History Repeats Itself: Bursting the AI Bubble

Bubble: a state of booming economic activity (as in a stock market) that often ends in a sudden collapse

Seinfeld is America’s favorite sit-com, OJ Simpson has fled the police in his white bronco, and Michael Jordan is taking a break from basketball. These are the cultural events that occupied most people’s minds in 1994, yet the most crucial development that year was the expansion of the internet. The battle of browsers (the fight for market share of web browsers) was just kicking off between Netscape Navigator and Microsoft. The internet was a rapidly growing technology, and these two were the first to capitalize on the World Wide Web as a potentially profitable venture. Estimates suggest that the internet was still relatively unknown, with only about 11 million users in 1994, but by 1996 that number had jumped to 45 million. Consumers of the Internet had finally beaten the “geek in basement” stereotype by 1998, as the total users reached roughly 147 million. Windows computers now came with Internet Explorer preinstalled, and users could order books and music from Amazon.com and send emails to their buddies. 

Most importantly, markets were buzzing with excitement at the sheer potential of the Internet. A company with zero earnings and an idea would receive millions of dollars of funding from venture capitalists by simply mentioning the word “Internet “in their business plans. By 1999, total U.S. venture capital investments rose to over $10,000,000,000. As seen in Figure 1, that number would grow exponentially per quarter, until it peaked in March of 2000.

This outburst of venture capital occurred for two reasons: (1) there was plenty of capital available and (2) there was lots of money to be made. The Federal Reserve (Fed) had increased interest rates in response to the Gulf War recession in 1994-1995. The Federal Funds Rate was up to 6.00% by the first quarter in 1995. After the soft landing of the economy, Fed chair Alan Greenspan believed it was time to loosen the money supply and proceeded to cut rates 6 times between 1995-1998. In 1996, the Telecommunications Act relieved the Internet sector of stringent regulations. Then in 1997, the Taxpayer Relief Act decreased marginal tax rates and capital gains tax rates. These three developments led to an abundance of capital primed for investment, and opened the door to venture capitalist financing internet startups. 

People quickly realized how much money could be made in the emerging and rapidly advancing internet market. Investors continued to pour money into these startups, even when interest rates went back up in 1999 in response to the overspeculation of internet stocks. Any company with an idea related to the internet –with zero earnings– would become flushed with cash within weeks, and within months would have an IPO with shares soaring to prices higher than blue chip companies unrelated to the Internet. For example, theGlobe.com (who even is that?…exactly) went public in late 1998. They had set their offer price at $9 per share, but opened at $87. The public frenzy was really taking off in 1999 and early 2000. The NASDAQ opened at 1574.10 on January 2, 1998. On March 10, 2000, the NASDAQ reached a peak of 1532.52, a 4 year gain of roughly 390%.

Then it all came crashing down. NASDAQ fell by 78% over the next two years, the U.S. entered a recession, and thousands of companies went out of business. 

What happened? The internet boom was a prime example of an economic bubble. The most simple definition of a bubble is when a good’s price becomes higher than its fundamental value. Most of the time, the fundamental value of a good matches its price. Human nature isn’t perfect though, and sometimes people overestimate the potential value of a good. In this specific case, the good was stock of internet companies, and people got way too excited. People saw the prices of these stocks skyrocketing and they didn’t want to miss out on an opportunity to make seemingly free money. To quote Steely Dan, “a world where all is free … only a fool would say that.” Of course, knowing the exact fundamental value of something is nearly impossible, but there is only so much demand in a market for goods. It was clear that too many internet companies were being invested in as if they had significant market share, before they even had a competent product. 

Recede, Reset, Repeat

So why is this story important today? Because history repeats itself. Although not always in the exact same form, bubbles have been occurring since the formation of modern economies. From tulip-mania in the 1600s to the housing bubble in 2005, people have a tendency to overvalue certain goods. The story of the dot-com bubble can be compared to our current situation with AI. Over the last 5 years, NVIDIA stock has increased about 3,000%. Microsoft and Apple are heavily investing in AI technology and now have market caps up to nearly 3 trillion dollars. For reference, Microsoft’s market cap in 2022 was 1.787 trillion. Every day, the news is discussing how tech stocks are pushing the S&P 500 forward, despite high interest rates and inflation. Are these stocks being overvalued? Maybe, but the truth is we are still early in the development of AI. 

