Commentary: Why AI Stocks Are Defying Gravity — And What Could Bring Them Down

Since the release of ChatGPT in late 2022, the stock prices of leading U.S. artificial intelligence companies—known collectively as the “Magnificent Seven”—have significantly outpaced the broader market. Following DeepSeek’s emergence in early 2025, top Chinese AI firms, mainly listed in Hong Kong, also experienced notable gains. In the U.S., despite rapid corporate earnings growth, risk premiums remain low, showing persistent investor optimism. This situation has sparked ongoing debate about whether AI asset prices are entering bubble territory. Rather than focussing on the technical definition of a bubble, the article analyzes how asset prices, innovation, and macroeconomic conditions interrelate[para. 1].

One way to rationalize high stock valuations is through falling interest rates, fueling speculation about possible U.S. Federal Reserve rate cuts. Historically, an inverse relationship exists between interest rates and risk asset prices, with rate cuts typically boosting stocks. However, current stock surges have occurred despite rising U.S. interest rates. Three possible dynamics are described: the classic scenario (rates affect stocks), stocks affecting rates (the AI-driven rally fuels demand and inflation, causing the Fed to keep rates high), and a third dynamic where both rates and stocks are influenced by global capital inflows. As of September 2025, foreign investors held $21.2 trillion in U.S. stocks (31.3% of the market), the highest level since WWII. Notably, AI-related capital expenditure has contributed about a third of annual U.S. GDP growth, and the wealthiest 10%—owning 85% of equities—now account for half of all U.S. consumption. High valuations can coexist with high rates, but if AI enthusiasm fades, reduced capital flows and spending could lower the natural interest rate, signaling market pessimism. Retail investors, empowered by digital platforms and generative AI, could amplify both gains and losses[para. 2][para. 3][para. 4][para. 5].

Another absorption mechanism for valuations is corporate earnings growth. Despite recent gains concentrating in a few mega-cap companies, there is optimism that AI, as a general-purpose technology, will eventually benefit all sectors. Investors must weigh the cost-benefit of AI implementation, given the high R&D and operational expenses inherent in large models. The sector has shifted from capital-light to capital-intensive, with tech giants replacing venture capital in AI startup funding. Commercial utility still lags, as many corporate clients are not yet equipped to use large language models effectively. Studies diverge on AI’s long-term economic impact: extrapolations forecast an annual U.S. GDP growth boost of 0.8-1.3 points, while task-based models predict as little as 0.07 points. China’s AI impact is projected at 9.8% GDP growth by 2035, or about 0.8% per year. Bridging the gap between high current costs and uncertain future profits remains a key challenge, fueling worries of overheating[para. 6][para. 7][para. 8][para. 9][para. 10][para. 11].

Recent advances, such as DeepSeek’s algorithmic innovations, highlight the role of economies of scale in chipmaking. While increased efficiency can fuel greater demand—a concept known as the Jevons Paradox—chips benefit from falling unit costs with scale, unlike natural resources such as coal. Major chipmakers, holding pricing power, can temporarily earn monopoly profits, but long-term trends point toward declining costs and prices. Open-source AI models are also shifting industry dynamics by lowering barriers and fostering global innovation. Energy consumption poses a constraint, with green energy offering scale-driven cost benefits, particularly relevant to large-scale AI adoption[para. 12][para. 13][para. 14][para. 15][para. 16][para. 17][para. 18][para. 19].

In conclusion, high AI stock valuations are supported by expectations of future earnings and current capital spending, both of which may prove unsustainable. Valuations could fall if chip scale economies weaken or if AI applications underperform. However, unlike real estate bubbles, technology bubbles, while painful when they burst, often accelerate innovation and can have positive long-term economic impacts through scale and spillover effects, whereas real estate collapses tend to bring broader systemic risks[para. 20][para. 21][para. 22].