Have you wondered if the stock market is in an artificial intelligence (AI) bubble? You are not alone. The number of web searches for the term “AI bubble” hit, well, bubble proportions in August 2025, according to Google Trends.
Searches for “AI bubble” made searches for terms like “crypto bubble” seem like a blip. Having lived through several bubbles (the Japanese asset price bubble, the commodity bubble, the U.S. housing bubble, and even meme stocks and NFTs), I can say that the 1990s technology bubble was the most captivating. It is also the one most frequently compared to today’s AI bubble (if, in fact, it is a bubble).
From 1992 through 1999, the S&P 500 gained nearly 250 percent. If you are not sure, that is a lot. I mean, a real lot. The 2020s are on track to do something similar. The 1990s are known as much for their stock market gains as the subsequent bubble that ravaged investors. The index was cut in half, and the tech-heavy Nasdaq fell to within 80 percent of its previous level. I know you are sure of this—that is a lot. That is why it is essential to examine this topic thoroughly.
The stock market today is extremely expensive compared to historical metrics. Is it a bubble? What if I told you that the stock market is not in a bubble, but that a bubble is forming? Should investors get out of the stock market now to avoid the subsequent pop of the bubble? Is a popping imminent? Or should investors ride the wave and get out once the bubble begins to deflate? A look back at history may help us find some answers to those difficult questions and map out a game plan.
My personal comparison of then vs. now
I bought my first stock in 1992. In 1994, I got my first job in the investment industry—an intern at Smith Barney (which has long since been absorbed by a larger company). By 1998, I was working with a boutique investment firm that specialized in macroeconomic allocation and sector selection. Like many of us back then, I was day trading stocks on the side. Admittedly, day trading stocks is a statistically losing endeavor (at least 70 percent of day traders lose money, and only a tiny fraction might achieve profitability). Nonetheless, in the frenzy and hype of the late 1990s, it was hard to lose (and hard not to conflate a bull market with brilliance).
Following that party came the biggest U.S stock market hangover in three decades. Having lived through it, I have revisited that history enough to consider myself a true student of the era. And yet, as much as these first five years of the 2020s feel like the first half of the 1990s, I could not tell you what the next five years will mean for the stock market. But I have some educated guesses.
In the mid-’90s, you could feel the future humming. I remember watching Netscape’s green browser window crawl across a dial-up modem and thinking, “This is going to change how we live.” Specifically, I recall working in my home office and marveling at how I could type a URL into the browser, and by the time I went downstairs to grab a glass of milk and return to my desk, the website would already be up on my screen. How frustrating would that be today!
Today, the hum is still there, it is just the sound of GPUs (graphics processing units) and data centers instead of 56k modems. The question investors ask today is: Since the last five years rhyme with the first five years of the 1990s, will the next five years rhyme with the previous five years of that decade?
Short answer: The parallels are real, the differences are important, and the stakes for portfolios are high. I have said before, and I stand by it: There is a non-zero chance of a bubble forming. However, I will add some new spice to it: We are dangerously close to a bubble now, based on Nick Colas of Datatrek’s rule of thumb, which states that whenever the S&P 500 doubles in a short period of time, it is a bubble. Colas says, “A double is a sign of speculative excess because macro conditions are never so different that asset prices should rise 100 percent over a short period of time.” If the S&P 500 rises another seven percent, it would have registered a double from its October 2022 low. Is three years a “short period of time”? Also, has the advent of artificial intelligence made it different this time?
It took six years into the 1990s for the index to double in value. From 1996 to 1999, it doubled again. So, what is in store for the S&P 500 for the rest of the decade? Another double? Or imminent demise?
The first five years: 1990–1994 vs. 2020–2024
Both periods began with shock and a recovery. The early 1990s started with a recession tied to the savings and loan crisis, followed by a rebound. The early 2020s began with a pandemic shutdown and the fastest bear-to-bull transition on record. By late 2023, GDP growth had re-anchored to around three percent, a profile reminiscent of the mid-’90s when the recovery quietly set the stage for a productivity boom in the back half of the decade.
Another shared thread is the adoption of important innovations, first the internet and now artificial intelligence. Both periods are industrial revolutions in their own right. The First Industrial Revolution lasted approximately 80 years and peaked in the late 1880s, when water and steam power mechanized factories. The Second Industrial Revolution lasted 40 years and immediately followed the first; the widespread adoption of electricity and internal-combustion engines characterized it.
The Third Industrial Revolution (the Digital Revolution) began silently in the 1950s and, in many ways, continued all the way right up until the handoff to the fourth. The third revolution was marked by the use of semiconductors, personal computers, wireless communication, and, of course, the internet. The transition to the current and Fourth Industrial Revolution became more evident in the previous decade, marked by the widespread adoption of 3D printing, advanced robotics, cloud computing, and, of course, more recently, artificial intelligence.
Internet use transitioned from niche use to a mainstream phenomenon, with roughly half of U.S. adults online by 2000 (up from about one in 10 in 1995), a diffusion curve that supercharged the profits of the era’s most significant companies (but also left many companies clamoring for market share and profit). In the current decade, generative AI has transitioned from curiosity to a daily tool at an even faster pace. McKinsey’s 2024 survey “The state of AI in early 2024” found 65 percent of organizations regularly using generative AI in at least one function, roughly double the share from 2023. On the consumer side, ChatGPT reached an estimated 100 million monthly users in just two months, the fastest user growth in app history.
Markets notice adoption curves. The S&P 500 delivered two outsized years in a row, in 2023 and 2024, a pattern that investors grew accustomed to in the 1990s but is actually relatively rare. Technology leadership is again dominant; by various measures, tech and tech-adjacent names account for nearly a third of the index weight. That concentration is a few percentage points lower than during the peak of the 1990s bubble, but close enough to cause unease—if you are looking for it. However, that level of concentration is a feature of eras when platform technologies begin to scale. It was desktop operating systems and the internet then; it is cloud computing and AI now.
