Evercore ISI raises the forecast for U.S. AI cloud computing capital expenditures and points out potential beneficiaries in U.S. stocks
Evercore ISI raised its forecast for U.S. cloud computing capital expenditures to a year-on-year growth of 44% by 2025, primarily driven by investments in artificial intelligence and cloud infrastructure. In the first quarter, capital expenditures of core U.S. hyperscale companies accelerated to $81 billion, a year-on-year increase of 71%. Capital expenditures from tech giants such as Google, Microsoft, Meta, and Amazon have all seen significant growth and are expected to continue driving future investment waves
According to the Zhitong Finance APP, Evercore ISI released a research report indicating that total capital expenditures of core hyperscale companies in the United States accelerated growth in the first quarter, reaching approximately $81 billion, a year-on-year increase of 71%. Driven by ongoing investments in artificial intelligence and cloud infrastructure, all major hyperscale companies saw significant year-on-year growth in capital expenditures in the third quarter. More importantly, Evercore ISI raised its forward-looking estimate for year-on-year growth in U.S. cloud computing capital expenditures in 2025 to 44% (previously expected at 38%), as hyperscale companies continue to increase investments in artificial intelligence infrastructure.
Additionally, Evercore ISI believes that the upward revision of capital expenditure forecasts should help alleviate any recent concerns about a pause/slowdown in data center capacity demand. The agency still believes that generative artificial intelligence could trigger a sustained wave of investment for years to come, driven by today's model training expansion use cases (from inference and other AI workloads). The expansion/expectation of capital expenditures is primarily driven by hyperscale tech companies such as Google (GOOGL.US), Microsoft (MSFT.US), Meta (META.US), Amazon (AMZN.US), and Oracle (ORCL.US).
Google: Capital expenditures in the first quarter increased by 43% year-on-year, driven by servers, followed by data center investments. Google still expects to spend $75 billion in 2025 (approximately a 43% year-on-year increase).
Microsoft: Capital expenditures in the first quarter increased by 53% year-on-year, and the company expects capital expenditures to continue to grow in the second quarter (indicating a significant increase in second-quarter capital expenditures); the company expects capital expenditures for the fiscal year 2026 (ending in June) to continue to grow, although at a slower pace.
Meta: Capital expenditures in the first quarter increased by over 100% year-on-year, and they raised their capital plan for 2025 from $60-65 billion to $64-75 billion.
Amazon: Capital expenditures in the first quarter increased by 72% year-on-year to $25 billion, primarily driven by technology infrastructure.
Oracle: Capital expenditures in the first quarter more than doubled, and it raised its capital expenditure guidance for fiscal year 2025 (ending in May) to $16 billion (more than double year-on-year).
Generative artificial intelligence is the incremental driver of data center spending: Generative artificial intelligence will create meaningful tailwinds for the broader data center ecosystem. From a technology infrastructure perspective, hyperscale companies will need to invest in GPU servers and auxiliary hardware (management servers/storage, networking) to run AI workloads, benefiting OEMs such as Arista Networks (ANET.US), Juniper Networks (JNPR.US), and Cisco (CSCO.US). Higher power/cooling density requirements will drive demand for suppliers like Vertiv Holdings (VRT.US). Finally, the connection-intensive model training clusters will drive demand for connectors/cables/fiber optics (Amphenol (APH.US), Ciena (CIEN.US), CommScope (COMM.US))
