Breaking Up Google

In late August, Judge Amit P. Mehta of US District Court for the District of Columbia found Google guilty of maintaining an illegal monopoly in online search. Google had paid billions to device manufacturers and browser developers—including Apple, Samsung, and Mozilla—to set its search engine as the default in web browsers and smartphones, enabling the company to hinder competition and dominate the market. This landmark ruling was followed by a Virginia judge formally initiating the trial of a second US Department of Justice (DOJ) antitrust suit against Google in September. This second lawsuit, originally filled in January 2023, denounced Google’s acquisitions related to its monopolistic digital ad technology.

The case demonstrates a transformation in the US government’s approach to antitrust. Since the late 1970s, US lawmakers have reduced  antitrust policy to a minimum, using it to guarantee “consumer welfare via prices.” This made it almost impossible to prevent any merger or acquisition—including the 1999 merger of Exxon and Mobil, which the US Federal Trade Commission (FTC) approved despite predicting that the merger between the second and fourth largest corporations in the energy market would increase concentration. The next year, the pharmaceutical giant Pfizer acquired Warner-Lambert for $90 million and became the world’s second largest drug company. Once more, the US FTC cleared the deal. Permissive antitrust policy arrived at a climax during Trump’s presidency, with fewer criminal antitrust cases brought to courts than any administration since the 1970s. 

Biden’s appointment of Lina Khan as the US FTC Commissioner signaled the start of the turn against this permissive regime. Khan is part of a collective of antitrust lawyers that define themselves as the “New Brandeis Movement,” aiming to retrofit antitrust by returning to a more comprehensive antimonopoly ruling that sees the concentration of economic power as a basis for the concentration of political power—thereby shrinking democracy. Instead of focusing on consumer welfare, the New Brandeis Movement proposes to look broadly at the effects of market power on workers, suppliers, and innovators. 

The August ruling against Google’s illegal monopoly in search and ads is an outcome of this policy shift. It took four years for the DOJ to gather the evidence to win the case, and it may take much longer to actually terminate Google’s monopoly. The DOJ’s proposed solutions,  from fines on Google to forcing the company to divest from its search engine tools, are insufficient to resolving the problem at hand. This is because they are rooted in a fundamental mischaracterization of the issue: Google’s search engine is not just a monopoly, but a natural monopoly underpinned by an intellectual monopoly.

Natural monopoly

Contemporary understandings of antitrust closely link it to the promotion of market competition. In fact, in many parts of the world, from the UK to Argentina, antitrust agencies are called competition authorities. Even within neoclassical theory, however, natural monopolies are an exception to this framework. In a natural monopoly,  a single company is more efficient than two or more offering the same product. These are generally markets with economies of scale due to high fixed costs. 

Google’s search engine constitutes a natural monopoly. Multiple search engines would be less efficient than one, particularly given that the engine improves with additional searches. Far beyond the illegal contracts with device manufacturers and browsers that the DOJ antitrust case found, the company’s success depends on social data and is powered by deep learning artificial intelligence (AI) algorithms that become better the more data they process. Google’s search results are the social outcome of our collective use of the same AI model. If searches had been scattered in a dozen search engines, these algorithms would be significantly poorer.

There is yet another peculiarity of search engines. Drawing on the work of Vilfredo Pareto, advocates of perfect market competition define maximum social welfare through the pursuit of an equilibrium in which no agent can be better off unless someone else is worse off. The very idea of maximization of social welfare in the competitive market, however, says nothing about how that welfare is distributed. If a company could spot the maximum willingness to pay of each consumer, it could charge differentiated prices so that it extracts the most from everyone. Social welfare would be at its maximum, but it would lie in the exclusive hands of one actor: the perfectly discriminating monopolist. This strategy is presented in the most used microeconomics textbooks, authored by Hal Varian, who became Google’s chief economist in 2002. 

Unsurprisingly, Google operates as a perfectly discriminating monopolist. Because Google knows exactly which ad a user will most likely select, it can run its ads market as a live auction in which each appropriated piece of the online space is charged at the maximum price that clients are willing to pay. The effect is the full extraction of social welfare in Google’s hands as it not only charges the maximum price to its clients in the ads markets, but operates in the same way with users, maximizing the extraction of data and time—attention—from each individual.

Intellectual monopoly

Intellectual monopolies pose another challenge to the neoclassical framework. They systematically capture intangibles and turn them into assets, accumulating at the expense of socially produced knowledge.

The US antitrust case against Google recognized that its technology is extraordinarily good. Yet Google’s frontier machine learning algorithms are co-created with a global network of thousands of organizations including universities and start-ups. Google is an intellectual monopoly because it coproduces its technology with thousands of others while keeping most of the associated and skyrocketing rents. Google has published thousands of scientific articles on AI. These articles rarely disclose the models but instead only summarize the findings, signaling Google’s central role in the field while preventing society from the spillovers of shared knowledge. Not even Google’s direct collaborators can profit. While over 80 percent of this research had at least one non-Google coauthor, only 0.3 percent of Google’s patents are co-owned with another organization.