At the moment, large language models (LLM) are the dominant form of AI technology available to the public. Companies like OpenAI have released ChatGPT and created other AI programs for companies like Microsoft and Google. LLMs are very impressive, and the technology has certainly inspired a frenzy of excitement from investors and the general public. But LLMs are only the tip of the iceberg for what AI potentially holds. That is why these stocks are rising so much. Since the price of a stock is a reflection of what a company’s future income looks like, investors are betting that AI technology will become even more advanced and integral in our everyday life. If I were to compare our current moment in time to the buildup of the internet bubble, I would place us around 1994. Although not everybody uses or is aware of LLMs, the corporate world is fighting to gain market share in the emerging market. Google is trying to get Bard up to speed with ChatGPT and Microsoft bought OpenAI to pursue their own AI search engine. This seems similar to the early stages of the browser wars in the mid-1990s. 

The main difference is that AI technology is simply being integrated by already well-established tech companies. Many people may have only heard the name NVIDIA in the last couple years, but the corporation has been leading the market for GPUs for more than a decade now. The Big 7 companies that push the S&P 500 and NASDAQ into record highs every month already have a very strong fundamental value. That is what makes it difficult to determine whether the AI boom is a bubble or not. At the moment, LLM technology alone isn’t enough to justify the massive gains these stocks have made. People are betting on new AI technology, like Edge AI. And although integrating AI has been interesting to see in some companies, it is still unprofitable. AI can’t be commercialized until the cost of building and maintaining massive databases and swaths of programming can become much, much less expensive. 

This bet that investors are making on tech stocks creates some potential risks. Companies like NVDA, AMD, MU and TSMC are all in the market of producing chips and AI processors. The demand for their products is derived from the tech companies (like Microsoft, Apple, Google) who are using these chips to create AI software. The tremendous increases in stock price for these chip manufacturing companies implies that demand is also expected to skyrocket. But demand has already skyrocketed, and now in order for demand to continue to increase AI needs to become more than just LLMs and become a profitable product that the average person wants. The problem is that AI advances have slowed down since this frenzy started, and without any new major advances, these large tech companies and chip manufacturers are bound to start losing traction in the stock market. A market correction (a decrease in stock market between 10%-20%) is possible, but the major bubble pop we saw in 2000 is years ahead. 

What becomes dangerous is when companies solely dependent on AI grow at extraordinary levels in the market. At the moment, there aren’t too many companies like this. But many of the chip producers and database companies are very dependent on the derived demand for AI. Stocks like AMD and Micron take a significant hit if AI can’t be commercialized effectively. Another problematic trend is companies with financial weaknesses increasing in value due to their relation with AI. For instance, Accenture is a consulting firm that released their earnings about a week ago. Earnings were lower than what Wall Street was hoping for (which usually means their stock price falls), but in their report they mentioned how they plan to take on more consulting for AI businesses. Their stock price went up. Although it may not seem like major news, what happens when every consulting firm begins advertising their potential AI customers? There are only so many AI businesses that need consulting, so some of the firms whose stock price only increased due to their hypothetical work with AI companies are bound to be overvalued. Expand this theory to all companies related to AI, and the market begins to look very fragile.

These are all early signs and warnings of a potential bubble. Whether or not we’re in a bubble now is hard to tell. But if my estimate is correct, and we’re around 1994 right now, as interest rates finally begin to fall and AI technology becomes more affordable and profitable in a couple years, we may begin to see what happened in 1996. That is, small startups with zero earnings, but an AI related business idea, will be flooded with cash by venture capitalists and large IPOs. Maybe history really will repeat itself.

The opinions expressed within this piece represent the views of the author alone and do not necessarily reflect the views of The Jefferson Independent.