Chairman Fitzgerald, Ranking Member Nadler, Chairman Jordan, Ranking Member Raskin, and esteemed members of the Subcommittee, I’m honored to be here today. My name is Joseph Coniglio and I am the Director of Antitrust and Innovation Policy and the Schumpeter Project for Competition Policy at the Information Technology and Innovation Foundation. I’ve worked on antitrust issues in high-tech markets for over a decade, first in government, then as an attorney in private practice, and now in public policy. I have written numerous pieces and spoken widely on the application of antitrust law to high-tech industries as well as the need for policies which foster the innovation that drives economic growth and U.S. techno-economic leadership.

The Antitrust-AI Interface

There are three key areas where AI and antitrust overlap: the AI industry itself, AI’s broader impact on competition throughout the American economy, and the effects of AI on the ability of the antitrust laws to police anticompetitive conduct.

First, with respect to the AI industry itself, competition exists at all key levels of the AI stack—foundation models, cloud computing, and chips. At the model level, entrants like OpenAI, Anthropic, Cohere, and Mistral thrive even without the troves of proprietary data enjoyed by the large digital incumbents like Google, Meta and X with whom they compete. At the resource-intensive cloud computing level, robust competition nonetheless exists not just between Amazon, Google, and Microsoft, but also IBM, Oracle and other players. And, while Nvidia’s revolutionary AI chips have unsurprisingly given the company an edge that will enable it to recoup its investments, firms like Amazon and AMD are poised to compete.

Some have expressed concerns about anticompetitive conduct in the AI space, and in particular partnerships between legacy big tech firms and innovative AI startups. However, worries that these arrangements are substantially lessening competition not only overlook the competition that exists at both the foundation model and cloud levels, but the procompetitive benefits that these arrangements can provide in accelerating AI innovation. That is, even if it is true—and a good thing—that investment is abundant in AI, partnerships like those between Microsoft and OpenAI can be critical to provide AI startups with patient capital that is unconstrained by a more short-term outlook and crucial to sustainable growth.

Rather than recreating a “Wintel-duopoly” scenario where a few market leaders use anticompetitive tactics to stifle innovation, the AI space is witnessing huge investment and expansion that belies any market failure. As evidenced by recent developments involving DeepSeek, Coreweave and Google’s next-generation Willow chip, the AI industry is dynamic and evolving consistent with the “Schumpeterian” competition that typifies high-tech markets: Firms like Nvidia reinvented themselves with tremendous success in chips, new players like OpenAI and Anthropic are flourishing at the model level, and digital leaders—rather than resting on their laurels—are using their scale to push the cloud revolution forward and compete vigorously in other areas of the AI stack.

Second, AI is already being heralded as the next general-purpose technology that, like the Internet, will transform our economy. While it is far too soon to speculate about how the AI revolution will generally affect competition in the American economy, the effects of the Internet revolution provide a helpful backdrop for analyzing what the impact of AI may be. For example, although some have expressed fears about increased concentration in the economy over the past two decades, these concerns are often overstated. Indeed, ITIF has found that concentration has not meaningfully increased and only 4 percent of U.S. industries are highly concentrated. In fact, the digital revolution may have even enhanced competition at the local level, such as through online e-commerce platforms that allow consumers to transact with millions of sellers around the world.

Moreover, even where concentration or markups may have increased over the past several decades, that does not mean competition isn’t working. On the contrary, studies continue to find that higher concentration and markups are not fueled by price increases, but instead cost reductions driven by the sort of efficiencies that bring lower prices for consumers—and which another AI wave of general-purpose technological innovation would continue to empower. In addition, from the standpoint of promoting innovation, concentration has long been recognized as often beneficial: Firms with established positions can have a greater incentive and ability to engage in innovation than firms that face high levels of product market competition. This has been confirmed by studies showing that the relationship between concentration and innovation may regularly take the form of an “inverted-U,” whereby industries that move from many competitors to a few big firms can see greater innovation, even if monopoly levels of concentration remain undesirable.

At bottom, despite the oft-heard refrain of “digital market failure” that motivated the antitrust policies of the Biden administration as well as policymakers around the world who are considering—and increasingly adopting—digital antitrust regulations like the European Union’s Digital Markets Act, the reality is that the Internet has brought tremendous benefits to the U.S. economy. In fact, the AI revolution itself is a testament to the continued flourishing of high-tech competition and the importance of an antitrust policy that fosters innovation. Over the past forty years the U.S. pursued deregulation and a consumer welfare antitrust policy conducive to the innovation and growth that helped to preserve its share of global wealth despite China’s rise. By contrast, Europe continued with an antitrust framework more suited to protecting competitors, missed the digital revolution, and saw its share of global wealth fall by about half.

Third, there is the question of how the AI revolution will affect the antitrust enterprise—namely, the prevalence of, and the ability of the antitrust laws to effectively police, anticompetitive conduct. For example, when it comes to combatting collusion, some have argued that AI algorithms may make it easier for firms to form and maintain cartel agreements, raising the specter of increased collusion throughout the economy. While the question of whether AI will lead to more cartelization is worthy of study, it is far too early to suggest that the antitrust laws are not up to the task. Not only has the Department of Justice already brought an AI-related algorithmic collusion case using its existing tools—which in the future could even be enhanced by AI technologies that make it easier to detect cartels—but AI may also increase firms’ incentives to disrupt cartel behavior, resulting in less collusion throughout the economy as a general matter.

