WE Think: How Long Can the AI Liquidity Siphon Last? Every Bubble Peaks in a Similar Way

In May, global equity markets saw an accelerating rally in AI-related sectors, accompanied by a powerful liquidity siphon effect. South Korea, Taiwan, and the Nasdaq have gained 101%, 54%, and 16% year-to-date, with May gains of 28.45%, 14.92%, and 8.36%, respectively. China's equity market has become increasingly polarized: the ChiNext and STAR Market significantly outperformed the Shanghai Composite and CSI 300, while AI infrastructure beneficiaries continued to reach new highs. In contrast, traditional sectors such as consumer and healthcare stocks continued to make new lows, creating an unusually stark market divide.  

Can AI Infrastructure Spending Continue to Grow at This Pace?

The emergence of ChatGPT triggered a global race among technology giants to develop large language models, making data-center construction a priority. Before 2024, annual capital expenditure by the four major U.S. cloud providers averaged around US$150 billion. Capex grew 65% in 2025 to US$413.5 billion and is projected to rise another 73% to US$715 billion in 2026. Initial guidance for 2027 points to US$970 billion, though growth is expected to slow to 36%. Chinese technology firms may reach peak investment one to two years later than their U.S. counterparts, albeit at roughly one-tenth the scale. Globally, 2026 is likely to mark the fastest growth year in this AI infrastructure cycle.  

Can Major U.S. Cloud Providers Keep Expanding Capex Indefinitely?

Combining the financials of Amazon, Google, Microsoft, and Meta reveals that before 2025, their capital spending remained at healthy levels relative to earnings. However, capex exceeded total profits for the first time in 2025. Based on current plans, capital expenditure is expected to significantly exceed profits in both 2026 and 2027. Compared with operating cash flow (OCF), capex is projected to surpass total OCF in 2026, with the gap widening further in 2027. Unless these investments generate stronger-than-expected revenue growth, continued aggressive spending is unlikely to remain economically rational.  

Why Did the Dot-Com Bubble Burst in 2000?

The internet boom of the 1990s also triggered a major wave of technology infrastructure investment, led by telecommunications operators. Their annual capital spending increased from US$30 billion in 1995 to a peak of US$160 billion in 2004—roughly 5.3 times higher than the starting level. Yet the internet stock bubble burst in 2000, long before total capex peaked. Historical data show that 2000 was the year when telecom capex growth reached its highest level; growth rates then declined year after year.  

The collapse of the dot-com bubble was not caused solely by slowing capex growth. In the second half of 1999, the Federal Reserve began raising interest rates aggressively to curb overheating and inflation, tightening market liquidity. High-growth technology stocks with weak profitability struggled under higher rates. At the same time, Microsoft's antitrust setback and disappointing earnings across the sector accelerated the reversal.  

Lessons from the Final Stage of the Dot-Com Bubble

The final surge of the internet bubble began around November 1999. Investors increasingly abandoned underperforming "old economy" stocks in favor of "new economy" winners. Mutual fund investors redeemed lagging funds and piled into top-performing technology funds, accelerating the liquidity siphon effect. History has already revealed the eventual outcome for those who joined the rush near the peak, although many investors have forgotten the lesson.  

A review of the best-performing U.S. growth funds in 1999 shows that many delivered disastrous returns in the years that followed. The most celebrated funds at the peak of a bubble are not necessarily those that successfully navigate the full market cycle. In fact, the strongest enthusiasm often leads to the greatest disappointment later.  

1999 was one of the most challenging years for Warren Buffett and Anthony Bolton. On December 27, 1999, Barron's published a famous cover story titled "What's Wrong, Warren?", criticizing Buffett for underperforming the market, refusing to buy technology stocks, managing an outdated business portfolio, and making questionable acquisitions. Berkshire Hathaway declined 23% that year, while the S&P 500 rose 18% and the Nasdaq surged 80%. Just a few months later, the internet bubble began to burst.  

Not every fund manager can afford to remain as disciplined as Buffett. In China, public fund managers often face intense pressure from investors. Recently, several large consumer-focused funds added technology-oriented co-managers with strong performance records, suggesting that markets may once again be entering a phase of "change your strategy—or be replaced."  

Every Bubble Follows a Similar Script

The strongest performers in U.S. markets this year have not been the established technology leaders, but rather cyclical companies benefiting from improving industry conditions, particularly memory-chip manufacturers. Similar patterns have emerged in China's hard-tech sector, where optical fiber, electronic fabric, and copper-clad laminate companies have seen earnings and share prices rise together. Comparable dynamics were also evident in 1999.  

At every cyclical peak, investors convince themselves that strong conditions will last indefinitely. Cyclical businesses are reclassified as growth companies and valued accordingly. This narrative is one of the most common—and dangerous—features of late-stage bubbles.  

Most investors follow the same behavioral pattern: redeem underperforming funds and chase the year's top performers. If investors could travel back to December 27, 1999, and choose between the leading technology funds and the struggling funds run by Buffett and Bolton, most would likely make the same mistake.  

The reason is simple: short-term comfort. Unfortunately, short-term comfort and long-term success rarely go hand in hand. Investors constantly face the same choice—whether to pursue immediate satisfaction or endure short-term discomfort in pursuit of long-term gains.  

In capital markets, the road to wealth creation is usually quiet, while the road to losses is often crowded. When everyone is pursuing the same opportunity, risk is often far greater than it appears.  

Today's rush into technology star funds is fundamentally no different from investors rushing into consumer-focused funds at the end of 2020. In both cases, investment decisions are driven by recent performance rather than long-term value. Investors leave during downturns for the same reason they joined during the boom—they were attracted by past success.  

Recent Strategy/ Recommendations

Although the current liquidity siphon has been uncomfortable for many investors, history suggests that such phases rarely last long and are typically measured in months rather than years. No one can accurately predict the timing of a bubble's peak. As in business and life, long-term thinking remains essential. Consistently doing the right thing over time is more important than trying to time market turning points.  

Investors should remain open-minded toward AI infrastructure beneficiaries whose earnings are improving and valuations have not yet become clearly excessive. At the same time, many excellent companies outside the AI theme are being overlooked because of the current capital concentration. For long-term investors, this may present attractive opportunities to build or increase positions. Ultimately, the stock market remains a weighing machine over the long run.  

There is little value in obsessing over short-term price moves or portfolio fluctuations. Instead, investors should revisit the original reasons for owning an asset and assess whether it remains attractive relative to available alternatives. The greatest challenge in investing is not forecasting markets—it is filtering out noise and staying true to one's principles.  

Do not pursue what feels comfortable in the short term but proves wrong in the long term. Continue doing what is right for the long term. Ignore short-term fluctuations and rankings; only then can investors successfully navigate one market cycle after another.

本期《偉志思考》簡體中文版鏈接：

伟志思考：虹吸行情还能持续多久？每一轮泡沫高潮的剧本都是类似的