Everbright Overseas: How does Microsoft's suspension of AI data center construction view the impact of tariffs on US stock cloud vendors?

The Zhitong Finance App learned that the overseas research team of Everbright Securities released a research report saying that in the short term, Trump's tariff policy may put pressure on the performance of US technology companies, mainly reflected in increased data center construction costs, potential digital service tariff countermeasures, and more cautious IT spending and cloud budgets for downstream customers. It is expected that this pessimistic forecast will make the 25Q2 performance guidelines of US stock technology companies conservative and suppress valuations.

The overseas research team of Everbright Securities said that in the long run, the US corporate tax reduction policy that Trump may introduce is beneficial to the release of corporate profits, and there is strong certainty that demand for AI inference will increase. Currently, US technology companies such as Google, Amazon, and Meta have not shown the will to reduce capital expenditure and slow AI investment. Microsoft (MSFT.US) suspends early data center projects and strategic adjustments in the context of aggressive investment for 24 years. It is recommended to focus on the US stock technology sector's profit recovery and valuation for a long time. Increase space and focus on Nvidia (NVDA.US), Meta (META.US), and TSM.US.

Microsoft's proposal to slow down or suspend construction of some data centers is a strategic adjustment in the context of aggressive investment in 24 years. In April 2025, Microsoft suspended a $1 billion data center project in Ohio. On April 10, the president of Microsoft's cloud computing business said that the AI infrastructure investment strategy requires “agility and refinement,” and that Microsoft is strategically adjusting the pace of investment to slow down or suspend some early projects. As mentioned in the 24Q3 Microsoft performance conference call, capital expenditure is expected to increase quarterly in '25, while capital expenditure in 24Q4 will increase 13% month-on-month, which is higher than the market's agreed expectations. Microsoft's cash capital expenditure in 24Q1-24Q4 accounted for 34.3%, 37.3%, 43.7%, and 70.9% of operating cash flow, respectively, increasing rapidly from quarter to quarter. The global macro environment and AI algorithms and hardware iterations may face greater uncertainty in 2025, but the increase in AI inference load demand is more certain, and Microsoft is expected to maintain the trend of capital expenditure expansion.

Additionally, Microsoft mentioned that AI and non-AI businesses are losing ground and may move some business centers back to traditional cloud businesses in the future. Microsoft mentioned during the 24Q4 performance call that since the sales team needed to balance AI and non-AI businesses, 24Q4 Azure AI revenue exceeded expectations, but Azure's non-AI business revenue was slightly lower than expected. In addition, due to the continuous increase in capital expenditure to support the expansion of AI infrastructure, the gross profit margin of the Microsoft cloud business in 24Q4 was 70%, down 2 pcts from the previous year, and the company expects the gross margin of the 25Q1 cloud business to continue to drop to around 69%. Considering that the AI cloud business has relatively high costs and low profit margin, Microsoft may shift its strategic focus back to the traditional cloud business to cope with the pressure of the macro environment.

At 25M4, Google reaffirmed its 2025 $75 billion capital expenditure plan, highlighting the huge opportunities in the field of AI. On April 9, 2025, the Google CEO emphasized at the Google Cloud Next press conference that it will invest 75 billion US dollars in 2025 to improve the chips and servers needed to improve core products such as search and advertising, while supporting the development of Gemini models and AI product services, in line with the guidelines given at the 24Q4 performance conference (capital expenditure in 2025 is expected to reach 75 billion US dollars, an increase of 43% over the previous year, which is significantly higher than the agreed forecast of 26.9%). The ratio of Google's cash capital expenditure to operating cash flow in 24Q1-24Q4 was 41.6%, 49.5%, 42.5%, and 36.5%, respectively. Capital expenditure had relatively little pressure on free cash flow.

Amazon also emphasized the importance of R&D expenses and capital expenses in the AI field at 25M4, but did not mention specific amounts. According to Reuters, on April 10, 2025, Amazon CEO emphasized continued commitment to AI investments to support products and services such as cloud services, e-commerce generated AI capabilities, and Alexa AI voice assistants. On the 24Q4 results conference call, Amazon mentioned that capital expenditure in 2025 will reach US$105.2 billion, an increase of 34.5% over the previous year. The ratio of Amazon's cash capital expenditure to operating cash flow in 24Q1-24Q4 was 78.6%, 69.7%, 87.1%, and 57.1%, respectively, which is more aggressive than Google's capital expenditure strategy.

