Worldwide data centre capital expenditure (capex) increased 59% year-on-year in the third quarter of 2025, marking the eighth consecutive quarter of double-digit growth. This surge comes as artificial intelligence (AI) investment expands beyond initial training to large-scale production deployments.
According to Dell'Oro Group's latest quarterly report, the sustained growth reflects both accelerated compute spending and broader infrastructure investment as cloud providers scale AI capabilities for inference-heavy workloads.
Major cloud providers raise spending
The four largest US cloud service providers, Amazon, Google, Meta, and Microsoft, have continued to raise their capex expectations for 2025. This is supported by increased investment in both AI and general-purpose infrastructure. Notably, Oracle is on track to double its data centre capex this year as it expands capacity for its ambitious Stargate project.
"What is notable this cycle is not just the pace of spending, but the expanding scope of investment," said Baron Fung, senior research director at Dell'Oro Group. "Cloud providers are simultaneously scaling accelerated compute, general-purpose servers, and the supporting infrastructure required to deploy AI at production scale."
Accelerated server spending surged in the quarter, driven by the ramp of NVIDIA’s Blackwell Ultra platform across US hyperscalers, neo-cloud providers, and sovereign AI deployments. This has also triggered a record rise in Ethernet data centre switch sales, as operators rush to build the high-speed back-end networks required for these clusters.
General-purpose server demand strengthens
While AI remains the headline, demand for general-purpose servers also strengthened during the quarter. Cloud providers are expanding compute and storage capacity to support traditional cloud services alongside AI inference tasks.
In the OEM market, Dell led in revenue, followed by HPE and Lenovo. These vendors gained share through strong shipments of NVIDIA-based systems and a refreshed x86 portfolio. However, white-box vendors continue to capture the majority of shipments, particularly within hyperscale AI deployments.
Efficiency and infrastructure transitions
As spending hits record levels, cloud service providers are placing increased emphasis on capital discipline. This involves optimising asset depreciation and lifecycles to maintain cash flow. For the optical communications industry, the report highlights a significant transition in physical infrastructure. To keep pace with AI-era power requirements, there is a rapid shift toward direct liquid cooling (DLC) and 800G Ethernet architectures. DLC is on pace to exceed $2 billion in yearly revenue as adoption broadens across large clusters.
The global outlook for data centre capex has been raised through 2026, reflecting sustained hyperscale investment and the continued prioritisation of AI-ready infrastructure.