In addition to cartels, there is a related worry that AI may facilitate oligopoly outcomes without firms having to enter into any anticompetitive agreement. Because such “tacit collusion” is lawful, AI may allow firms to avoid antitrust scrutiny despite nonetheless harming consumers through oligopoly behavior. Here again, although the possibility of greater tacit collusion through AI certainly merits monitoring, it is unlikely to justify the inclusion of exploitative offenses into antitrust law. Rather, not only is it a fundamental premise of antitrust law that high prices are an important element of a free-market system, but the “plus factor” analysis that courts consider for inferring an agreement should prove sufficiently flexible to adapt—as antitrust law has always done—to new economic realities. What’s more, the Federal Trade Commission already has the established authority to condemn unilateral facilitating practices under Section 5 of the FTC Act.

Finally, similar issues have been raised about AI undermining effective antitrust enforcement of predatory conduct by a monopolist, which is often analyzed by asking whether the monopolist’s behavior made “economic sense.” For example, some have suggested that AI will both reduce variable costs to the degree that below-cost pricing or profit-sacrifice becomes practically non-existent, as well as provide firms with analytical tools that ensure their behavior makes economic sense but for any anticompetitive effects. However, such concerns aren’t just speculative, but they beg the question that the existing tests for evaluating predatory behavior shouldn’t apply: If AI is making below-cost pricing and other behavior that does not make economic sense rarer, anxieties about the antitrust laws failing to condemn anticompetitive predatory behavior should diminish—not suggest that greater enforcement is needed.

General Frameworks for Policymakers

In the face of these complex issues involving the intersection of antitrust and AI, there are at least four general frameworks policymakers can adopt.

First is the view that in the face of a new general-purpose technological wave like AI, vigorous antitrust enforcement is more important than ever before. A common justification for this approach is that AI will create powerful network effects and increasing returns to scale, which will lead to a winner-take-all economy where first movers enjoy unassailable market positions that they can use to exploit consumers. Similarly, high increasing returns to scope may also have concentrating effects—especially with the advent of even more advanced AI technologies—that further entrench dominant firms that are vertically integrated.

Although it is premature to predict the economy-wide effects of AI on concentration, there are many reasons to believe that AI will create competition—not destroy it. For one thing, the cost-reducing potential of AI may generally lower entry barriers throughout the economy in a way that empowers small- and medium-sized firms to better compete, much the same way that cloud computing made it easier for smaller firms to achieve scale. Moreover, even if AI does increase concentration, to the extent that this is driven by efficiencies that lower prices for consumers, it would only further—not undercut—the consumer welfare goals of U.S. antitrust law.

Second, there is the contrary view that because AI is so disruptive and efficiency-enhancing, there is little need to enforce the antitrust laws at all: Market forces, not antitrust intervention, can be trusted to remedy whatever market failures may persist. Indeed, a similar debate was had during the Internet revolution and noted in the seminal Microsoft case, where the court discussed how, in technologically dynamic markets, monopoly power may be only temporary and better remedied by Schumpeterian competition than time-consuming antitrust lawsuits.

While, like the Internet, AI will surely bring enormous benefits and disruption to the American economy, the past forty years have taught us to be wary of formalizing a priori theoretical assumptions of market self-correction into antitrust analysis. That is, even though markets often do correct their own failures, the need for antitrust action should be determined by market realities, which themselves can counsel strongly against intervention. Robust antitrust enforcement to promote consumer welfare and innovation should remain a policy focus—even in the age of AI.

Third, there is the notion that AI may undermine the very market competition upon which antitrust law is premised. Put another way, markets—and therefore antitrust law—may no longer become the dominant form of economic organization. This channels Marx’s view during the First Industrial Revolution that market capitalism embodied an economic contradiction that would ultimately bring about a socialist central planning state. Today, some commentators have analogously claimed that AI may render markets obsolete by enabling more sophisticated calculation techniques that make regulation a generally more effective mechanism for economic governance.

Besides being sweeping speculation, the central problem with this perspective is that it is based on a static understanding of how market competition works. That is, even if AI will no doubt improve our ability to do sophisticated economic analysis, a world of even more dynamic competitive processes driven by AI could at the same time make state central planning and regulation more difficult and far less necessary: Not only will the economy become even more fast-moving and complex, but the sort of persistent market failures which justify regulatory intervention could be more readily corrected by AI-enabled disruptions.

Fourth, rather than enforcing the antitrust laws too much, too little, or moving to a regulatory economic model, policymakers can continue to rely on an approach aimed at promoting consumer welfare and innovation to police anticompetitive behavior both in key AI markets, as well as the myriad industries that AI will touch—all while appreciating that AI is, and appears likely to remain, a highly dynamic space. In other words, as a common law statute that has evolved for over a century, the existing antitrust framework can be trusted as sufficiently well-founded and adaptable to help guide the American political economy through the AI revolution.

I think this is the right approach. Antitrust law’s fundamental focus on promoting competition and innovation by proscribing collusive and exclusionary anticompetitive conduct has, since the passage of the Sherman Act, withstood and helped manage the gales of creative destruction that helped make America the world leader that it is today. And, ensuring that competition policy remains tethered to these principles is essential for preserving American techno-economic leadership into the 21st century amidst the new wave of general-purpose technological change that is AI.

Thank you and I look forward to your questions.