John B. Kirkwood explains six ways in which Big Tech’s alliances with AI startups could harm competition, making clear that the antitrust agencies have good reason to monitor and investigate them.
Editor’s Note: This article is part of a symposium which asks experts to evaluate the anticompetitive harms of Big Tech investments in AI startups in light of recent investigations from antitrust agencies on both sides of the Atlantic. See here to read the other contributions from Matt Perault and Vivek Ghosal. Return tomorrow for the final contribution from Stacey Dogan.
The federal antitrust agencies are increasing their scrutiny of the tech giants. They have already brought lawsuits against Amazon, Apple, Facebook (Meta), Google, and Microsoft, accusing them of monopolization or making illegal acquisitions. The Federal Trade Commission (FTC) has moved to block Microsoft’s acquisition of Activision and unwind Facebook’s purchases of Instagram and WhatsApp, and last fall the FTC sued Amazon for monopolizing segments of the online marketplace. The Department of Justice (DOJ) has been even more active. It has filed three monopolization actions against the tech giants: one accuses Google of monopolizing general search services and search text advertising (which just resulted in a district court ruling in the DOJ’s favor); a second alleges that Google monopolized the buying and selling of digital ads; and a third claims that Apple has monopolized the sale of smartphones.
Now, the agencies have turned to the alliances the tech giants have recently entered into with artificial intelligence startups. These alliances allow the tech giants to license the startups’ products and data in return for providing the startups with key inputs like capital and computing power. The collaborations have proliferated in the last few years and, despite their obvious benefits for the parties and consumers, the agencies are concerned that the alliances may suppress competition. The issue is so important that four antitrust enforcers on two continents—the European Commission, the United Kingdom’s Competition and Markets Authority, the DOJ, and the FTC—have issued a joint statement describing their concerns.
Those concerns have led to the opening of multiple investigations. The FTC is investigating Microsoft’s partnerships with OpenAI and Inflection AI, and Amazon and Google’s alliances with Anthropic. The DOJ is investigating Nvidia’s alliance with Arm. Other collaborations—such as Amazon’s deal with Adept, Apple’s investment in OpenAI, and Arm’s alliances with Microsoft and Apple—may also become targets of federal scrutiny. Authorities in the U.K and EU have replicated many of these investigations.
This article explains how these alliances could reduce competition. It begins by describing the ways in which the alliances could stimulate competition and then sets forth six types of anticompetitive harm. While the magnitude of these harms will depend on the facts of each case, the agencies have ample reason to be concerned.
It is easy to see how Big Tech-AI collaborations could enhance competition. A Big Tech firm gets to license an AI startup’s products and data, which expands the firm’s product line and may speed up its development of new products. Given that this is occurring in at least three of the Big Tech companies, these firms could see more competition between them as a result. At the same time, the startup gets inputs from the Big Tech firm that it could not, or could not readily, obtain on its own, which may sharply increase its ability to compete.
Access to huge amounts of capital may be the most important input. While America’s capital markets are the largest and most efficient in the world, an AI startup may obtain more money more quickly through an alliance with a tech giant than by raising funds on the open market. For example, OpenAI turned to Microsoft for an infusion of $13 billion. Likewise, a tech giant may supply knowledge and expertise that a startup lacks. Similarly, a large tech firm’s enormous computing power, vast reservoirs of data, and specialized microchips may be critical to an AI startup’s success. A tech firm’s established channels of distribution and skills in marketing may enable a startup to grow more quickly.
Although there is procompetitive logic to these partnerships, there are also six significant risks that warrant the current antitrust investigations.
First, the alliances may result in the elimination of an AI product or computer chip developed by the startup. This is likely to occur when a startup’s product or chip competes with a product or chip made by the Big Tech firm and the latter concludes that its profits would be greater if the startup’s product or chip were dropped. Microsoft, for example, may decide that one of OpenAI’s products competes too closely with one of Microsoft’s products and eliminate the OpenAI product. This is the same dynamic that operates in a “killer acquisition,” in which an established firm acquires an entrant and then shuts down the entrant’s competing product or development project.
There is no evidence, to my knowledge, that such a direct reduction in competition has actually occurred in any of the Big Tech-AI alliances. Nevertheless, the antitrust agencies should monitor for such behavior, especially as the Big Tech firms ramp up the rollout of their AI products such as Lex (Amazon) and Gemini (Google). One critical issue is whether the loss of an AI product would be quickly made up by the creation of a similar product by another firm. That is possible, since there is a great deal of AI development going on, but it takes considerable time and expense to create a new product.
