Economists warn: an AI bubble has burst

Ask AI · After the AI bubble bursts, can profit growth continue to support high valuations?

Nvidia’s fourth-quarter revenue rose 73% year over year, but one economist says this AI-driven surge in performance may be difficult to sustain. Photo source: David Paul Morris/Bloomberg—Getty Images

Debate over whether there is a bubble in AI stocks has been heating up through the second half of 2025, and now, this bubble has actually burst. That is the conclusion reached by John Higgins, chief market economist at Capital Economics. However, he is more concerned that another bubble is in the making.

The so-called “bubble” typically refers to situations where asset valuations far exceed their intrinsic value, and it often shows up like this: even when companies’ financial performance is strong, their stock prices are still being driven higher and higher. Higgins told Fortune: “If you use overextended valuations as the basis for deciding whether there is a bubble, then you could say this bubble has already burst.”

In a client report published this week, Higgins noted that over the past few years, the price-earnings ratio of the information technology sector and the entire “Big Tech” segment (the ratio of stock price to earnings per share) has continued to rise, signaling signs of inflated valuations. But starting around October 2025, the indicator began to fall, and it is now down to the lowest level since the pandemic. He also mentioned that the internet bubble around the turn of the century followed a similar trajectory; however, at that time, the price-earnings ratio levels were much higher—at the start of the 21st century, the IT sector’s price-earnings ratio exceeded 150%, and the peak for this cycle appeared at the end of 2024, close to 75%.

Valuations in the AI space have been soaring all the way up. According to data from tech market intelligence platform CB Insights, as of the fall of 2025, there are 498 global AI unicorn companies with total valuations of $2.7 trillion, including 100 founded in 2023 or later. In addition, more than 1,300 AI startups have valuations exceeding $100 million. OpenAI CFO Sarah Friar said the company’s valuation reached $27k last month, while less than six months earlier (October 2025), that figure was $730B.

However, the technology sector as a whole has already begun to “return to reality.” One reason is the so-called “SaaS endgame”—because investors worry that “agentic AI” could easily replace traditional software business models, software-as-a-service (SaaS) stocks have been hit by rapid selling. Since the beginning of this year, the market values of Salesforce and ServiceNow have both shrunk by about 30%.

Higgins said: “Investors have realized that the software services industry is one of the relatively fragile segments under the AI wave, so we’re seeing a significant valuation retracement in that industry.”

Higgins also pointed out that it’s not only the SaaS industry that has been hit. The semiconductor industry has also shown signs of slowing recently: on the one hand, surging demand has driven a chip shortage; on the other hand, tariffs and geopolitical tensions such as the Iran war have also disrupted supply chains.

_****

AI will face a rare bubble

****_

Higgins believes there may be another bubble hidden behind the challenges these industries are facing. Over the past few years, tech companies’ profits have surged dramatically, raising a key question: how long can this growth last? Bloomberg Intelligence expects the earnings growth rate of the “Magnificent Seven” to be about 18%, while the other 493 companies in the S&P 500 are growing at only 11%. Last month, Nvidia (Nvidia) reported fourth-quarter revenue of $68.1 billion, up 73% year over year.

Higgins said: “Actually, it’s possible that a bubble is appearing in the fundamentals themselves—this is quite rare. Usually, the bubble we talk about is when prices decouple from fundamentals… but this time, the bubble may be in earnings themselves.” Higgins’ remark directly targets the core argument of “anti-bubble” tech supporters: that the large publicly listed tech companies dominating the “Magnificent Seven” are generating astonishing profits. In other words, the question Higgins is asking is: what happens if these companies’ profits start to decline?

AI earnings may face a “cliff” soon and trigger a market pullback, driven by multiple factors. First, Higgins said AI demand may be far lower than initially expected, making it difficult for tech companies to digest the heavy investment costs—according to estimates by Goldman Sachs, tech companies’ AI capital expenditures in 2026 will reach $500B. McKinsey data shows that although 88% of companies say they are normalizing the use of AI, the actual rollout may stall because employees worry about being replaced by technology.

Higgins believes the bigger risk facing AI earnings is that the economy remains in an unstable state. The ongoing Iran war has already led to Qatar’s shutdown of helium production, and Qatar supplies about one-third of the world’s helium—this colorless, odorless gas is an important raw material for manufacturing computer chips. The conflict not only turns data centers into attack targets; rising energy prices also raise the costs of investing in these facilities.

Higgins said: “If the overall economy weakens, even if AI demand itself doesn’t clearly decline, it will still drag down stock market performance and weaken the earnings of companies that rely on profiting from AI deployment.” (Fortune Chinese website)

Translator: Liu Jinlong

Reviewed by: Wang Hao