising even as the per-token inference price is falling. The gap is funded by the operating cash flow of the platform businesses. If that cash flow is compressed by a regulatory order, the build-out is not stopped. It is repriced. Either the company accepts a lower return on its AI capital base, or it pushes the cost of compute into the prices charged to enterprise customers, or it slows the deployment of new capacity relative to competitors that are not subject to the same remedies.

The structural frame

This is the part of the story that the legal coverage tends to miss and the energy coverage tends to miss in the opposite direction. The Commission is not regulating Android because it dislikes bundling. It is regulating Android because the platform's distribution advantage is the input that the company is using to subsidise a different business, in a different market, with a different competitive geometry. AI compute is being financed, in significant part, by the margins of mobile software distribution. A regulator that wants to constrain the second business has, in the first, a handle.

The same handle is available to the company's competitors, and they know it. App-store rivals, alternative billing providers, and the European handset makers that have long complained about the terms of their Google contracts have been pushing the Commission toward exactly this set of remedies for the better part of a decade. They will not be the only beneficiaries. The enterprise cloud and AI infrastructure market, in which the same company competes, is now watching a proceeding that may change the cross-subsidy that has, until now, let that business absorb capital costs that its rivals could not.

What to watch before the end of 2026

The decision is a finding of infringement. The remedies will be set out in a separate instrument, with a consultation period that runs into the autumn. The fine, if one is imposed, will follow the usual Commission practice of being calculated on the basis of group turnover in the relevant geographic market, and is likely to be material but not catastrophic on its own. The structural remedies are the part that will outlive the fine.

The energy story does not have a comparable regulatory milestone on the horizon. The 37% capacity figure is a snapshot of contracted, not consumed, power. The consumed figure will appear in subsequent environmental disclosures, and the gap between the two will be the most useful single number for tracking whether the build-out is being delivered on schedule, repriced, or trimmed. That is the figure worth watching from outside the company, because the antitrust decision and the capacity disclosure are not two separate stories. They are the same story, told in two registers, about a platform whose cost of winning the next cycle just went up.

Sources: The European Commission's Android antitrust decision of 1 July 2026 and the underlying 2018 decision (Case AT.40099), as reported by Reuters and the Financial Times; Alphabet's most recent sustainability and infrastructure disclosures, including 10-K and CDP filings, cross-referenced with state utility commission dockets in Virginia, Ohio and Ireland; industry analyses of hyperscaler power-purchase agreements published by BloombergNEF and S&P Global Commodity Insights; the Digital Markets Act designations of Google as a gatekeeper, and the Commission's prior DMA decisions on self-preferencing in Search.

Desk note: Monexus has tracked the EU's Google cases as a single arc since the original 2018 decision, and has flagged in earlier coverage that the European approach to platform regulation diverges from the United States on remedy design. This piece is the first to read those cases alongside the company's energy disclosures as two surfaces of the same platform-under-pressure story.