AI Bubble? No, But Some Tech Stocks Show Signs

Like past transformative technologies, from railways and radio to the internet, artificial intelligence has attracted optimistic investors and inflated valuations, and with them speculation of a bubble market - when eager investors drive up asset prices beyond their underlying values.

The AI sector is not in a bubble market, but the prices of some individual AI companies - including Alphabet, Google's parent company - behave with the price variability and exuberance characteristic of a bubble, according to new Cornell research.

"We hear the term 'bubble' used frequently, but the more important question is whether we can identify and quantify the underlying characteristics associated with one," said Martin Wells, the Charles A. Alexander Professor of Statistical Sciences in the Cornell Ann S. Bowers College of Computing and Information Science and in the ILR School. "Our analysis suggests that some segments of AI-related companies exhibit characteristics consistent with speculative bubble dynamics, while others do not. Treating 'AI' as a single category therefore obscures important heterogeneity across firms and sectors."

Wells and Abir Sarkar, a doctoral student in the field of statistics, are the authors of "Is There an AI Bubble? Robust Data-stamping for Periods of Exuberance," which was published in July in the American Institute of Mathematical Science's Frontiers in Mathematical Finance.

According to the pair's analysis, Alphabet currently shows strong evidence of overvaluation. Its stock price has increased more than 70% in the last year, outpacing the entire NASDAQ Composite, which has grown by around 20%.

In the paper, researchers developed and used a new statistical method - a Stochastic Volatility-robust Augmented Dickey-Fuller (SV-ADF) framework - that examines daily stock prices to pinpoint when bubbles originate and collapse, and to drill down beyond volatile price swings to identify individual companies that exhibit genuine bubble-like dynamics or exuberance. Existing models lack this specificity and too often label entire sectors as a bubble when there isn't one, the researchers said.

"Our method is not saying that everything is a bubble. That is the key distinction from the standard bubble-detection methods," said Sarkar, the paper's lead author. "Our method is robust enough to identify transient volatility spikes and is sufficiently selective to distinguish bubble episodes at the individual-stock level, even within closely related industry groups."

Researchers analyzed AI-exposed stocks from 2020 until April 2026, including the "Magnificent Seven" - big U.S. tech companies like Apple, Amazon, Nvidia and Meta - semiconductor and AI infrastructure companies such as TSMC and Broadcom, and cryptocurrency.

They found that nearly all semiconductor companies were in a bubble following ChatGPT's release in November 2022. The model found Tesla, too, showed signs of overvaluation in 2020. In their analysis of cryptocurrency, researchers found exuberance with Bitcoin and Ethereum starting in December 2020. All these bubbles have since collapsed, although new exuberance eventually emerged in a few cases, researchers found.

Wells and Sarkar intend to expand this research with Robert Jarrow, the Ronald P. and Susan E. Lynch Professor of Investment Management in the Cornell SC Johnson College of Business and a leading researcher in asset-price bubble estimation theory.

"Identifying stocks with bubbles is important because bubbles eventually burst, which result in dramatic price declines," Jarrow said. "Knowing which stocks contain bubbles enables investors to make better investment decisions."

Louis DiPietro is a writer for the Cornell Ann S. Bowers College of Computing and Information Science.