Since March of this year, shifting U.S.-Iran tensions and the change in Federal Reserve leadership have led to volatile expectations for U.S. interest rate hikes, unsettling global markets.
In earlier reports, we have extensively discussed the potential impact of the Fed's rate hikes. However, during recent roadshows, we continue to sense significant concern regarding the Fed's actions, particularly their potential effect on technology stocks.
Upon observation, a key reason this concern persists is a widespread, intuitive belief: that the previous tech bubble was burst by the Federal Reserve raising interest rates. Is this truly the case? We believe it is necessary to deconstruct the detailed circumstances of 1999-2000.
This retrospective is not about directly comparing today's AI boom to the 1999 internet experience, as the external environments are vastly different and no historical scenario can be perfectly replicated. Instead, this article aims to use this classic case study to examine, from first principles, how rate hikes, liquidity contraction, and rising interest rates actually impact technology stocks during an industry boom.
The 1999-2000 Rate Hike Path and Context: Preliminary Guidance Began in February 1999
Following the end of the Asian Financial Crisis (the Fed cut rates three times by a total of 75 bps from Sep-Nov 1998), the global deflationary cycle reversed, and commodity prices began recovering. Additionally, early 1999 saw OPEC and non-OPEC producers jointly cut production, coinciding with the outbreak of the Kosovo War (Mar-Jun 1999). Threats to Balkan and Mediterranean transport raised fears of supply disruptions, pushing oil prices from around $10/barrel to over $30/barrel.
At this point, U.S. CPI inflation also accelerated rapidly. The Federal Reserve re-entered a tightening cycle in June 1999, raising the federal funds rate from 4.75% to 6.5% by May 2000. The approximate timeline is as follows:
1. Feb-Mar 1999 FOMC meetings: The Fed did not signal explicit hikes but began expressing concerns about inflation risks and questioning the appropriateness of the then-loose monetary policy.
2. May 1999 FOMC meeting: The Fed essentially confirmed that rate hikes would begin soon. The Nasdaq Composite saw a short-term correction of about 8% before resuming its upward trend.
3. June 1999: The Fed officially began the hiking cycle. The Nasdaq showed no significant adjustment and continued its uptrend.
4. August 1999: The Fed's second rate hike. The Nasdaq showed no significant adjustment and continued its uptrend.
5. November 1999: The Fed's third rate hike. The Nasdaq showed no significant adjustment and continued its uptrend.
6. February 2000: The Fed's fourth rate hike. The Nasdaq showed no significant adjustment and continued its uptrend.
7. March 2000: The Fed's fifth rate hike. However, the Nasdaq had already peaked before this hike, marking the top of the tech cycle. From the first Fed hike to the bubble's peak, the Nasdaq gained over 90%.
8. May 2000: The Fed's final hike of the cycle, a single 50 bps increase, bringing the total cycle hikes to 175 bps.
From the Moment Hikes Were Confirmed in May '99, the Dow Jones Was Suppressed First, Peaking Well Before Tech Stocks
The Dow Jones Industrial Average faced a pullback in Q3 1999 under the dual pressures of high oil prices and rate hike expectations, briefly rallied in Q4, and then peaked in January 2000—significantly earlier than tech stocks.
This indicates that from May 1999, when the Fed's tone essentially confirmed imminent hikes, the Dow entered a volatile phase. From a pricing logic perspective, this highlights two points:
1. Sectors with average demand-side dynamics are likely to face more fundamental pressure from rising interest rates (financing costs) and high oil prices compared to sectors experiencing explosive demand.
2. For companies with less volatile earnings cycles, valuation often relies on long-duration discounted cash flow models. When the interest rate and oil price environment shifts higher, it tends to suppress the valuation of such companies.
Therefore, the slower-growing Dow Jones index (with EPS compound growth around 3% in 98-99), even if not expensive, was the first to be suppressed by the rate hike expectations.
However, Through the Fed's First Four Hikes Starting June '99, the Nasdaq Surged Over 90%
From the Fed's first hike in June 1999 until the tech bubble finally peaked in March 2000, the Nasdaq Composite rose 91%, with the peak lagging the start of hikes by 9 months. This also illustrates several points:
1. During periods of explosive demand, while rate hikes have some impact, companies are less concerned with the increase in interest rates (financing costs), or at least it is not the core of their business decisions.
2. For companies with short-term earnings explosions, investors understand this surge is likely short-term (2-3 years), not long-term (5-10 years). Therefore, this short-term EPS and cash flow cannot be heavily discounted. From a pricing perspective, it has little relation to a rising interest rate environment.
3. When rate hikes lead to liquidity contraction, investors may prioritize selling companies with slow or no earnings growth, crowding instead into the few sectors with booming prospects.
Thus, when the EPS growth of the Nasdaq 100 index surged from 15% in 1998 to 60% in 1999, even though the index's PE valuation was already at a static 55x and dynamic 45x in 1998, its valuation expanded significantly during the Fed's consecutive rate hikes.
What Drove the Nasdaq's Earnings Explosion in 1999?
Around 1995, as the internet trend was just starting, the Nasdaq 100's growth gradually accelerated but remained in the 20-30% range. Between 1997-1998, the fundamentals of tech giants had already begun to weaken. However, the anticipated order surge from Y2K created a "temporary boom" for leading internet companies in 1998-1999.
