CITIC Securities: North American cloud providers' capital expenditure continues to grow rapidly; pay attention to undervalued computing power and high-dividend targets
In Q2 2026, capital expenditures by leading North American cloud service providers will continue to grow rapidly.
According to Financial Associated Press APP, CITIC Securities has released a research report stating that, in Q2 2026, North America's leading cloud service providers will continue to see strong growth in capital expenditure. The total capex guidance for the four major North American cloud companies for 2026 is approximately $720 to $745 billion, reflecting their ongoing expansion around AI infrastructure.
Currently, the AI computing power industry chain remains highly prosperous, but the market has recently seen a significant correction, resulting in potential oversold conditions. Going forward, it is recommended to continually monitor: 1) The ARR of large models, especially the growth of ARR in Coding scenarios, since recent price reductions and promotions by North American large models may create short-term bottlenecks in ARR growth, potentially affecting market expectations for future computing power demand; 2) The deployment and development of large models in application scenarios outside of Coding; 3) The price trends of the computing power inflation chain in the coming period; and 4) The financing conditions and market risk preference across various segments of the AI industry chain.
In addition, considering the heavy AI industry chain holdings in the second quarter and recent market adjustments, with signs of rotation between high and low valuation stocks, it is suggested to focus on undervalued, high-dividend targets.
CITIC Securities' main viewpoints are as follows:
In Q2 2026, North America's leading cloud service providers will continue to ramp up capex, with quarterly expenditures of approximately $54.2 billion for AWS, $41.0 billion for Microsoft (MSFT.US), $44.9 billion for Google (GOOGL.US), and $31.1 billion for Meta. The total is $171.2 billion, maintaining rapid sequential and year-on-year growth.
For the full year, Amazon's total capex guidance is about $220 billion, further up from around $200 billion previously planned. Microsoft's guidance is approximately $175 billion for the calendar year 2026; disclosure standards have been adjusted as some data center leases are now classified as operating rather than financing leases. Alphabet's capex is between $195 and $205 billion, with a midpoint of $200 billion. Meta's capex is about $130 to $145 billion, midpoint $137.5 billion, including principal repayments for financing leases. Throughout 2026, the total capex guidance for the four major North American cloud companies is approximately $720 to $745 billion, with a midpoint of about $732.5 billion—an obvious increase over the Q1 guidance midpoint of $710 billion. This reflects ongoing major expansion around AI servers, GPU clusters, data centers, as well as power and network infrastructure among North American cloud giants, with AI infrastructure construction still in full swing.
In Q2 2026, revenue from cloud and AI-related businesses among North America's leading cloud companies continued strong growth: AWS, Microsoft Intelligent Cloud, and Google Cloud reported revenues of $42.2 billion, $39.3 billion, and $24.8 billion, up around 37%, 32%, and 82% year-on-year, respectively, with a combined total of about $106.3 billion—a 43% year-on-year and about 15% sequential increase. Notably, AWS saw its fastest growth rate in nearly 18 quarters, with annualized AI business revenue exceeding $25 billion and maintaining triple-digit growth. Microsoft's broader Microsoft Cloud revenue reached $59.3 billion, up 27% year-on-year; Azure and related cloud services grew 43%. Google Cloud experienced a notable acceleration in revenue growth, driven by GCP, AI solutions, AI infrastructure, and TPU system sales. Meta's total Q2 revenue was $60.8 billion, up 28% year-on-year, of which advertising revenue was $59.4 billion, an increase of 27%, indicating that its current AI investments are being monetized indirectly through improved recommendation algorithms, advertising efficiency, and user interactions. Overall, North American cloud companies are seeing AI revenue growth evolve from computing power rentals to model services, enterprise AI applications, and advertising efficiency gains, with robust revenue growth fundamentally supporting expanding capex.
Currently, the AI computing power industry remains buoyant, though the market has recently shown a clear correction and some oversold conditions. Moving forward, it is advised to closely monitor: 1) The ARR of large models, particularly in Coding scenarios, since North American large model providers recently cut prices, and any short-term ARR growth bottleneck could affect market expectations for future computing power demand; 2) Large model application development and deployment in scenarios beyond Coding; 3) The trend of computing power inflation chain prices in the near future; 4) Financing and market risk preference across the AI industry chain.
Moreover, given heavy sector holdings of the AI industry chain in the second quarter, recent market adjustments, and signs of sector rotation, it is recommended to focus on undervalued, high-dividend targets.
Risk Warning
Changes in the international environment may affect the security and stability of the supply chain and impact the pace of overseas expansion of related companies; tariff impacts may exceed expectations; the AI industry may grow slower than expected, impacting demand for cloud computing industry chain companies; intensified market competition could cause profit margins to decline rapidly; currency fluctuations may affect the foreign exchange income and profit margins of export-oriented companies, including those in ICT equipment and optical module/device sectors; digital economy and Digital China initiatives may fall short of expectations; cloud computing business progress among telecom operators may be below expectations; operator capex may fall short of predictions; cloud vendor capex may be weaker than expected; demand for communication modules and smart controller sectors may also lag expectations.
Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.
You may also like
Report: The production capacity of the three major storage manufacturers for 2027 is already sold out in advance, confirming that "next year will be the most supply-short year for storage"
According to DIGITIMES, the entire annual production capacity of DRAM and HBM from Samsung, Micron, and SK Hynix for 2027 has already been fully allocated, and the NAND Flash capacity has also been nearly sold out in advance. Previously, SK Group warned that 2027 will see the most severe supply-demand imbalance in history. Although price increases are expected to slow down compared to 2026, "high prices becoming the norm" has already become a new normal, and those who have not locked in supply ahead of time will face a shortage and inability to purchase.
Can the yen be supported by the joint efforts of the US and Japan? Wall Street believes that "if Japan doesn't raise interest rates, any intervention is useless"
The United States and Japan coordinated intervention in the yen, using nearly $100 billion over two days—the largest amount on record—but the effects have already begun to fade. Institutions such as Goldman Sachs believe that the root cause of the yen's depreciation lies in Japan's high debt suppressing interest rate increases, as the Bank of Japan is reluctant to actually raise rates and the intervention can only buy time. Goldman Sachs predicts that the Bank of Japan’s next rate hike may be delayed until January 2027, at which point carry trades will return and the yen could weaken sharply again. The US Treasury will also face losses from this intervention.
After the collapse of the South Korean stock market, retail investors "lost confidence": Remember two principles—First: do not buy Korean stocks, Second: adhere to the first rule.
In July, the Korean stock market plunged by 22%, suffering a historic crash. Previously, driven by FOMO sentiment, retail investors heavily bought into AI-related stocks such as Samsung and SK hynix, and leveraged their positions with single-stock leveraged ETFs introduced by the government, resulting in severe losses. Analysts warn that deleveraging and volatility in semiconductor stocks may persist for several months, and that rebuilding market confidence will require a longer period of recovery.
Data center "power engine" experiences explosive growth! The AI surge spills over to the power supply chain; ON Semiconductor (ON.US) Q2 cash flow quadruples and outlook beats expectations
The company's AI data center "Power Engine" business has completely taken off—demand for power management chips used in artificial intelligence data centers has surged, driving the company's stock price to soar by more than 7% in after-hours trading on the US stock market.
Crypto prices
More
