Dell’Oro Group expects the global data center physical infrastructure market to reach $120 billion by 2030, with artificial intelligence shifting from a rapid growth cycle into a sustained source of infrastructure investment. The firm’s latest forecast projects manufacturer revenue to grow at a 22% compound annual rate from 2025 through 2030. The market could add nearly 200 GW of data center capacity during that period, while new IT capacity will account for most of the expansion rather than a major rise in infrastructure spending per megawatt.

Higher-density computing will also reshape how operators allocate infrastructure spending inside data centers. Liquid-cooled architectures and denser racks will shift investment across thermal, power and supporting infrastructure categories, reducing the impact of infrastructure content per megawatt on overall market growth. Dell’Oro expects annual net capacity additions to post their strongest year-over-year growth in 2026 before growth moderates through the rest of the decade. Despite that moderation, annual capacity growth should remain in double-digit territory through 2030.

“The AI buildout has moved past the point where it can be treated as a surge. It is now the baseline against which the rest of the market is measured,” says Alex Cordovil, Research Director at Dell’Oro Group. “What has changed in this forecast is where the risk sits. Demand is no longer the open question, delivery is. Equipment lead times, construction labour, grid interconnection and community consent all remain constrained, especially with the first statewide data centre moratorium now in force.”

Hyperscalers remain the largest infrastructure buyers

Hyperscalers should remain the largest individual contributors to DCPI revenue through 2030, although their growth rate should moderate from the exceptional pace recorded in 2025 and 2026. Colocation providers will continue to play an important role as hyperscalers increasingly use third-party facilities to accommodate workloads, particularly in markets outside the U.S. Powered shell developments are also reshaping infrastructure procurement as tenants increasingly take responsibility for equipment purchases. As a result, revenue can shift between customer categories without necessarily changing the underlying occupancy of the data center building.

Dell’Oro has also created a separate category for AI-specialized cloud providers in its forecast. The group includes neoclouds and AI model builders and could become one of the fastest-growing areas in the research. Their emergence reflects a broader change in the data center customer base as companies build businesses around AI compute without operating like traditional hyperscalers. Enterprise demand should continue growing, but at a slower pace than the wider market.

North America leads as EMEA faces tighter constraints

North America is projected to lead capacity additions through 2030, with China ranking second. The regional outlook becomes more complicated beyond those two markets, as power availability, permitting and local acceptance increasingly determine which proposed facilities can progress. EMEA is the only region that Dell’Oro revised downward from its January forecast, reflecting slower access to power and a more difficult permitting environment. Community resistance has also become a material constraint, delaying or blocking a meaningful share of announced projects. 

The regional picture reinforces a broader point about the next phase of AI infrastructure. Capital availability and customer demand may support large project pipelines, but those factors alone cannot guarantee construction. Developers still need equipment, skilled construction labor, grid connections and approvals to convert announced capacity into operational capacity. The market’s next bottleneck therefore sits increasingly in execution rather than demand.

Delivery becomes the defining data center challenge

Dell’Oro’s forecast ultimately describes an industry entering a more disciplined phase of expansion. AI will continue to drive capacity additions, but the infrastructure required to support that growth is becoming more specialized, more power-intensive and more dependent on local constraints. Cooling technologies, higher-capacity UPS systems, medium-voltage designs and emerging solid-state transformer architectures will increasingly influence how developers configure facilities. Meanwhile, hyperscalers, colocation operators and AI-focused cloud providers will compete for the physical capacity needed to support the next generation of workloads.

The $120 billion projection therefore represents more than a market-size milestone. It captures a structural shift in which data center infrastructure becomes a central investment category for the AI economy, while delivery constraints determine how much of the projected demand can actually become operational capacity. The winners will likely be suppliers and developers that can align power, cooling, equipment availability and construction schedules rather than simply secure large customer commitments. For the industry, the question through 2030 is becoming less about whether AI needs more data centers and more about who can build the infrastructure fast enough to support it.