From Silicon to Cooling: Dell’Oro Maps the AI Data Center Buildout

For much of the past decade, data center growth could be measured in incremental gains: another efficiency point here, another capacity tranche there. That era is over. According to a cascade of recent research from Dell’Oro Group, the AI investment cycle has crossed into a new phase, one defined less by experimentation and more by industrial-scale execution.

Across servers, networks, power, and cooling, Dell’Oro’s latest data points to a market being reshaped end-to-end by AI workloads which are pulling forward capital spending, redefining bill-of-material assumptions, and forcing architectural transitions that are rapidly becoming non-negotiable.

Capex Becomes the Signal

The clearest indicator of the shift is spending. Dell’Oro reported that worldwide data center capital expenditures rose 59 percent year-over-year in 3Q 2025, marking the eighth consecutive quarter of double-digit growth. Importantly, this is no longer a narrow, training-centric surge.

“The Top 4 US cloud service providers—Amazon, Google, Meta, and Microsoft—continue to raise data center capex expectations for 2025, supported by increased investments in both AI and general-purpose infrastructure,” said Baron Fung, Senior Research Director at Dell’Oro Group. He added that Oracle is on track to double its data center capex as it expands capacity for the Stargate project.

“What is notable this cycle is not just the pace of spending, but the expanding scope of investment,” Fung said. Hyperscalers are now scaling accelerated compute, general-purpose servers, and the supporting infrastructure required to deploy AI at production scale, while simultaneously applying tighter discipline around asset lifecycles and depreciation to preserve cash flow.

The result is a capex environment that looks less speculative and more structural, with investment signals extending well into 2026.

Accelerators Redefine the Hardware Stack

At the component level, the AI effect is even more pronounced. Dell’Oro found that global data center server and storage component revenue jumped 40 percent year-over-year in 3Q 2025, driven overwhelmingly by AI accelerators and their cascading dependencies.

“Growth in the quarter was led by NVIDIA’s Blackwell Ultra ramp across US hyperscalers and neo cloud providers,” Fung said, citing not only GPUs but the associated demand for HBM, back-end NICs, CPUs, and storage. Custom accelerators from hyperscalers such as Google and Amazon also surged, driven by both training and inference requirements tied to foundational models and large enterprise AI agreements.

Dell’Oro now forecasts the server and storage systems component market to grow 48 percent in 2025, with high-end accelerators remaining the dominant growth driver into 2026 as platforms such as NVIDIA’s Vera Rubin and AMD’s MI400 come online.

This, in turn, is reshaping expectations for rack density, power delivery, and thermal management, pressures that ripple outward into every other layer of the data center.

Two Networks, One Data Center

Nowhere is that ripple effect more visible than in networking. Dell’Oro reports that Ethernet data center switch revenue exceeded $8 billion in 3Q 2025, more than doubling from three years ago, a remarkable result given the size and maturity of the market.

“While the surge in Ethernet switch sales was primarily fueled by accelerated adoption in AI back-end networks, front-end network sales were also strong,” said Sameh Boujelbene, Vice President at Dell’Oro Group. As AI clusters scale, additional front-end capacity is required simply to feed them.

The data underscores a structural bifurcation: front-end networks still account for well over half of total Ethernet switch sales, dominated by 100/200/400 Gbps speeds, while 800 Gbps shipments are concentrated almost entirely in AI back-end networks.

That back-end is also undergoing a critical transition. Ethernet now accounts for more than two-thirds of switch sales in AI clusters, up sharply from less than half a year ago. While InfiniBand continues to grow, as boosted by NVIDIA’s Blackwell Ultra ramp, Boujelbene noted that Ethernet growth is outpacing InfiniBand, driven by demand from large AI cluster builders including hyperscalers and emerging AI infrastructure providers.

The implication is clear: AI networking is no longer a niche overlay but a parallel, first-class architecture inside the modern data center.

Physical Infrastructure Catches Up

As compute and networks race ahead, physical infrastructure is no longer a quiet follower. Dell’Oro reports that the Data Center Physical Infrastructure (DCPI) market grew 18 percent year-over-year in 3Q 2025, with North America up 26 percent, led by hyperscaler AI buildouts.

“Operators are scaling capacity at unprecedented speed, and both power and cooling architectures are evolving to keep pace with AI-era requirements,” said Alex Cordovil, Research Director at Dell’Oro Group.

Thermal management stands out. Direct liquid cooling expanded 85 percent year-over-year and is now on pace to exceed $2 billion in annual revenue, while higher-voltage distribution and emerging DC power topologies are beginning to shape next-generation facility designs.

