The stock market might have flashed an ominous warning on August 29 that the AI bubble is starting to deflate. While most pundits talk of “bursting” bubbles that cataclysmically explode into thin air, the most damaging nature of a bubble is to relentlessly lose air until it is all gone. Bubbles don’t suddenly burst and then go away. They burst, then they take years to deflate. The biotech and dotcom bubbles exhibited this behavior. And the bursting of the AI bubble is likely to do the same.

By the early 1980s, biotech was seen as a new, investable asset class. Breakthroughs in gene splicing, monoclonal antibodies, and the promise of the Human Genome Project fueled repeated speculative waves in 1980, 1984, and 1987. Investors poured capital into startups with little more than promising technology, betting that science would quickly translate into marketable products.

Despite the hype, the industry was unproven. By 1991, only about 15 FDA-approved biotech products existed, and most companies were years from generating revenue. The largest surge came in 1991–1992 after Amgen’s patent acquisition and FDA approval of Neupogen, which convinced many investors that biotech had become viable. The bubble burst in 1992, and the fallout included sharp declines in biotech stock values, a drying up of financing, and a prolonged downturn that lasted into the late 1990s.

The dotcom bubble was not dissimilar. In early 2000, the Fed began hiking interest rates, making borrowing more expensive and triggering a sell-off in overvalued tech stocks. Market psychology changed, and the rampant speculation turned into panic selling. Moreover, dotcom companies had excessive valuations that were unjustified given their lack of cash flow.

The Nasdaq reached its all-time peak of 5,048 in March 2000 and then collapsed for two years, hitting a low of 1,139, which represented a 78 percent loss of value. The crash led to reduced investment, layoffs in the tech sector, and a broader slowdown in the U.S. economy, though it did not trigger a full-scale recession.  The NASDAQ would take 15 years to recover to its March 2000 peak, finally reaching it in April 2015. 

Overall, both the dotcom and biotech bubbles are cautionary tales of overvaluation, investor hype, and market psychology, demonstrating how rapid technological enthusiasm can lead to dramatic financial consequences. They featured higher interest rates, excessive valuations, unproven business models, and unbridled speculation. While there is no way to truly predict the bursting of a bubble, those conditions mirror what is happening today in the AI arena. Moreover, those markets featured prolonged selloffs with FALLING volatility. After the initial crash, the VIX would have actually dropped during the majority of the time those bubbles were bursting. Very different from a 2008 crash or a March 2020 breakdown.

Last week, VIX futures prices fell at the same time stocks were falling. That occurs when markets have an orderly repricing of value downwards. Investors appeared to be balking at market prices, but in an orderly fashion. No crash had occurred, but a stock market selloff amid dropping volatility is a warning sign. It says that rather than being surprised by a selloff, investors expect more of the same. In this case they buy if prices fall; otherwise, they do nothing. Sellers, meanwhile, would be beyond panic. They would be resigned to lower prices, just glad to be lightening their risk. Hence the fall in the VIX. The warning was brief but serious. Just days later, Fed Governor Waller expressed a more dovish path for the Fed, and hence a better environment for stocks. This is not a coincidence.