The much-discussed AI stock bubble into late 2025 has already burst. That’s the conclusion of John Higgins, chief market economist at Capital Economics. He is more worried about what is still happening.
A bubble is when an asset is valued far above its true value, usually seen when stock prices soar despite solid evidence of strong performance. “If you’re determining whether a bubble exists in relation to things like valuation growth, then there’s an argument that the bubble has burst,” Higgins said. luck.
In a note to clients this week, Higgins said information technology and other Big Tech companies have seen their current price-to-earnings-per-share ratios rise in recent years, leading to inflated valuations. However, as of October 2025, its price-to-earnings ratio has declined to its lowest since the pandemic. Higgins pointed out that the dot-com bubble at the beginning of this century followed much the same pattern, but the price-to-earnings ratio was much higher, exceeding 150% in the IT sector in the early 2000s, compared to a peak of nearly 75% in late 2024.
The reputation of AI is certainly on the rise. As of fall 2025, there were 498 AI unicorns with a total valuation of $2.7 trillion, 100 of which were founded after 2023, according to data from technology market intelligence platform CB Insights. More than 1,300 AI startups are valued at more than $100 million. OpenAI’s valuation hit $730 billion last month, up from $500 billion less than six months ago in October, according to CFO Sarah Friar.
But the technology sector is coming back to reality, partly as a result of the “SaaSpocalypse,” a sharp decline in software-as-a-service (SaaS) stocks, as investors worry that agent AI could easily replace traditional software business models. Salesforce and ServiceNow have both lost about 30% of their value since the beginning of the year.
“Investors have taken note to some extent that the software services industry group is one of the sectors that is relatively vulnerable to the deployment of AI,” Higgins said. “Therefore, we have significantly reduced valuations, especially for that sector.”
Higgins argues that it’s not just the SaaS industry that is hurting. The semiconductor industry has also slowed recently, with high demand fueling chip shortages and recent geopolitical tensions such as tariffs and the Iran war creating supply chain challenges.
AI’s next bubble will be rare
Higgins says there may be another bubble hidden within these stories of industry failure. Tech companies’ profits have soared in recent years, raising questions about how sustainable this level of growth is. Bloomberg Intelligence estimates that the Magnificent Seven’s earnings growth is about 18%, compared to 11% for the remaining 493 S&P 500 companies. Last month, NVIDIA reported fourth-quarter revenue of $68.1 billion, an increase of 73% year-over-year.
“There may be one [bubble] “In fact, it’s very rare that there’s a problem with the fundamentals of things,” Higgins said. “Usually we think of bubbles as prices moving away from the fundamentals themselves… In this case, the bubble may actually be in the earnings itself.” By doing so, Mr. Higgins was referring to a key issue focused on by tech advocates in the anti-bubble camp: the huge profits generated by the largest publicly traded tech companies that control the Magnificent Seven. In other words, he is asking what happens if these profits decline.
There are several reasons why AI revenues could soon reach a cliff and enter a market correction. As an example, Higgins said demand for AI may be much lower than initially expected, and tech companies need to consider that AI capital spending will reach an estimated $539 billion in 2026, according to Goldman Sachs. According to McKinsey, 88% of companies report using AI on a regular basis, but adoption may be stalled as a result of employee fears about technology replacing their jobs.
Higgins suggested that the risks to AI’s returns would be greater if the economy remained unstable. The ongoing war with Iran has halted helium production in Qatar, which is responsible for about a third of the world’s supply of the odorless gas used to make computer chips. Not only are data centers targeted during conflicts, but energy prices can also drive up input costs for these facilities.
“If the economy weakens more generally, that could weigh on stock markets and the earnings of companies that profit from deploying AI, even if demand for AI itself hasn’t actually weakened that much,” Higgins said.
