Data center capital expenditures witnessed a remarkable 57% growth last year, reaching an unprecedented $726 billion, according to the Dell’Oro Group. The upward trend is anticipated to persist into 2026, with projections indicating a growth rate exceeding 50%, pushing data center expenditures towards the $1 trillion mark—an estimate that was previously set for 2029.
Baron Fung from Dell’Oro emphasizes that this surge is fueled by the escalating competition in the AI sector. As companies invest in more advanced computing architectures, spending is not only escalating in GPU resources but expanding into infrastructure, networking, and storage solutions. The four major cloud providers—Amazon, Google, Meta, and Microsoft—saw a staggering 76% increase in their data center capex.
Recent earnings reports further indicate this trend will continue. Amazon invested $131 billion in capital expenditures in 2025, primarily for data center enhancements, with expectations of this figure climbing to approximately $200 billion in 2026 due to substantial demand, particularly in AWS.
Amazon’s backlog now stands at $244 billion, representing a 40% increase year-over-year, highlighting a significant demand for AWS services. Simultaneously, Google plans to allocate around $180 billion in capex for 2026, reporting a backlog of $240 billion, with a notable increase in multi-billion dollar deals over the past year.
This growth trajectory is largely attributed to increased AI investments. Companies in the AI sector are striving for more compute power to train advanced models, while businesses implementing AI technologies are rapidly expanding their inference capabilities. A recent survey by the Boston Consulting Group indicates that businesses are poised to double their AI spending this year, with over 90% of CEOs committed to maintaining or increasing their AI investment levels.
However, this explosion in data center spending has accompanying challenges. Enterprise businesses are encountered with rising hardware costs, particularly for memory, which can constitute nearly half of a server’s total cost. High demand for memory driven by hyperscalers contributes to increased commodity prices, leaving smaller enterprises struggling to compete.
As a result, many enterprises are forced to curtail their hardware purchases, opting to extend the life of existing servers or pivoting to cloud infrastructure rather than investing in on-premises solutions. There’s speculation that the hyperscalers may be engaging in a strategy of acquiring vast amounts of memory to raise prices and boost cloud adoption, although this remains unsubstantiated.
Fung suggests that companies contemplating capital expenditures for AI infrastructures should first analyze their cloud capabilities to ensure they can effectively utilize AI resources without incurring excess idle time, which could adversely affect returns on investment.