Meta is actively guiding capital expenditure in 2025, announcing the construction of a new data center cluster, and a firm AI strategy. According to a Bloomberg report on April 5, 2025, Meta plans to invest $1 billion to build an AI data center in Wisconsin to support Llama series model development and Meta AI inference load, as well as optimize ad recommendation algorithms and short video content ranking algorithms. The 24Q4 Meta results conference mentioned that capital expenditure in 2025 will reach US$60-65 billion, an increase of 52.9%-65.7% over the previous year. At the same time, the company will continue to optimize capital allocation efficiency, including deploying customized ASIC chips to reduce computing costs and extend the life of servers and network equipment. The ratio of 24Q1-24Q4 Meta's cash capital expenditure to operating cash flow was 33.3%, 42.2%, 33.4%, and 51.5%, respectively. This ratio increased significantly in 24Q4.

Trump's tariff policy has brought more uncertainty to the return on investment in AI data centers, the global macroeconomy, and the financing environment. Since Trump proposed an aggressive tariff policy on April 2, the global macroeconomic and technology industry chain has faced greater uncertainty. The AI data center industry chain covers various regions such as the United States, Mexico, mainland China, Taiwan, Southeast Asia, etc., and its return on investment cycle will be widely and complicated, and may be the trigger for US technology companies to strategically adjust capital expenditure.

1) GPU servers: GPU servers are not included in the Trump administration's tariff exemption list, so all GPUs exported from Taiwan to the US are subject to 32% tariffs. However, according to the USMCA, digital processing units exported from Canada and Mexico are eligible for tax exemptions regardless of their place of origin, so US customers can avoid paying GPU server tariffs by re-exporting to Mexico.

2) Data center infrastructure: According to SemiAnalysis estimates, Trump's tariff policy may increase data center infrastructure costs by a single digit, and have a great impact on wafer manufacturing and optical modules. Regarding wafer manufacturing, information disclosed by US builder Exyte shows that US fab infrastructure construction costs about twice that of Taiwan. TSMC founder Zhang Zhongmou said that American-made chips are 50% more expensive than Taiwan, and that wafer manufacturing equipment also relies more on imports from ASML, TEL, etc.; for optical modules, according to the China Commercial Industry Research Institute, more than 80% of the world's optical module production capacity will be located in mainland China in 2024.

From the supply side, due to the long payback period for data center investments and the cost pressure brought about by superimposed tariffs, US technology companies may withdraw some of their investments and pass on costs to customers through price increases. According to the 24Q2 Microsoft performance conference call, Microsoft's investment in AI data centers will be monetized over a period of 15 years and more. The FY25Q2 (24M12-25M2) Micron performance call will mention that the company is expected to pass on the impact of tariffs to customers. Micron has imposed surcharges on memory modules and SSDs on April 9, 2025 to meet the cost increase challenges brought about by tariffs. Cloud vendors are also likely to continue to channel cost pressure downward. From the demand side, downstream customers face wider economic uncertainty, and IT spending and budgets for cloud services will tend to be conservative in the short term.

The increase in cloud service costs brought about by tariffs may weaken the competitiveness of large cloud vendors and make tech giants' capital expenditure plans more cautious. For example, according to the 25M4 DCD interview, data service company StarzData moved cloud workloads to cheaper GCP and OVHcloud after discovering that AWS prices were rising rapidly. Since Microsoft Azure mainly serves large customers, its average price is about 20% higher than AWS. If large cloud vendors pass on the cost pressure brought about by tariffs to customers, the competitive landscape in the public cloud market may become fragmented, making the capital expenditure plans of tech giants such as Microsoft, Amazon, and Google more cautious.

Although digital services in most regions are not affected by tariffs, the EU is still likely to counteract tariffs or strengthen content reviews on digital services, putting pressure on the overall revenue of US technology companies. According to a report in the Financial Times on April 14, the European Commission President is considering taxing digital ads from tech giants such as Google and Meta to deal with Trump's tariff policy. The main restrictions on imported digital services in the EU region are: 1) Digital Market Act (DMA): officially implemented in 24M7, anti-monopoly measures issued by the EU against tech giants to avoid vicious competition; 2) Digital Services Act (DSA): came into effect in 22M11, which emphasizes strengthening platform content review; 3) Digital Services Tax (DST): France began taxing digital services in 19M1, which had a great impact on US internet companies such as Google and Facebook.