Summary of Key Points
This article reviews the entire process of the 2000 dot-com bubble burst (from its peak formation to collapse), identifying the "left-side warning signs" and "right-side confirmation signals" before and after the bubble's collapse. By comparing this with the current AI market, it suggests that AI is currently in a mid-course adjustment phase similar to the early stages of the dot-com bubble. However, there is a need to be cautious about the debt risks of highly leveraged infrastructure companies. In the future, the bubble may burst starting with the bonds of these companies, rather than directly leading to a stock market crash.
I. How Did the Dot-Com Bubble's "Peak" Happen?
The peak of the Nasdaq in 2000 did not come about gradually; instead, it was a rapid acceleration: in the five months before reaching its peak, it didn't even touch the 60-day moving average (indicating an extremely strong trend). However, within a month of hitting the peak, it broke through all long-term and medium-term moving averages. The root cause of this madness was easy access to money combined with the market's fanatic belief in the "internet myth":
- During the 1998 Asian financial crisis, the Federal Reserve cut interest rates three times, fueling the already overheated internet market.
- In 1999, the Fed began raising interest rates, but the market initially didn't believe it. The Nasdaq fell by 10%-15% in January (due to concerns about rate hikes and the Microsoft monopoly case), but investors thought the internet would outperform the rising interest rates and instead rose another 30%, turning the decline into a "cleaning" of the market, until it reached a new all-time high on March 10.
The fear associated with reaching the peak is twofold: trying to escape before the peak may result in missing out on further gains (for example, selling too early only for the market to rise another 30%), while trying to escape after the peak means facing a significant pullback (for example, selling just as the market starts to fall, potentially incurring substantial losses).
II. "Left-Side Signals" Before the Bubble Burst: Warning Signs Hidden in Industry Details
More than 20 days before the Nasdaq peaked, the index was still setting new highs, but there were already signs of weakness within the industry—these were not direct negative signals, but indications of slowing growth:
1. Leading companies started to show signs of slowdown: The CEO of Cisco stated that orders had normalized (meaning growth was no longer as fast as before), similar to Meta selling excess GPUs, which raised doubts about future growth prospects.
2. Analysts began to discuss capital expenditure: Although operators were still investing, the pace was not as frantic as in previous years. Once discussions about this increased, stock prices began to weaken.
3. The upstream parts of the supply chain showed signs of fatigue: The optical communications sector (such as JDS Uniphase) saw increased volume but stagnating growth—those with the best understanding of the industry had already started selling.
4. IPOs became irrational: Companies without profits were still experiencing huge gains on their first day of trading (for example, VA Linux Systems), as funds sought better investment opportunities and turned to companies with attractive narratives.
The common theme among these signals was a decrease in the second derivative of growth (i.e., growth was not stopping altogether, but the rate of growth was slowing down). However, due to previous false declines, the market was reluctant to exit the market early.
III. "Right-Side Signals" After the Bubble Burst: How Did the Dominoes Fall?
After hitting its peak on March 10, the Nasdaq fell by 36% in one month. At this point, the signals indicated a confirmed collapse:
1. The most speculative sectors collapsed first: 2B internet companies (without clear business models) were the first to experience sharp declines—everyone suddenly realized that not all internet companies could be profitable.
2. Increasing profit warnings: Analysts lowered their forecasts for communications companies, and as this trend intensified, it led to a negative feedback loop (the more prices fell, the more sales declined).
3. The financing chain broke: Many internet companies relied on stock market financing; when stock prices dropped, they could no longer obtain funding, exacerbating the panic in the market.
4. "Black swan" events accelerated the collapse: The Microsoft anti-monopoly case (unrelated to the internet sector) led to a massive redemption of funds by investors, forcing fund managers to sell stocks and triggering a sharp market drop.
The collapse was not caused by a single event; it was a chain reaction involving **doubts → seeking evidence → the stock market crash affecting the industry → the breakdown of the financing chain.
IV. At Which Stage of the Dot-Com Bubble Does the Current AI Market Resemble It? What Are the Risks?
The current AI market has not yet reached the stage of a bubble burst; it is more akin to a mid-course adjustment similar to 1998 or early 2000, as demand for tokens is still growing rapidly. The four key industry indicators of the dot-com bubble (sufficient bandwidth, reduced capital expenditure, slowing orders, and excess fiber optic capacity) have not yet appeared. However, there are risks of secondary crises:
- The dot-com bubble lasted longer due to subsequent debt issues (such as WorldCom's accounting fraud). The risk with AI may not lie in the stock market itself but in highly leveraged infrastructure companies: pure GPU leasing firms and highly leveraged IDC companies that rely on debt to build data centers and purchase GPUs, relying on long-term contracts.
- The bonds of these companies could be the first to face problems (for example, soaring yields on high-yield bonds, or failures in bond issuances by Oracle). This is more concerning than the Nasdaq's decline because large internet companies have now shifted their risks to third-party infrastructure providers.
In short, AI has not yet reached its peak, but investors should watch out for companies that are borrowing money to build computing power. The potential collapse of these companies could be the precursor to a broader bubble burst.
V. Lessons for Ordinary Investors: How to Act on Left-Side and Right-Side Signals?
- Left-side selling: Pay attention to industry details (such as slowing growth in leading companies or stagnating performance in upstream sectors), but the error rate is high, so you may miss out on further gains if you sell too early.
- Right-side selling: Look for signals from both the stock market and the industry (such as sector collapses or breaks in the financing chain); the error rate is lower, but you need to act quickly at critical moments.
- In the current AI market, there is no need for panic, but avoid highly leveraged AI infrastructure companies, as they are the most likely to face financial problems.
In conclusion, history may not repeat itself exactly, but the logic behind bubbles is similar: madness stems from easy access to money and inflated expectations, and the collapse begins with cracks in growth projections, which are then amplified by debt issues. Ordinary investors should focus on industry details and stay away from high-risk leveraged assets.