A new study has found that artificial intelligence stocks had a 97% probability of being in a bubble as of September 2026, according to its strongest model. However, two other versions put the probability closer to 40%. The findings come from researchers who analyzed nearly 800,000 financial news articles going back 170 years. Importantly, the 97% figure measures how closely today’s market resembles past bubbles rather than the likelihood of an imminent stock market crash.
According to the Financial Times, researchers Marco Grotteria, Afonso Monteiro, and Constantine Yannelis examined 795,964 articles from The Economist published between 1843 and 2013. They found that financial news articles historically suggested there was a 20% chance of a major market crash within the next six months. However, historical data showed that crashes actually happened only 5.7% of the time.
This means that the warnings overstated the actual risk by roughly three and a half times.
However, the researchers also found that changes in the intensity of those warnings contained useful information. Their model compared news coverage with 22 historical market bubbles and tracked signs of excessive speculation and concerns about expensive stock valuations.
When tested, the model correctly ranked a randomly selected bubble month above a normal month 74% of the time. That figure increased to 94% for the dot-com bubble.
Why the Latest AI Bubble Warning Stands Out
The latest reading is particularly high compared to previous market crises. The same model recorded a 92% probability near the peak of the dot-com bubble in 2000 and 97% during the 2007 financial crisis.
There are also financial reasons why AI valuations are receiving attention. The Federal Reserve recently reported that debt issuance related to advanced computing had reached $680 billion by October 2026, roughly five times the amount issued during all of 2023.
That matters because companies must continue making debt payments even if their AI investments generate less revenue than expected. Meanwhile, a recent Federal Reserve survey found that 87% of respondents considered AI a financial stability risk, up from 50% in May.
However, the researchers pointed out some limitations in their study. For example, their historical analysis relied on one publication, while the AI models used to examine old articles might have recognized events from their training data. They also pointed out that technologies such as the internet created genuine economic value despite experiencing major investment bubbles.
Is SPY ETF a Buy, Sell, or Hold?
Turning to Wall Street, TipRanks data shows that analysts have a Moderate Buy consensus rating on State Street SPDR S&P 500 ETF Trust (SPY), which tracks the overall stock market in the U.S. This is based on 416 Buys, 81 Holds, and six Sells assigned in the past three months. Furthermore, the average SPY price target of $913.08 per share implies 25.2% upside potential. (See SPY Stock Forecast).



