The recent volatility in global technology stocks has intensified once again. The underlying causes of this fluctuation, whether last week's price hikes for Apple's product line or Meta's announcement this week to lease computing power, revolve around a core concern: will future demand for computing power slow down, and can the high growth in capital expenditure (Capex) be sustained? This is the focal point of market attention.
Market concerns about Capex growth essentially reflect worries about a peak in tech stock earnings growth. The fundamental driver of the current AI wave is the growth in global computing power demand, with Hyperscalers' capital expenditure being the core engine pulling the performance of global technology stocks. For US stocks, whether hardware or software, the consensus year-over-year earnings expectations for the next 12 months are strongly correlated with the year-over-year growth of Hyperscalers' capital expenditure.
We will not delve into predicting the future path of Capex growth at this moment, as it is inherently difficult to forecast. Firstly, market expectations are constantly adjusted based on industry developments and company guidance: from 2024 to 2026, the initial market expectations for capital expenditure growth at the start of each year differed from the final actual values by approximately threefold. Secondly, the timing of corporate capital expenditure deployment is somewhat subjective, and published capital expenditure guidance is frequently revised according to their own business progress.
However, recent market concerns have prompted us to consider a question: does a slowdown in Capex inevitably lead to the end of a market rally? In other words, if earnings growth decelerates from a high level, will the stock price inevitably fall? This is a question worth exploring in depth. We will examine this core market concern by reviewing historical cases of leading companies from the dot-com era and incorporating corresponding experiences from the A-share market.
Does Slowing Growth from High Levels Inevitably Cause Stock Price Declines? A Case Study of Microsoft from the Dot-Com Era
Given that US stocks are the core driver and emotional "anchor" of the current global AI rally, we first summarize historical lessons using the example of Microsoft (MSFT), a leading company from the dot-com era.
During the entire dot-com rally in the 1990s, while Microsoft's stock price trended upward overall, it did not maintain accelerating growth throughout. Between 1996 and 2000, Microsoft actually experienced two peaks in its earnings growth rate.
The first was in mid-1997. At that time, global PC penetration reached a stage of saturation, demand for new installations slowed marginally, corporate IT procurement entered a phase of consolidation, and coupled with the high base effect from previous strong performance, Microsoft's TTM EPS year-over-year growth peaked and began to decline from a high of around 70%.
The second, more widely recognized peak, was in March 2000. As the "Y2K" replacement cycle prematurely exhausted capital expenditure demand, leading to a sharp contraction in orders for major companies, Microsoft's TTM EPS year-over-year growth declined from over 50% and even turned negative.
However, these two peaks in earnings growth resulted in two截然不同的 outcomes for the stock price. The peak in mid-1997 only caused Microsoft's stock price to flatten; it did not lead to a systematic decline. When earnings growth resumed its upward trajectory in 1998, the stock price also began a new round of上涨. In contrast, the peak in March 2000 confirmed the stock price's high point. Throughout that year, Microsoft's stock price followed earnings growth downward, only stabilizing when earnings growth bottomed out in 2001.
Why the difference? We believe the "absolute level" of earnings growth is crucial, with 30% potentially being a key threshold. Although growth began to decline in mid-1997, it remained around 30% at its lowest, still within the "high growth" range. Therefore, the impact on the stock price was merely a flattening rather than a systematic decline. However, the decline starting in March 2000 fell below 30% and even turned negative. The breakdown of the "high growth" narrative led to a sustained downward trend in the stock price.
Why 30%? The valuation center for leading US tech manufacturing companies is generally around 30 times P/E. A 30% earnings growth rate represents a relatively comfortable level relative to this valuation.
Therefore, by reviewing Microsoft's experience during the dot-com era, we draw an important conclusion: a deceleration in growth from high levels does not necessarily lead to the systematic end of a rally. If earnings can be maintained within a high-growth range that matches the current valuation, the stock price may not necessarily experience a systematic decline. A 30% earnings growth rate is a significant threshold.
Furthermore, historical review suggests that within a large, difficult-to-disprove industry trend, the market has its own criteria for judging whether the "high growth narrative can be sustained." Only when signals of a breakdown in the growth narrative are sufficiently strong will the market completely abandon the original pricing logic; otherwise, it often awaits clearer fundamental confirmation. During the dot-com era,阶段性渗透率放缓 was insufficient to end a major industry trend rally. What truly led to the systematic end of the rally were core demand拐点 that undermined the fundamental支撑 of capital expenditure, such as "the disappearance of replacement cycle demand."
Applying this to the current cycle: First, the爆发 of Capex follows the progress of AI commercialization, which is often non-linear and unpredictable. Even if growth阶段性放缓 at some future point, it does not necessarily mean the end of the rally. As long as the earnings growth of leading companies can be maintained within a high-growth range, even if growth decelerates marginally, stock prices may not systematically decline. Second, it is essential to distinguish the underlying logic causing the Capex slowdown. If it stems from confirmed slowdowns in end-demand for the AI industry, caution is warranted. Otherwise, one can await clearer verification signals.
Does the Same Hold True for the A-Share Market?