Google has acquired over 200 companies, thus acquiring their talent and intangible assets. DeepMind, Google’s AI heart, was acquired in 2014. Yet, at least as important for perpetuating its intellectual monopoly are the thousands of firms that it nurtures with venture capital funds. In 2023, Microsoft, Google, and Amazon invested two-thirds of the $27 billion raised by generative AI start-ups. By February 2024, Google ranked as the third-largest funder of AI companies, in terms of the number of AI companies that had received its funding. Only two venture capital firms, Techstars and Y Combinator, were investing in more AI start-ups. Expanding the analysis to the whole start-up universe, the relevance of corporate venture capital becomes all the more apparent. The practice has spread among Big Tech, and Google is at the forefront: 2,445 active start-ups have Google among their top five investors by the same date.

Corporate venture capital is a form of control with partial ownership that distracts regulators while Big Tech companies steer start-ups’ development and gain privileged access to their technologies. A recent example is the generative AI start-up Anthropic, founded by former OpenAI employees that left the company once Microsoft started investing in it. Only three months after the release of ChatGPT, Google decided to invest $300 million in Anthropic (Anthropic was ultimately unable to develop and train frontier AI models without a Big Tech sponsor, given high costs and a lack of access to key pieces of the computing stack, which are monopolized by Big Tech).

A network of start-ups and other firms that do not receive Google funding are equally controlled. They work on different steps of the AI value chain that are offered as services on Google’s cloud. Just like in the case of Google Pay—Google’s platform app, which has also been subjected to legal scrutiny—Google gets a cut from every computing service sold on its cloud.

The open-source community also benefits Google’s accumulation. One key example is TensorFlow, a  software library for machine learning open sourced by Google. Software libraries are sets of predefined tools that perform common or repetitive tasks for software development. Around one third of the contributors to improve TensorFlow are not from Google. Open sourcing TensorFlow has made it the industry standard: everyone uses TensorFlow for machine learning models. This leads to more tutorials and discussion forums explaining how to use it or how to solve common errors. More shared knowledge further incentivizes its use and creates network effects. As more developers produce applications and other complementary products using TensorFlow, the better these solutions will work as complements to Google technologies. The result is a structure in which the whole industry sooner or later depends on Google technologies or produces technology for Google, sometimes without even realizing it. Other cloud giants like Amazon or Microsoft similarly take advantage of blind knowledge transfers. Google profits from developers’ free labor, as open-sourced technologies receive thousands of contributions from outside of the company. Improved technologies are then refined with other pieces of software and models kept secret by Google, allowing the technology to become economically useful. Microsoft and, more recently, Meta have actively used this strategy. 

Tracing these links offers a panopticon view of the AI value chain. Just like Microsoft and Amazon, Google weaponizes this interdependence, setting the rules of the whole network from AI research to applications. Traditional ways to dismantle illegal market behaviors will not affect its stranglehold on the AI value chain.

Competition is not the solution

Precisely because Google is a natural and an intellectual monopoly, the remedies for its excessive market power ought to be different from those applied to a regular market monopoly. Under a natural monopoly, boosting competition would come at the expense of efficiency and worsen search results. Moreover, increasing competition in this market would not make it free or fair. The winners would be Apple and Microsoft, whose search engines would receive more traffic. A far cry from Robin Hood, this solution would be taking from the rich to give it to the other rich. Even if Google were forced to divest and disaggregate, the new companies could still share databases and research results to maximize extracted rents, leaving the intellectual monopoly untouched. In China, Alibaba is legally separated from Ant Group, but they still cross-reference datasets to boost each business and each of the companies equally relies on knowledge captured from other organizations as a basis of their intellectual monopoly. 

The problem of the DOJ’s potential divestment remedy—which would require Google to divest from parts of its search engine tools—is that companies share information on a need-to-know basis both inside the firm and within their networks. Microsoft and OpenAI share intangibles as part of their agreement. The same could happen with the resulting new firms from a carved-up Google. These mid-size Googles could still sign strategic partnership agreements to codevelop solutions and collaborate in research and development while continuing knowledge and data extraction from their ecosystems, cross-fertilizing and perpetuating their intellectual monopolies. Such strategic agreements are often signed between Google’s strategic solutions teams and other leading corporations. Many are kept secret to avoid antitrust scrutiny.

If the harms of natural and intellectual monopolies cannot be mitigated through the imposition of competition or their disaggregation, what can  prevent a giant corporation from amassing billions by controlling the organization and classification of the Internet’s information?

State-run companies often provide public utilities that would otherwise create natural monopolies. One step forward could be to treat Google like the private companies that offer public utilities at regulated prices. Setting maximum prices for ads and limiting ad space would reduce the price—in attention and data—that users pay for Google searches. An international agreement among states would be required to enforce such a regulation. Indeed, online search engines should be seen as a public utility not only because they are part of our everyday life, but also because they are endlessly improved with our queries. 

Google should additionally pay withholding taxes country by country for its monetization of freely harvested data. Such a tax could consider the percentage of the population with internet connectivity and the number of hours that people spend on the internet. But the disagreements concerning a digital tax levied on Big Tech suggests that the US state would reject such an initiative. Nonetheless, we should aim to build international cooperation around such a proposal. 

Taxing data and enacting price controls will pierce Google’s main profit source. But Google would keep profiting from society’s data and knowledge. To break the company’s intellectual monopoly, we must first recognize that the world’s largest companies are robbing us all. International organizations should defend the control of our data and publicly funded knowledge, moving us towards a society that democratically determines what data is collected, who can access it, and for what purpose.