CITIC Construction Investment: North American CSP capital expenditure accelerates growth, continues to recommend the AI computing power sector

According to the Zhitong Finance APP, CITIC Construction Investment has released a research report stating that four CSP manufacturers in North America have successively announced their third-quarter reports, with capital expenditures accelerating growth, and all companies maintaining a positive attitude towards future infrastructure investments. In the short term, driven by rumors of increased orders, performance expectations for the third quarter, and the easing of China-U.S. relations, the AI computing power sector, including optical modules, has seen significant increases. The firm believes that from a mid-term perspective, it still recommends continued attention to the AI sector. On one hand, it continues to recommend the AI computing power sector, including core companies in the North American and domestic computing power supply chains; on the other hand, it also suggests paying attention to the AI application sector, including progress in edge AI, such as IoT module companies.

The main points of CITIC Construction Investment are as follows:

The four CSP manufacturers in North America have successively announced their third-quarter reports, with capital expenditures accelerating growth, and all companies maintaining a positive attitude towards future infrastructure investments. The total capital expenditures of the four major cloud manufacturers in North America for Q1 2025 amounted to $77.3 billion, a year-on-year increase of 62%. For Q2 2025, total capital expenditures reached $95.8 billion, a year-on-year increase of 64%. In Q3 2025, the total capital expenditures of the four CSP manufacturers in North America amounted to $113.3 billion, a year-on-year increase of 75%, and a quarter-on-quarter increase of 18%.

Amazon (AMZN.US): Capital expenditures for Q3 2025 were $35.095 billion, a year-on-year increase of 55.15%. In the past 12 months, new data center capacity exceeded 3.8 GW, with plans to add another 1 GW in Q4 2025. The company expects total capital expenditures for 2025 to reach approximately $125 billion, with further increases planned for 2026, primarily directed towards data centers, power, and chips required for AI.

Microsoft (MSFT.US): Capital expenditures for Q3 2025 were $34.9 billion, a year-on-year increase of 74.5%, with about half allocated to short-term assets (mainly GPUs and CPUs), and ongoing replacements of decommissioned servers and network equipment. The company plans to increase total AI capacity by over 80% this year and to approximately double the total area of data centers within the next two years.

Google (GOOG.US): Capital expenditures for Q3 2025 were $23.953 billion, a year-on-year increase of 83.39%. The vast majority of capital expenditures were invested in technological infrastructure, with about 60% invested in servers and 40% in data centers and network equipment. Expected capital expenditures for 2025 are projected to be between $91 billion and $93 billion, higher than the previous expectation of $85 billion. Looking ahead to 2026, the company expects capital expenditures to increase significantly.

Meta (META.US): Capital expenditures for Q3 2025 were $19.37 billion, a year-on-year increase of 110.5%, primarily invested in servers, data centers, and network infrastructure. The company expects capital expenditures for 2025 to be between $70 billion and $72 billion, higher than the previous forecast of $66 billion to $72 billion, and currently anticipates that capital expenditures in 2026 will grow significantly more than in 2025 Currently, capital expenditures of CSP manufacturers in North America are still in a phase of rapid growth, and the outlook is optimistic. The bank believes that the user penetration rate of AI large models is still relatively low, and the development of large models is still in the early to mid-stage, with the industrialization cycle just beginning. The investment in computing power brought about by large models is on the rise, and capital expenditures will grow alongside the increase in large model revenues, with a potentially high ceiling for investment.

In the short term, driven by rumors of order increases, expectations for third-quarter performance, and the easing of China-U.S. relations, the AI computing power sector, including optical modules, has seen significant gains. As expectations materialize, and given the situation where "demand is optimistic, but supply capacity may be insufficient," the sector is bound to experience fluctuations and adjustments. However, the bank believes this does not signify the end of the industry's prosperity trend, and from a mid-term perspective, it still recommends continued attention to the AI sector. On one hand, the bank continues to recommend the AI computing power sector, including core companies in the North American and domestic computing power supply chains; on the other hand, it also suggests paying attention to the AI application sector, including advancements in edge AI, such as IoT module companies. Additionally, the bank recommends focusing on the submarine cable and optical fiber cable sectors, which have not seen significant increases previously but may improve in the future, the military communication sector, and the quantum technology sector, which is being heavily invested in by domestic and foreign giants.

Risk Warning:

Changes in the international environment may impact the security and stability of the supply chain; the development of the artificial intelligence industry may fall short of expectations; market competition may intensify; exchange rate fluctuations may affect the foreign exchange earnings and gross margins of export-oriented enterprises; the development of the digital economy and digital infrastructure may not meet expectations; the cloud computing business development of telecom operators may fall short of expectations; capital expenditures of operators may not meet expectations; capital expenditures of cloud vendors may not meet expectations; demand in the communication module and smart controller industries may not meet expectations