The US Federal Trade Commission is closely watching "clouds on the horizon" of the competitive landscape for AI large language models, despite strong existing competition among key players like OpenAI, Google, Meta, Anthropic, xAI and DeepSeek. Commissioner Melissa Holyoak warned that market conditions — high fixed costs, economies of scale and network effects — may eventually drive global consolidation, reducing competition. Such consolidation could raise national security and data privacy concerns if a foreign adversary-subsidized AI firm gained dominance.
Competition is healthy in the market to create AI large language models, but the conditions exist for that market to consolidate — and the US Federal Trade Commission will be “vigilant” for early signs of that change, Federal Trade Commissioner Melissa Holyoak said today.Noting that OpenAI, Google, Meta Platforms, Anthropic, xAI and Deep Seek are all competing to build more powerful, cheaper large language models, Holyoak said the current state of competition is sunny. But she warned that there are “strong indicators based on both economic theory and the long history of industrial organization” that the conditions could emerge for the global — not just national — LLM market to be consolidated into the hands of a few competitors.
“The field right now is a competition enthusiast's paradise, right? Tons of competition,” Holyoak said, speaking at an AI policy summit* in Washington, DC. “But if you look carefully, you may see gray clouds forming on the horizon.”
There is evidence that AI models have similar characteristics to multi-sided platforms that could consolidate in the hands of a few competitors through network effects, Holyoak said.
“Foundation models are similar to multi-sided platforms that can exhibit indirect network effects,” she said. “As they improve through access to large data sets and attract more developer integration, they become more attractive to enterprise users, customers and potentially advertisers. In industrial organization economics, when a market is characterized by high fixed costs, barriers to entry, economies of scale and network effects, that market is likely to consolidate around a small number of competing firms.”
That is not the case right now, Holyoak said, when investors are injecting billions of dollars of capital in multiple competitors. But if that flow of investment starts to slow, “there will be increased pressure in the market to consolidate,” she said.
Moreover, because an AI LLM can instantly move between prompts and outputs in English, French, Arabic, Chinese or any other language, that market consolidation could produce global dominance, not just national market power, Holyoak said, underscoring the importance of antitrust scrutiny.
There could be national security and data privacy implications for that global competition. The commissioner did not name China and DeepSeek in that context, but she warned about the national security danger of an “AI firm subsidized by a foreign adversary, an adversary that offers deeply below-cost pricing for a future AI product” to gain market power.
“This could be good for consumers to benefit from low prices, but from a national security perspective, this could cause significant concerns,” Holyoak said. “The foreign firm could use its low prices to attract American consumers and gain access to their personal data and other potentially sensitive or actionable information. Such an outcome would justify careful scrutiny across the federal government.”
Repeating a foundational view within the Trump administration that the US antitrust agencies — the FTC and the US Department of Justice — “are antitrust enforcers, not regulators,” Holyoak said the US would not protect US AI companies from foreign competition or otherwise try to shape the development of the AI market or its products.
“If we try to impose our beliefs about how the AI market should look, about how many competitors there should be, or how much they should charge for the products, or what features the products should have, we are more likely to dampen innovation and allow foreign firms to win the market than we are to actually get the desired outcome,” she said.
“Gray clouds on the horizon do not guarantee a storm, but it does suggest that we pack the umbrella,” Holyoak said. As a US antitrust enforcer, “that means we have to remain vigilant about the accumulation and abuse of market power, even if the market is competitive, like today.”
*ForumGlobal, “The USA AI Summit 25,” US Leadership in AI: The Emerging Global Order.” Washington, DC. June 3, 2025.
Please e-mail editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.