The second anticompetitive possibility is more subtle. Instead of killing a competing product, the alliance may slow down its development. Microsoft may reduce its investment in OpenAI products that threaten to take business from its own products or may press OpenAI to steer its R&D in other directions. Mark Lemley and Matt Wansley summarize this fear in a recent New York Times op-ed. They argue that “in recent years, a handful of incumbent tech companies have sustained their dominance. Why? [They] have learned how to co-opt potentially disruptive start-ups before they can become competitive threats.” These alliances supply the means to curb disruption.
Similarly, Microsoft may reduce or redirect its own R&D to avoid creating products that compete with Open AI’s products. Most of the Big Tech firms are competing to build their own AI chips and large AI models. Coordination with the startups risks strategic suppression of product development where access to firm secrets convinces Big Tech managers they would not benefit from competition or do not have the comparative advantage to compete.
Often, a reduction in innovation may be difficult to detect without internal documentation that reveals it. In some instances, however, a Big Tech firm takes public and material steps that affirmatively undercut a startup’s ability to innovate. For example, Microsoft has hired away many of the key employees of its alliance partner, Inflection AI: its co-founder and CEO, its chief scientist, and several of its senior engineers and researchers. While this massive poaching may have helped Microsoft, it almost certainly reduced the capacity of Inflection to conduct R&D. The Wall Street Journal reports that deals like this are becoming more common and that many in the industry call them “acquisitions in everything but name,” a characterization that may not be entirely accurate since these hires do not transfer voting control of the startup to the tech giant.
The third anticompetitive possibility is a standard antitrust effect. The Big Tech-AI alliances may not reduce the number of products offered or slow their development, but they may cause prices to rise. Often the easiest way for competitors to reduce competition among themselves is to agree, expressly or tacitly, to charge higher prices. To my knowledge, no one has detected such collusive price increases, but if a Big Tech-AI partnership is likely to increase the risk of express or tacit collusion, it can be enjoined. Determining whether this is the case requires the same kind of analysis that is used to evaluate a proposed merger. Similarly, a Big Tech-AI alliance may cause higher prices through a different form of collusion—the exchange of current non-public price information.
The fourth anticompetitive possibility is vertical. Unlike the horizontal theories of harm described above, this theory posits that a Big Tech firm uses its alliance with an AI startup to induce the startup to refuse to deal with, or raise the costs of, a competitor of the Big Tech firm. By disadvantaging a competitor, the Big Tech firm gains the market power to increase its prices or otherwise exploit its customers. Suppose, for example, that Amazon uses its partnership with Anthropic to induce the startup to refuse to provide its best AI product to an emerging competitor of Amazon in the cloud services markets. If Anthropic’s product is superior to any other product available to the nascent competitor, its ability to grow and challenge Amazon will be reduced, depriving consumers of an alternative to Amazon and possibly raising prices or diminishing quality in the cloud services market.
The fifth anticompetitive possibility is related to the fourth. Not only may the strategy of raising a rival’s cost enable a Big Tech firm to entrench its existing market power, but it may also allow the firm to extend its market power into another market. Suppose that Meta decides to enter the cloud streaming gaming market and partners with or outright acquires an AI startup that is developing a potentially critical input to cloud streaming gaming. If the input does prove crucial, Meta could gain market power in the gaming market by preventing the sale of that input to competitors in that market except at discriminatory prices. That could raise the prices of cloud streaming games and choke off innovation.
The final threat relates to the labor side of the market. Alliances between a Big Tech firm and an AI startup may give the two firms monopsony power over the highly skilled and specialized workers they employ. If these workers have few other options, the collaboration, by increasing concentration in the labor market, may enable the firms to reduce wages or degrade working conditions.
In short, the antitrust agencies have ample reason to scrutinize Big Tech-AI startup alliances, even if material harm to competition is not yet demonstrable. While the impact of a particular alliance will depend on its facts, this reviews indicates there are substantial grounds for concern. Lemley and Wansley agree:
The tech giants are old. Each was founded more than 20 years ago—Apple and Microsoft in the 1970s, Amazon and Google in the 1990s, and Facebook in 2004. Why has no new competitor emerged to disrupt the market? . . .
The tech giants have learned to stop the cycle of disruption. They invest in start-ups developing disruptive technologies, which gives them intelligence about competitive threats and the ability to influence the start-ups’ direction.
Author Disclosure: the author reports no conflicts of interest. You can read our disclosure policy here.
Articles represent the opinions of their writers, not necessarily those of the University of Chicago, the Booth School of Business, or its faculty.
