Summary of Key Points
The AI stock market correction in July 2026 (with the NASDAQ 100 index falling nearly 10%) was not a bubble bursting, but rather resembled several manageable adjustments before the 2000 Internet bubble. At that time, the issues were localized (valuation, quality of earnings, macroeconomic shocks), and did not affect the fundamental health of the industries. The 2000 bubble burst because it created a vicious cycle: falling stock prices led to difficulties in financing, reduced demand, decreased earnings, and further declining stock prices. The current AI correction is a result of a combination of cyclical concerns, scrutiny of earnings quality, and macroeconomic risk aversion. While the overall AI trend remains positive, some companies have valuation bubbles that need attention. To determine if a bubble has burst, five key indicators should be watched; ultimately, only AI companies with actual profits will be worth investing in.
Detailed Analysis
1. Why were previous market crashes before the 2000 bubble able to recover?
The 2000 Internet bubble burst was not the first major crash; it was the fifth time that the market failed to recover. The reason for the previous recoveries was that the negative news remained at the market level and did not affect the fundamental health of the industries:
- 1996: Excess chip supply: Only some chip prices fell, while demand for PCs and the Internet continued to grow. Companies did not cut their IT budgets, and prices rebounded after adjustments in valuations.
- 1997: 3Com's earnings warning: This was a failure by individual companies, but leaders like Cisco still benefited from network expansion. The market shifted from a general upward trend to focusing on the best performers.
- 1997: Asian financial crisis: Employment and consumption in the U.S. remained stable, and there were no disruptions to the funding flow in the tech industry.
- 1998: LTCM crisis: The Federal Reserve cut interest rates to stabilize the market. Once liquidity was restored, Internet user growth and financing continued, and the financial shock did not halt industry development.
The 2000 bubble burst because falling stock prices destroyed the fundamental health of the industries: Many Internet companies relied on financing to survive. As prices dropped, IPOs and refinancing opportunities disappeared, customers had no money to buy equipment, and suppliers experienced a sharp decline in orders and profits, creating a irreversible cycle.
2. How does this current AI correction compare to historical events?
The current adjustment is more akin to a combination of the 1996, 1997, and 1998 scenarios, but it has not reached the level of a complete bubble burst:
- Similar to 1996: There are concerns about reaching a cyclical peak. Stocks in storage and chip sectors have fallen sharply, despite SK Hynix reporting a six-fold increase in profits—consumers are no longer concerned with short-term gains but are questioning whether these profits represent the last peak before further expansion (similar to the DRAM chip price cycle).
- Similar to 1997: There is a focus on evaluating earnings quality. Google's capital expenditure (44.9 billion) exceeded its operating cash flow (39.1 billion), and Microsoft increased its capital spending by 70%. The market began to distinguish between companies that could convert AI investments into profits (e.g., Microsoft's cloud AI services) and those that were merely wasting money (and thus sold off).
- Similar to 1998: Macroeconomic factors have amplified the decline. Rising interest rates, high U.S. bond yields, and geopolitical conflicts have caused funds to withdraw from crowded sectors, but there was no systemic leverage crisis like the LTCM one, and the credit market did not fail.
In short, what is happening now is a process of “deflating bubbles,” not a complete bubble burst.
3. Not all AI-related sectors are affected equally; these three types of companies are most at risk
The overall AI trend is sound, but the distribution of valuation bubbles is uneven. The following three types of companies may never return to their previous highs:
- Cyclical suppliers: Those that profit from chip shortages, price increases, and early customer purchases (e.g., storage chips). Once capacity expansion is completed, prices and profits will plummet.
- Heavy-capital-dependent financiers: Data center operators and cloud computing service providers. They need continuous funding to build infrastructure and purchase hardware. If financing opportunities dry up, their cash flows will be interrupted.
- Companies with a superficial AI label: Those without core technologies or customer loyalty, relying solely on the “AI” concept to drive valuations. Without these fundamentals, they will be vulnerable in market downturns.
Remember: Industry growth does not equate to company growth, and company growth does not necessarily mean shareholders will make profits (many companies survived the Internet bubble but their stock prices never recovered).
4. How to determine if an AI bubble has truly burst?
There’s no need to predict the exact peak of the market; instead, focus on whether positive feedback loops are reversing:
- Signal 1: Major customers are cutting spending. Cloud providers like Microsoft and Google have reduced GPU and data center investments for two consecutive quarters, citing that AI revenue did not meet expectations.
- Signal 2: Both prices and utilization rates are declining. Cloud GPU rental prices have fallen, and many GPUs are idle (low utilization). Even if usage increases, profits will decrease.
- Signal 3: Orders, inventory, and cash collection are all struggling. Chip and server companies are receiving fewer orders, accumulating inventory, and customers are failing to pay on time—demand is truly weak.
- Signal 4: Financing-dependent companies are defaulting on loans. Cloud computing service providers and AI startups are having difficulty repaying debts, leading to wider credit spreads.
- Signal 5: Falling stock prices are affecting real demand. Companies are laying off employees and cutting investments, which in turn reduces suppliers’ profits and drives further price declines, creating a vicious cycle.
The appearance of multiple of these signals indicates the beginning of a bubble burst.
5. Conclusion: The AI trend is correct, but don’t invest blindly
The overall AI trend (cloud revenue and model usage) is strong, but the current correction highlights real issues: overcrowded valuations, uncertain returns on capital investments, and pressures from the semiconductor cycle.
In the future, there will be a differentiation among companies. Only those that can demonstrate that their investments generate cash flows and have competitive advantages (e.g., Microsoft, NVIDIA) will see price increases. Those that rely on hype will struggle to maintain their valuations, even if they survive the current market downturn.
This is a necessary part of a technological revolution—moving from “storytelling” to “showing real performance.” Just as after the Internet bubble, only companies with solid foundations (like Google and Amazon) survived.
(Risk Warning: This analysis is based on historical comparisons and does not constitute investment advice. Market conditions should be monitored dynamically.)