The Y2K "millennium bug" replacement wave was driven by warnings from the U.S. scientific community, media, and even the government about a potential massive network risk: if servers, systems, and PCs were not updated to the latest versions before 2000, the millennium change could cause code and system crashes.
Under policy directives from agencies like the OCC, FDA, and Department of Defense prioritizing system bug fixes, global governments and enterprises embarked on a "panic buying" spree for old servers, mainframes, personal PCs, and software operating systems in 1998-1999, igniting a capital expenditure frenzy for replacements. This led to another earnings surge for many hardware and software companies, pushing the Nasdaq 100's EPS growth to its fastest level in the 1990s: 60% in 1999.
If Not Rate Hikes, What Actually Burst the Bubble?
The capital expenditure boom driven by Y2K not only helped the Nasdaq withstand consecutive rate hikes starting in 1999 but also formed the largest tech bubble in history. However, the same factor that fueled it also led to its downfall.
The feared Y2K crisis in the year 2000 did not materialize. The pre-deployed hardware purchases had透支 future demand, making the next phase of hardware capital expenditure unsustainable, and the supply chain faced high inventory pressure.
During 1998-1999, global enterprises massively front-loaded IT capital expenditures to prevent potential widespread system failures from the "millennium bug" date-coding issue. However, when 2000 arrived, it became apparent Y2K was less severe than feared; even without full hardware/software replacements, simply modifying code could avert network risks.
This signaled an impending decline in tech equipment procurement demand, leading to order shrinkage and increased inventory pressure for leaders like Dell, HP, and IBM. Global semiconductor industry inventory levels also reached a historical high by late 2000.
Furthermore, in Q1 2000, negative signals for the tech industry and core company fundamentals emerged one after another:
Feb-Mar 2000: U.S. media extensively reported that the Justice Department's antitrust case against Microsoft was nearing a verdict, with substantial evidence suggesting monopolistic behavior, negatively impacting market sentiment.
Starting with earnings reports in March 2000, data showed disappointing sales of tech products during the 1999 Christmas holiday season.
Microsoft's April earnings report showed revenue below expectations and projected a slowdown in PC market growth. IBM's Q1 report, also released in April, missed revenue expectations.
Several once-high-flying new internet companies began to fail (e.g., luxury fashion e-tailer Boo.com, online news platform APB online, online retailer Value America).
Ultimately, the Nasdaq 100's EPS growth plummeted from 60% in 1999 to just 12% in 2000.
Returning to the Present: If Rate Hike Fears Trigger U.S. Recession Concerns, AI Would Not Be Spared
Returning to the current AI situation, aside from changes in AI's own demand, if U.S. recession fears intensify due to rate hike expectations, concerns would grow that a recession could impact the traditional core businesses of U.S. Cloud Service Provider (CSP) giants. This, in turn, could affect their cash flow and future capital expenditure capacity, ultimately impacting overall computing demand.
Around April 2025, as global markets priced in U.S. recession fears, sectors like GPUs, optical modules, and PCBs experienced one of their most significant declines in recent years.
But do rate hike expectations necessarily trigger recession fears? The most typical recent case is 2022-2023, which also faced high oil prices (Russia-Ukraine war) and high inflation, with the Fed executing the fastest rate hikes in history. At that time, most investors believed the U.S. would enter a recession between 2022-2024. However, expansive fiscal policies, including three major bills, ultimately offset the impact of rising rates.
The current environment is similar in many global economies: monetary policy shows signs of tightening or has begun to tighten, but major large economies still employ very proactive fiscal policies, ensuring relative stability in traditional demand.
Historically, during war crises, an official recession is typically triggered only when the war causes a sustained, sharp surge in oil prices, combined with the U.S. economy itself being in a fragile state at the end of an expansion cycle with high inflation and transitioning into a rate hike cycle. If the war's impact on oil prices is brief and other fundamental drivers are strong, a recession may not necessarily occur even during a hiking cycle.
Simultaneously, a characteristic of the U.S. economy since the pandemic has been "downturns without clear recessions, upturns without robust recoveries." Supporting factors have shifted through several rounds of drivers, but a state of全面 overheating has not emerged.
Without a sharp暴涨, there is no sharp暴跌. If the U.S. economy were at a cyclical peak with all indicators全面过热, then facing high oil prices, high inflation, and rate hike expectations would be far more dangerous. Currently, U.S. fundamentals are supported by resilient consumption, high-growth AI investment, and potential fiscal stimulus driven by the political cycle. Looking ahead, growth momentum may slow, but recession fears are premature.
Otherwise, Rate Hikes Are Not the Primary Fear; The Core Returns to the AI Industry Itself
In fact, the best defense against rate hikes, rising interest rates, and liquidity contraction is rapid爆发 on the demand side. Historical instances in both A-shares and U.S. stocks of industry booms occurring during liquidity contraction have proven this point.
Finally, a pointed question: Do rate hikes increasing financing costs affect the capital expenditure of CSP cloud giants? This also seems to depend on the demand within the AI industry itself. If capital expenditure is driven merely by FOMO (Fear Of Missing Out), then the increase in financing costs would give the giants serious pause. But if driven by持续爆发的云服务订单和需求, would these giants slow their capital expenditure because of a 25 bps increase in financing rates? In the first two decades of the 21st century, did Chinese people stop buying houses because interest rates were high?
Comments