CICC's research report indicates that capital expenditure is an effective indicator for assessing the progression of stock market trends. Currently, the growth rate of AI-related capital expenditures remains high, with robust demand continuing to support AI investments. Furthermore, compared to the dot-com bubble era, the current AI market rally may not have peaked yet, and the stock price increases of leading companies in this cycle are more driven by earnings rather than valuation expansion, suggesting a healthier uptrend.
For the current AI market cycle, although there are concerns about the sustainability of AI capital expenditures, major AI companies such as Alphabet (GOOG.US), Amazon (AMZN.US), Meta (META.US), and Microsoft (MSFT.US) all reported Q1 2026 earnings that exceeded market expectations. The continuation of capital expenditures signals that companies are still sending positive signals to the market, suggesting the current AI rally may not be over in the short term.
CICC's main views are as follows:
Global markets rebounded in April, with technology narratives leading gains. Japanese and South Korean stock markets and the Nasdaq led the advance. Eased liquidity conditions contributed to a weaker US dollar, while the Renminbi performed strongly. Gold showed relative weakness.
Specifically, Japanese and South Korean markets, along with the Nasdaq, which have high exposure to semiconductors and AI, led the gains. The South Korean Kospi index rose 3.1% for the month, the Nikkei 225 gained 1.6%, and the Nasdaq advanced 1.5%. The performance of A-shares was better than that of European and other emerging markets. The CSI 300 rose 0.8%, while Germany's DAX and the UK's FTSE 100 gained 0.7% and 0.2%, respectively. India's Sensex rose 0.7%, and Brazil's IBOV fell 0.1%. Market style shifted back to large-cap growth. The MSCI World Large Cap and Growth indices rose 1.0% and 1.3%, respectively, while the Small Cap and Value indices gained 1.0% and 0.8%, respectively. The 10-year US Treasury yield rose 10 basis points to 4.4%, and the US Dollar Index fell 1.9% to 98.1. Gold continued its slight decline, falling 1.1% to $4,618 per ounce.
The pricing of geopolitical risks receded, with the technology sector leading gains and AI remaining the main trading theme. On one hand, the easing of geopolitical tensions that suppressed the market in March drove the rebound, with previously leading gainers/losers reversing roles. This indicates that geopolitical disturbances primarily affect market sentiment and liquidity; once the suppressing factors ease, major asset prices largely revert to their original trajectories. On the other hand, the market returned to trading on fundamentals, with AI remaining the core theme. Since the global equity rebound in late March, Japanese and South Korean markets, which are heavily reliant on energy imports, have led gains, suggesting that under the dominance of the tech rally, concerns about energy security stemming from geopolitical risks have temporarily subsided. The US Mag 7 index and China's AI sector have once again outperformed the broader market since late March. Across US, Chinese, Japanese, and South Korean markets, the technology sector led the gains. Furthermore, within these markets, the industrial sector also performed well, indicating the market has begun to reprice safety assets represented by manufacturing strength.
The AI market rally has made a comeback following a correction in Q4 2025. A key question is: how long can the AI thematic rally that has led the market since 2023 continue? Which indicator should be used to judge the trend of the AI rally? Reviewing two major US technological revolutions (semiconductors and communications/internet) since WWII and their investment opportunities, capital expenditure (Capex) is an effective indicator for signaling the progression of market trends. Although there are current concerns about the sustainability of large tech companies' Capex, Capex in the AI field is still in a phase of rapid growth, suggesting the AI thematic rally may persist over the next few quarters.
First, the semiconductor and information technology revolution of the 1970s. Jack Kilby of Texas Instruments and Robert Noyce of Fairchild Semiconductor independently invented the integrated circuit in 1958 and 1959, respectively. In 1968, Robert Noyce and Gordon Moore co-founded Intel, which released the first commercial microprocessor, the 4004, in 1971 and went public the same year. In 1972, Intel released the second-generation 8008 microprocessor, driving rapid development in computer hardware and software, followed by the launch of personal computers by Apple and IBM. The development of semiconductor technology spurred rapid growth in related investments. Information technology investment accounted for only 7% of equipment investment in 1968, but with the proliferation of microprocessors, this share rose to 56% by 1998. Increased semiconductor investment drove up stock prices of related companies like Intel. From 1973 to 1980, Intel's stock price rose over tenfold. As US-Japan semiconductor trade friction intensified in the 1980s and US semiconductor investment growth slowed, Intel's stock price began to decline in tandem.
