A claim has surfaced that the huge investment burden borne by U.S. frontier AI companies such as OpenAI and Anthropic could become a risk factor as low-cost Chinese AI models catch up. The analysis says it could be difficult to sustain today’s high corporate valuations if the price gap for high-performance models narrows.

On Sept. 25 (local time), Business Insider reported that Scott Wilson, chief investment officer for Washington University in St. Louis’ endowment, voiced a skeptical view at an investment event in Fort Worth, Texas. "OpenAI and Anthropic are highly valued relative to the obligations they have taken on," he said. He also argued that profitability at frontier companies that continue massive infrastructure spending could come under pressure if free or low-cost AI models rapidly increase.

Wilson pointed to Chinese companies such as DeepSeek, Alibaba Qwen, Zhipu AI and Tencent, saying they are narrowing the performance gap while emphasizing lower costs. He also said that talks with founders of companies being considered for investment showed some were moving to cheaper open-weight models. On the model brokerage platform OpenRouter, requests for DeepSeek’s text model accounted for 25.3 percent, topping OpenAI’s 18.6 percent and Anthropic’s 2.9 percent. This is a limited indicator that shows only OpenRouter usage.

Anthropic is preparing for an initial public offering, and Reuters reported that the listing process could begin around mid-October. Wilson said he had reviewed an opportunity to invest in OpenAI early on but did not participate because he lacked confidence in long-term profitability.

There is also a counterargument. Vinod Khosla (비노드 코슬라), an early investor in OpenAI, said the core of competitiveness lies less in the model itself than in how much a company controls the overall infrastructure, including chips, data centers and software. OpenAI has jointly developed its own inference chip, Jalapeño, with Broadcom and is pursuing a strategy to reduce reliance on Nvidia and external clouds and cut inference costs.

Which strategy will gain a long-term edge, between the spread of low-cost open models and building large-scale in-house infrastructure, is emerging as a key variable for the AI industry.