Concerns are mounting over the sustainability of the rapid growth seen in the artificial intelligence (AI) sector, as companies in the United States and China compete fiercely for dominance. This competition, while driving innovation, also raises questions about the sector’s long-term financial viability and potential market repercussions.

In the United States, AI firms have attracted unprecedented levels of investment, buoyed by soaring valuations and strong revenue growth. Nvidia, the leading chipmaker powering AI applications, recently hit a market capitalization of approximately $5 trillion, a tenfold increase over four years. The company reported second-quarter sales of $96 billion, projecting a 70 percent increase over the coming year. Similarly, Anthropic, known for its Claude AI model, is preparing for an initial public offering this year with an expected valuation of around $2 trillion and revenues that have rapidly grown to approximately $65 billion since 2022. The sector is projected to invest about $7 trillion in data centers by 2030, a level of expenditure that will require substantial financing, likely through significant borrowing.

Despite these promising figures, some experts question whether the necessary revenue growth to justify such investment is attainable. David Cahn of Sequoia Capital highlighted the challenge of generating the annual revenues needed to cover the upfront costs of AI infrastructure, noting the figure has ballooned to an estimated $3 trillion—surpassing the total revenues of the entire U.S. technology industry. Lloyd Blankfein, former Goldman Sachs CEO, expressed concerns about an overly optimistic market appetite for AI-related revenue prospects, noting early signs of a slowing in customer spending. For instance, uptake of Anthropic’s latest Claude Fable 5 model has plateaued, reportedly due to the premium pricing.

Meanwhile, Chinese AI companies have adopted a contrasting strategy focused on affordability and accessibility. Chinese firms invest far less in data centers—less than one-tenth of their American counterparts’ spending—allowing them to offer AI models at significantly reduced prices. Models such as Moonshot AI’s Kimi K3, estimated to perform at 95 percent of the level of America’s Claude Fable 5 but at roughly one-third the cost, exemplify this approach. Chinese firms have also demonstrated rapid innovation cycles, narrowing performance gaps with U.S. technology within months. Chinese AI platforms tend to be more open, permitting hosting and customization on third-party servers, a model that contrasts with the more proprietary approach taken by many U.S. companies. Temporary restrictions by the U.S. government on providing the most advanced American AI models to foreign users have further boosted China’s market share, which now exceeds that of the U.S. in volume terms.

Other countries, including the UK, France, and Canada, have also developed AI offerings, often embracing open models to increase accessibility. This growing diversity contributes to increased competition and tends to push prices downward.

While the commoditization of AI services—as seen previously in software markets—may benefit consumers and businesses adopting AI, it poses challenges for investors expecting high returns on “frontier” models. If American AI firms cannot maintain premium pricing, their valuations, which constitute a substantial portion of major stock indices, could decline sharply. Given the current levels of corporate debt and the fragile state of U.S. public finances, a downturn in AI-related stocks might have broader implications for financial markets, potentially impacting investors globally.

Historical parallels are drawn with previous technological booms, such as the 19th-century railway expansions in the UK and the U.S., which, despite their transformative impact, were followed by severe financial crashes and widespread investor losses. Should AI follow a similar trajectory, the long-term benefits of the technology may still outweigh its economic disruptions, but caution remains warranted.