IPOs and animal spirits: then and now
Stock market fervor can be measured in part by investor appetite for new issues. Netscape’s 1995 initial public offering (IPO) debut cracked open the going-public window; by 1999, a record 544 U.S. offerings had raised approximately $69 billion ($130 billion today, adjusted for inflation), and first-day pops became cocktail-party fodder.
The IPO window got slammed shut from 2022 to 2024 under Federal Trade Commission Chair Lina Kahn, who kiboshed many deals. The IPO window creaked open again in the third quarter of 2025, the biggest issuance quarter since 2021, with 64 IPOs raising $14.6 billion. That is a generous bounce from the freeze under Kahn, but still far from hysteria.
Also in favor of bullish investors today, the rush of companies to IPOs in the 1990s was largely unprofitable internet hopefuls, trading on a combination of hope and hype. There is a similar energy with recent transactions, but companies are generally much more profitable and have demonstrated success.
The Fed voices: “irrational exuberance” vs. “fairly highly valued”
No comparison to the 1990s is complete without Alan Greenspan’s famous line. In December 1996, he asked, almost rhetorically, “How do we know when irrational exuberance has unduly escalated asset values?” The market proceeded to double again before the bust.
Jerome Powell’s tone has been cooler, but the echo is there. In September 2025, he noted, “By many measures … equity prices are fairly highly valued,” while adding he did not see elevated financial-stability risks. The market flinched, then moved on, giving off a distinctly 1990s vibe.
In both decades, the Fed acknowledged risks without calling the top. The job is not to time the quote but to manage the risk around it.
Risks that rhyme, and those that don’t
The 1990s had genuine scares: the Mexican peso crisis (1994), the Asian contagion (1997), the Russian default, and the implosion of the Long-Term Capital hedge fund (1998). Each pulled a chunk out of returns; each also kept complacency at bay, ironically extending the stock market’s gains (the market climbs a wall of worry).
The list of crises in the 2020s is heavier, yet still similar in that they each took a significant toll on returns: a pandemic, war in Europe and the Middle East, the worst inflation surge in 40 years, a regional-bank wobble, and trade policy comparisons matter most for the bull’s argument.
- Productivity: The late-1990s surge in worker output allowed the economy to grow faster without igniting inflation, giving the Fed room to let the economy run hot right up until the end. Early signs suggest that we may (eventually) be on that path again, as AI diffuses beyond Big Tech and into companies across all industries.
- Stimulus: In the 1990s, short-term interest rates spent years between five percent and 6.5 percent while the expansion chugged and stocks surged. We may be in a similar regime now, where inflation is running above target but relatively stable, and the labor market has cooled while the broader economy continues to expand. The Fed can make monetary policy more accommodative (i.e., cut interest rates) and let the economy run above trend as it attempts to stabilize the labor market.
What the next five years could look like if we rhyme with 1995–1999
Here is the optimistic rhyme. In the latter half of the 1990s, the U.S. experienced a rare alignment: disinflation, a productivity boom, a wealth effect that boosted consumption, and an investment super-cycle centered on networking and the commercial web. The S&P 500 compounded at an annual rate of more than 20 percent from 1995 to 1999. Volatility was not absent, but the trend was powerful.
If AI diffusion and the Fourth Industrial Revolution over the next five years mirror the Digital Revolution of the 1990s, corporate earnings will likely drive stock prices. Year to date, the S&P 500’s roughly 14 percent rise has been driven chiefly by earnings growth (eight percentage points) and dividends (one percentage point), with just five percentage points from multiple expansion. Over the last five and a half years, the S&P 500 has increased by approximately 125 percent, with roughly 77 points attributed to earnings growth, 19 points to dividends, and 30 points to multiple expansion. That is a fundamentals-led stock market rise, not a speculative melt-up.
If the next five years resemble 1995–1999, we will experience more upside than many investors expect and more drama than some are comfortable with. My base case is a civilized version of that rhyme: earnings-led gains, periodic air pockets, and a genuine productivity uplift that keeps inflation tame enough to let compounding of stock market returns do its quiet work. I am confident of that path, but I am not blind to the issues, especially the stock market’s valuation.
While acknowledging that a whopper of a correction could come at any time, it is also worth noting that bubbles are more vulnerable to the central banks raising interest rates than they are to the sky-high prices alone. That is good news for bulls in this environment.
The Federal funds rate (the interest rate set by the Federal Reserve) hit a low of two percent in September 1992, following the 1991 recession. Interest rates began to steadily increase in February 1994, reaching a high of 6.5 percent in May 2000. That interest rate hiking cycle was a primary culprit in deflating the tech bubble. In the current cycle, today’s interest rates are more than two percentage points beneath the peak level of the 1990s. Today’s interest rates are near the levels of May 1994 and are expected to decline, not rise.
Greenspan gave us an immortal phrase; Powell gave us a sober reminder. The right approach for growth-oriented investors may be to remain in the market but be prepared to jump out if necessary.
Allen Harris is an owner of Berkshire Money Management in Great Barrington and Dalton, managing more than $1 billion of investments. Unless specifically identified as original research or data gathering, some or all of the data cited is attributable to third-party sources. Unless stated otherwise, any mention of specific securities or investments is for illustrative purposes only. Advisor’s clients may or may not hold the securities discussed in their portfolios. Advisor makes no representation that any of the securities discussed have been or will be profitable. Full disclosures here. Direct inquiries to Allen at AHarris@BerkshireMM.com.