At the individual stock level, a similar conclusion can be drawn by observing leading company Zhongji Innolight Co.,Ltd. (300308). In mid-2024, as global capital expenditure and the net profit growth of leading companies passed the fastest growth phase of the early industry low-base period, Zhongji Innolight Co.,Ltd.'s earnings growth peaked and began to decline. In hindsight, the resulting stock price volatility was merely a minor disturbance before the "main上涨 wave" (with most of the decline attributed to sentiment-driven shocks like the DeepSeek narrative and trade tensions). By mid-2025, as global capital expenditure entered a new爆发 cycle and earnings growth bottomed out and rebounded (stabilizing at 74%, indicating the high-growth logic was not systematically impacted), the stock price also entered its "main上涨 wave" phase.
Does this规律 hold at the broader statistical level? We categorized all stocks into different景气 types based on their current earnings growth rate and its change relative to the previous period, using 30% as the threshold for "high growth." We then calculated the median超额收益 relative to the entire A-share market for different景气 types from 2010 to 2025.
"High growth with accelerating景气" is undoubtedly the most favored景气 type by the market. In some years, the景气 type transitioning from "medium-low speed景气 acceleration to high growth" also receives more积极的定价 from the market due to demonstrating strong earnings elasticity or growth logic.
More importantly, when high-growth industries or companies experience景气降速 and a decline in earnings growth, their stock price performance does not necessarily decline. Instead, it depends on the level of earnings growth: 1) If it remains above 30%, maintaining high growth, they can still achieve relatively high超额收益. 2) If it drops from >30% to a medium-low growth rate of 0-30%,超额收益 will significantly收敛. 3) If it directly drops into negative growth,超额收益 will also turn negative, underperforming the market.
Therefore, statistical规律 from the A-share market lead to the same conclusion: "High growth with accelerating景气" is固然最好. However, if high-growth companies experience景气降速, their stock prices may not necessarily fall. Whether "growth falls below 30%" is a key threshold determining whether超额收益 significantly收敛.
What Real-World Verifications Should Be Monitored Regarding Capex Going Forward?
Having clarified the逻辑 of how Capex growth and earnings growth affect stock prices, regarding Capex growth itself, during the current earnings off-season, narratives from companies like Apple and Meta are essentially "stories." The upcoming US earnings season才是真正的验证窗口. What real-world verifications should be the core focus going forward?
First, the earnings season will be a crucial window for upward revisions in capital expenditure expectations. Focus on whether new capital expenditure guidance from cloud providers can打破 the expectation that Capex growth will peak in 2026. Historical experience shows that每当财报季 listed companies disclose new guidance, it serves as a key window for the market to upwardly revise Capex expectations. Based on current market consensus expectations, the third and fourth quarters of this year might be the peak for capital expenditure growth. If new guidance can打破 the expectation of peaking Capex growth, it would undoubtedly serve as another strong catalyst for the stock prices of core companies.
Second, the market expects the存量自由现金流 of North American cloud providers to turn negative this year, raising questions about the sustainability of capital expenditure. However, more importantly, the focus should be on whether future growth in operating cash flow can support high capital expenditure growth. The earnings reports of several major core cloud providers in late July are the key verification window. As long as the growth in operating cash flow generated by cloud and AI业务收入 exceeds the growth in capital支出, future free cash flow will also turn positive, corresponding to sustainable high capital expenditure. Since the beginning of this year,远期营收预期 for core North American cloud providers have been continuously revised upward. In late July, the four major North American core cloud providers will陆续披露财报, providing a key window to assess whether AI业务收入 can support their high capital expenditure growth.
Finally, fundamentally, changes in Capex are primarily determined by the return on investment (ROI), which currently is linked to the Annual Recurring Revenue (ARR) of leading large model developers. This data will become more transparent after Anthropic's上市,有望继续支撑 the "spiral上升" of ROI and Capex. As long as computing power continues to enhance AI performance and the ROI of token generation remains attractive, leading large model developers will持续加码 capital investment. The indirect收益 gained by cloud service providers as "computing power suppliers" is sufficient to支撑 their further increase in capital expenditure.
Current Allocation Strategy Considerations
As July enters the earnings season for both China and the US, the relative strength of景气 and changes in relative performance remain core线索 for allocation.
For sectors with strong market consensus on景气, such as AI computing power and upstream resources (especially AI-related materials like non-ferrous metals, chemicals, and fiberglass), which were significantly affected by overseas volatility earlier,随着国内中报业绩预告披露, they are有望逐步进入再配置区间. Within the storage, optical fiber and cable, liquid cooling, and electronic special gases sectors, several companies have already disclosed亮眼中报业绩预告, validating the high景气度 of the AI industry chain in the second quarter. Subsequent业绩预告 from leading core companies有望是更加强劲的催化.
Among these, for North American computing power chains represented by optical communication and PCB, the earlier computing power rally "缩圈" to the storage industry chain, and recent declines受海外影响较大. Currently, the valuation spread between A-share North American computing power chain leaders and domestic computing power chain leaders has accelerated its回落 to levels seen in June last year. The配置 signal significance of this development warrants attention.
Additionally, before the US earnings season arrives in mid-to-late July, the market may still seek domestic low-valuation, high-performing sectors for补涨. Currently, these mainly include: the lithium battery industry chain, innovative drugs, securities firms, agricultural chemicals, petrochemical refining, and broiler chickens.
Risk Warnings
Economic data volatility, policy easing falling below expectations, Federal Reserve interest rate cuts falling short of expectations, escalation of geopolitical tensions.
Comments