• Insights, Thoughts and Insights

Competition in the AI Stack: A New Focus for Antitrust Enforcers? 

By Wyatt Fore

With news that OpenAI, the maker of ChatGPT, may soon be valued at $100 billion, artificial intelligence is now big business.1  As with any emerging technology, a handful of firms have grabbed the lead.  Is this a sign of market dynamism, allowing successful companies to earn outsized profits for their world-changing innovation?  Or does this raise concerns about whether competitive market conditions will shape the development of AI, hopefully for the betterment of society moving forward?  Difficult questions with uncertain answers abound.2

However, one thing remains certain: competition enforcers are taking a look.  From examining licensing strategies by chip manufacturers underpinning AI applications,3 to investigating investments by major technology firms into emerging AI companies,4 global enforcers are watching the space with a close eye. 

Winner Take All? 

Leading commentators of the artificial intelligence / machine learning industry have observed that many models, especially those that rely on deep or continuous learning, are characterized by strong feedback loops.  Such network effects are not new; they have characterized the digital age and have been noted by antitrust courts including the landmark Microsoft decision in the late 1990s.  This makes intuitive sense; the more users interact with an application, the more data it has to train the model to improve.  The better the product, the more users are attracted to the service, giving the application more data to improve yet again.  The cycle is self-reinforcing.  Scale thus can matter a lot, and data can be out of reach for all but a handful of firms.

 Economists have long observed that feedback loops, including in the presence of network effects, tend to create ‘winner take all’ markets.  This is why many networked industries, from telegraphs to railroads, have been called natural monopolies.  But not all network industries tend toward monopoly outcomes, and just because a market tends towards monopoly does not mean antitrust has no role to play.  In fact, just the reverse; some of antitrust law’s most foundational precedents (not the least of which, the iconic breakup of AT&T) involve networked industries in which a firm has abused its market power to exclude competition.  And more recently, a District Court found that Google Search was “an example of data network effects since each search query contributes to refining the Google Search algorithm.”5  Such a finding was critical to the conclusion that Google’s default arrangements and revenue share agreements had an exclusionary effect, by depriving its rivals of the ability to scale through access to user data. 

Such a finding, should it survive appeal, has major implications for artificial intelligence models and applications, which similarly rely upon external sources of data, including but extending beyond users, to refine their product.  Perhaps AI tools, especially deep learning models, do not need user data to improve, but rather can rely on other external sources.  But to the extent that AI models and applications deprive rivals of data, a critical “factor of production,”6 such arrangements are likely to catch the eye of antitrust enforcers. 

Strategic Partnerships or Problematic Investments? 

In addition to exclusionary conduct, antitrust enforcers have already begun to investigate investments by Big Tech into emerging AI companies.  As nearly every step of the AI supply chain (from microchips to cloud computing to the models and applications themselves) is already dominated by a handful of firms, such investments raise immediate antitrust questions.   

OpenAI, as the market leader, has already come under scrutiny.   For example, Microsoft is a major investor in OpenAI, and perhaps more importantly, has provided it with the vast computing resources needed to train generative AI systems like ChatGPT.7  With this background, Microsoft helped to resolve turmoil involving OpenAI’s board and the role of its founder, Sam Altman.  As part of the deal, Microsoft obtained a non-voting observer seat on the nonprofit board that controls OpenAI.  Antitrust commentators immediately raised concerns about this arrangement, which could allow Microsoft to observe (and perhaps steer) competitive decision-making.8  As a result, global enforcers, including the European Commission, the U.S. Federal Trade Commission, and the U.K. Competition and Markets Authority, began to investigate whether the arrangement violated anti-merger rules by effectively letting Microsoft gain a controlling influence over a rival and critical trade partner.  Microsoft ultimately abandoned its board seat.9   

For similar reasons, any deal or partnership involving an allegedly dominant firm and an important AI company is likely to raise eyebrows among competition enforcers.  Already, the Federal Trade Commission and the U.K. Competition & Markets Authority have launched broad inquiries into “strategic partnerships” between existing tech giants and AI start-ups, including not only Microsoft’s investment in OpenAI, but also two other tech giants’ partnerships with Anthropic, another AI model developer and rival to OpenAI.10  And of course given the national security implications of machine learning algorithms, such as the United States’ attempted forced divestiture of Tik Tok by Chinese firm ByteDance,11 such government unease about ownership is only likely to increase. 

Nobody knows how all these existing and future investigations will turn out.  But one thing is certain: global competition enforcers are keeping a close watch on how the industry develops. 

  1. Tom Dotan, Apple, Nvidia Are in Talks to Invest in OpenAI, Wall St. J., Aug. 29, 2024. ↩︎
  2. See generally ABA Antitrust Law Section Big Data Task Force, Competition Implications of Big Data and Artificial Intelligence / Machine Learning (2021).  ↩︎
  3. E.g., DOJ Launches antitrust probe of Nvidia, following complaints from rivals, report says, CBS News, Aug. 5, 2024; D. Michaels & A. Fitch, Nvidia draws antitrust scrutiny as enforcers signal early interest in AI, W.S.J., Sept. 9, 2024. . ↩︎
  4. Infra. ↩︎
  5. Order at 231, ECF 1033, United States et al. v. Google LLC, No. 20-cv-3010 (D.D. C. Aug. 5, 2024). ↩︎
  6. Lianna Brinded, Microsoft CEO Satya Nadella: We need to do to data what we did with electricity, Yahoo! Finance, Jan. 24, 2019. ↩︎
  7. Matt O’Brien, US antitrust enforcers will investigate leading AI companies Microsoft, Nvidia and OpenAI, Associated Press, June 6, 2024. ↩︎
  8. E.g., U.K. Competition & Markets Authority, Microsoft/OpenAI partnership merger inquiry (Dec. 8, 2023). ↩︎
  9. Ina Fried, Microsoft gives up observer seat on OpenAI board, Axios, July 10, 2024.   ↩︎
  10. David McCabe, Federal Trade Commission Launches Inquiry into A.I. Deals by Tech Giants, N.Y. Times, Jan. 25, 2024; Martin Coulter, Google parent’s partnership with AI startup Anthropic receives UK scrutiny, Reuters, July 30, 2024. ↩︎
  11. Aimee Picchi, After Biden signs TikTok ban into law, ByteDance says it won’t sell the social media service, CBS News (Apr. 26, 2024). ↩︎

Other Insights