Second, the communications and internet technology revolution of the 1990s. The world's first website appeared in 1991. The first graphical web browser, Mosaic, appeared in 1993. Its creator, Marc Andreessen, later founded Netscape, which went public in 1995, marking the beginning of the internet asset bubble. The US Telecommunications Act of 1996 signaled a wave of investment in large-scale fiber-optic networks and wireless communication equipment. The number of internet servers surged from 23,000 in 1995 to 30 million by 2001. Real IT investment related to computers, communications, and software grew at an average annual rate of 24% between 1995 and 2000, boosting GDP growth by approximately 0.75 percentage points. The communications and internet investment wave drove rapid stock price increases for related companies like Cisco. From Q1 1995 to Q1 2000, Cisco's stock price rose nearly 40-fold, while the Nasdaq index rose nearly 6-fold. In Q1 2000, the sequential growth rate of communications equipment investment peaked, coinciding with the peak of the US stock market bubble, where the Nasdaq's PE valuation approached 70 times and Cisco's PE reached nearly 110 times. Subsequently, Nasdaq and Cisco stock prices declined in tandem with the slowdown in communications equipment investment growth.
Historically, capital expenditure has been an effective indicator for judging the progression of stock market trends during technological advances. Logically, capital expenditure represents long-term operational decisions by companies. During periods of major technological change, at the micro level, higher capital expenditure is more likely to indicate more advanced products and greater market share. Companies reduce capital expenditure when long-term profit expectations are poor or when they choose to exit competition. At the macro level, higher capital expenditure signifies that iteration of new technology products is still accelerating. Capital expenditure growth overall declines when technological development reaches a plateau or when the marginal return on capital expenditure decreases. In other words, during major technological shifts, capital expenditure is a tangible choice reflecting companies' long-term profit expectations. As long as tech companies' capital expenditure maintains rapid growth, it indicates confidence in future growth. Even in hindsight, capital expenditure at the end of a technological cycle may have some blind spots, but it can still support market confidence and, consequently, stock prices.
Currently, AI capital expenditure growth remains elevated, and robust demand will continue to support AI investment. Compared to the dot-com bubble period, the current AI rally may not have peaked, and the stock price increases of iconic companies in this cycle are more driven by earnings rather than valuation, making the uptrend healthier.
Currently, US investments in communications, computer equipment and software, and data centers are growing rapidly. Particularly for computer equipment and data centers, the former has risen from around $150 billion in 2023 to over $300 billion currently, while the latter has increased from $10 billion in 2023 to nearly $30 billion currently. AI-related investment now accounts for close to 6% of GDP. Compared to capital expenditure during the communications/internet era, the current pace of AI investment is not particularly fast overall. Using Q1 1995 and Q4 2022 as the starting points for the two technological cycles, communications/internet investment increased by about 60% by 1998, while current AI investment has increased by about 40%. However, for specific sub-categories, the current investment growth rate for computer equipment and data centers is faster than the investment growth rate for communications equipment and infrastructure at that time. According to Q1 2026 earnings guidance, the combined capital expenditure of Alphabet, Amazon, Meta, and Microsoft for 2026 is projected to reach $725 billion, a significant 77% increase from $410 billion in 2025.
Rapid growth in AI capital expenditure has driven stock prices of related companies like Nvidia. Since Q4 2022, Nvidia's stock price has cumulatively risen over 12-fold. Currently, AI demand remains strong. According to US Census Bureau survey data, the proportion of businesses using AI has risen from 3.7% in 2023 to 10% currently, while the proportion expecting to use AI has increased from 6.3% to 14%. Particularly since February this year, OpenAI and Anthropic have launched Agent platforms for enterprise users, gradually improving AI business models and further driving hardware demand.
Furthermore, compared to the dot-com bubble era, the current AI rally may not have reached its peak. On one hand, looking at the overall stock market, from Q1 1995 to the bubble peak in Q1 2000, the Nasdaq rose over 6-fold. Since Q4 2022, the Nasdaq has risen about 2-fold, a trajectory more consistent with the Nasdaq's performance from 1995 to 1998. The overall valuation trends of the Nasdaq during the two cycles are also relatively similar. On the other hand, looking at iconic companies, Cisco's stock price rose over 40-fold from Q1 1995 to Q1 2000, with valuation expansion contributing nearly 6-fold of that increase. In the current AI cycle, Nvidia's stock price has risen 12-fold since Q4 2022, exceeding Cisco's 7-fold gain over a comparable initial period. However, Nvidia's rise has been almost entirely driven by earnings growth, with its valuation actually declining from 40 times in Q4 2022 to around 20 times currently. Overall, compared to the dot-com bubble era, the current AI rally is relatively moderate, and the performance of iconic companies is healthier.
For the current AI market cycle, although there are concerns about the sustainability of AI capital expenditures, major AI companies such as Alphabet, Amazon, Meta, and Microsoft all reported Q1 2026 earnings that exceeded market expectations. The continuation of capital expenditures signals that companies are still sending positive signals to the market, suggesting the current AI rally may not be over in the short term.
A potential risk is that as AI investment intensifies, the capital expenditure of AI companies has already exceeded their free cash flow, meaning AI investment will rely more on debt financing and become more sensitive to financing costs. If medium to long-term interest rates rise sharply due to external shocks, the risk of a periodic correction in the AI market could increase.
Comments