Skip to main content

Data centre capex to hit $1.7tn by 2030 as AI infrastructure enters new expansion phase

Data centre image representing accelerated spending driven by hyperscalers, neo clouds, and sovereign AI initiatives

Dell’Oro forecasts accelerated spending driven by hyperscalers, neo clouds, and sovereign AI initiatives (Credit: Volodimir Zozulinskyi / Shutterstock.com)

Dell’Oro forecasts accelerated spending driven by hyperscalers, neo clouds, and sovereign AI initiatives.

The multi-year AI expansion cycle is projected to drive worldwide data centre capital expenditure (capex) to $1.7 trillion by 2030, according to a recently published forecast from Dell'Oro Group. 

Hyperscale and neo cloud service providers, along with sovereign AI initiatives, are entering a new phase of infrastructure expansion that is outpacing earlier industry projections. Global data centre capex is now expected to approach $1tn in 2026 alone, reaching this major industry milestone sooner than anticipated.

Hyperscale dominance and the rise of neo clouds

The growth is being led by the top four US hyperscale cloud service providers: Amazon, Google, Meta, and Microsoft; which are entering 2026 with combined data centre capital expenditures approaching $600 billion. According to Baron Fung, senior research director at Dell'Oro Group, these companies continue to invest aggressively despite increased scrutiny around AI infrastructure returns.

"Hyperscalers are supported by large cash reserves and a long-term focus on market share," Fung said. "This growth is being driven by the deployment of larger and more complex AI clusters, which are increasing demand for high-performance networking, storage, inference capacity, and advanced power and cooling infrastructure."

Technical implications for optical networking

The findings have direct implications for optical networking vendors. As clusters scale, networking is shifting from a supporting role to a primary capex driver, with high-speed interconnects now representing a significant double-digit percentage of the total cluster cost. 

Diversification of AI infrastructure spending

Beyond the top hyperscalers, AI model builders, neo cloud providers, and sovereign cloud initiatives are accelerating their own data centre deployments. Dell'Oro projects that accelerated servers for AI training and domain-specific workloads could account for approximately two-thirds of total data centre infrastructure spending by 2030.

The transition to 1.6T and co-packaged optics

The transition from 800G to 1.6T optical modules is accelerating as a result, while co-packaged optics (CPO) and linear pluggable solutions are being evaluated for scale-up applications requiring extreme bandwidth density and power efficiency.

While the top four US hyperscalers are expected to represent about half of global data centre capex by 2030, emerging AI model builders and neo cloud service providers are projected to grow at significant rates. However, investment outside the hyperscale segment remains more constrained. Enterprise data centre spending continues to face headwinds from tariffs, monetary policy uncertainty, and unclear returns on AI investments.

The Dell'Oro Group Data Centre IT Capex 5-Year Forecast Report provides historical data from 2014 and includes tables covering data centre and server capex, along with server unit shipments forecasts segmented by customer type and infrastructure equipment categories.
 

Media Partners