The competitive response has been swift. Dell’Oro highlights accelerated consolidation, including acquisitions by Eaton, Vertiv, and Daikin, alongside rapid growth from newer entrants such as Aaon and Modine as hyperscalers broaden their supplier bases.

Open standards are also gaining traction, with new specifications introduced through the Open Compute Project helping to accelerate qualification cycles and ecosystem maturity.

Liquid Cooling Crosses the Line

If there is a single inflection point that ties Dell’Oro’s research together, it is cooling. According to the firm’s latest forecasts, the data center liquid cooling market is expected to approach $7 billion in annual revenue by 2029, roughly doubling in 2025 alone.

“Liquid cooling has crossed a critical threshold,” Cordovil said. “What was once treated as an optional efficiency upgrade is now a functional requirement for large-scale AI deployments.”

Dell’Oro projects that leading-edge GPU TDPs will exceed 4,000 watts by 2029, pushing air cooling beyond practical limits. As a result, single-phase direct liquid cooling has consolidated its position as the dominant architecture, supported by hyperscaler experience and accelerating vendor investment.

While two-phase and immersion cooling continue to advance, Dell’Oro expects their adoption to remain more selective in the near term, i.e. focused on pilots and specialized deployments, until thermal flux reaches levels that force broader architectural shifts.

The Numbers Behind the Shift

Across Dell’Oro’s recent research, a consistent quantitative pattern is emerging; one that helps ground the AI narrative in hard deployment and spending signals rather than isolated headlines.

Taken together, Dell’Oro’s recent releases point to an AI-driven infrastructure cycle that is not only accelerating, but broadening across every layer of the data center. Several data points help quantify the scale and durability of that shift:

• Capital spending is no longer episodic. Global data center capex rose 59% year-over-year in 3Q 2025, marking the eighth straight quarter of double-digit growth, with Dell’Oro raising its growth outlook through 2026; a signal that AI-driven investment is becoming structurally embedded rather than front-loaded.
• Component demand is compounding, not substituting. Server and storage component revenue grew 40% Y/Y in 3Q 2025 and is forecast to rise 48% for full-year 2025, reflecting not only AI accelerator ramps but parallel strength in CPUs, NICs, HBM, and storage required to support both training and inference workloads.
• AI networking is now a volume market. Ethernet data center switch revenue exceeded $8 billion in a single quarter, more than doubling in three years, with AI back-end networks accounting for over two-thirds of switch sales in AI clusters: an inflection point that confirms Ethernet’s growing dominance alongside, and increasingly ahead of, InfiniBand.
• Physical infrastructure growth is keeping pace with IT. The Data Center Physical Infrastructure market expanded 18% Y/Y globally and 26% in North America, with hyperscalers driving more than 30% growth in the Top 10 Cloud segment, evidence that power and cooling are no longer lagging indicators.
• Liquid cooling is crossing from adoption to standardization. Direct liquid cooling grew 85% Y/Y within DC physical infrastructure and is on track to exceed $2 billion in annual revenue, while the broader liquid cooling market is forecast to approach $7 billion by 2029, underpinned by projected GPU thermal design power exceeding 4,000 watts later in the decade.
• Supply chains are consolidating and professionalizing. Accelerated M&A activity, paired with rapid share gains by emerging vendors, signals that liquid cooling and high-power infrastructure are shifting from bespoke engineering exercises to scalable, repeatable product markets.

Together, these metrics do more than describe growth. They define the operating context the industry now has to plan against.

What This Means for 2026 and Beyond

With those signals in view, Dell’Oro’s research suggests the AI buildout has entered a new chapter. The question is no longer whether AI will drive data center growth, but how quickly the industry can align silicon, networks, power, and cooling into repeatable, financeable systems.

Capex is accelerating, but so is discipline. Hardware is densifying, but so is infrastructure complexity. And technologies once treated as optional, liquid cooling chief among them, are rapidly becoming baseline requirements.

As the industry moves deeper into 2026, the winners will be less defined by headline megawatts and more by execution: standardized designs, resilient supply chains, and infrastructure stacks capable of scaling AI from prototype to production, again and again.

At Data Center Frontier, we talk the industry talk and walk the industry walk. In that spirit, DCF Staff members may occasionally use AI tools to assist with content. Elements of this article were created with help from OpenAI's GPT5.

Keep pace with the fast-moving world of data centers and cloud computing by connecting with Data Center Frontier on LinkedIn, following us on X/Twitter and Facebook, as well as on BlueSky, and signing up for our weekly newsletters using the form below.