Dive Brief:
- Data center capital expenditures increased 53% year over year to $134 billion during the first three months of 2025, according to Dell’Oro Group research published Tuesday. The spike was driven by a surge in hyperscaler spending on AI infrastructure, particularly Nvidia Blackwell GPUs and custom accelerators, the research firm said.
- The four companies with the largest cloud footprints — AWS, Google, Meta and Microsoft — accounted for 44% of Q1 data center capital investments. Enterprise infrastructure spending — the second biggest category — accounted for one-third of the total, according to Dell’Oro.
- “Despite some project cancellations by U.S. cloud providers, overall CapEx remains on track, with hyperscalers adjusting capacity rather than cutting investments,” Dell’Oro Group Senior Research Director Baron Fung said in the report. “Enterprises, facing tighter budgets and tariff-related risks, are more cautious, prompting slight downward revisions to their CapEx forecasts.”
Dive Insight:
An ongoing data center building boom sparked by generative AI adoption showed no signs of an early-year slowdown as hyperscalers raced to add compute capacity.
Amazon reported $24.3 billion in Q1 capital expenditures, primarily to expand AWS’ AI cloud infrastructure. Microsoft and Google weren’t far behind, reporting $21.4 billion and $17 billion in CapEx for the first three months of the year, respectively.
Dell’Oro Group expects the trend to continue. The firm forecasted a 30% year-over-year bump in data center investments for 2025, despite mixed economic signals triggered by tariff concerns and supply chain challenges. Last year, data center infrastructure capital expenditures grew 51% year over year to $455 billion.
“Tariff-related uncertainties are not expected to materially alter hyperscaler spending plans given their diversified global supply chains,” Fung said in the report.
AWS, Microsoft and Google Cloud plan to invest more than $250 billion in buildouts this year, in part to ease capacity constraints and satisfy growing customer demand for AI processing power.
“The availability of GPU and supporting infrastructure is supply constrained,” Fung said in an email. “Demand is so strong that the top 4 U.S. cloud service providers have had to turn away smaller customers.”
Amazon signaled two massive U.S. data center construction projects earlier this month — a $20-billion hub in Pennsylvania and a $10-billion project in North Carolina. The company is also planning a nearly $13 billion buildout in Australia, according to a June 14 announcement.
“As fast as we actually put the capacity in, it’s being consumed,” Amazon CEO Andy Jassy said during a May earnings call. The executive noted that demand for AI compute services is “unlike anything we’ve seen before” and characterized AI as a “once-in-a-lifetime reinvention of everything we know” in an April letter to shareholders.
AWS and its hyperscaler peers are jockeying to establish early market share, Fung told CIO Dive. Realizing a return on the investments will take time, he said.
