Key Highlights

• Between 2026 and 2030, nearly 100 gigawatts (GW) of new data center capacity will be added globally, doubling the total global capacity. By 2030, the global data center industry is projected to expand at a compound annual growth rate (CAGR) of 14%, which will necessitate energy innovations to alleviate grid constraints. Hyperscalers will continue to be the primary drivers of industry growth, while simultaneously pursuing a dual strategy of both leasing and building their own facilities.
• By 2030, AI could account for half of all workloads, with inference emerging as the primary driver. In 2025, AI will account for only about one-quarter of all data center workloads, with the majority of demand coming from training. However, a significant shift is expected in 2027: inference workloads may surpass training to become the primary driver of AI demand.
• The data center industry is currently undergoing a supercycle of infrastructure investment, with potential funding requirements reaching as high as US$3 trillion by 2030. It is projected that approximately 100 GW of new capacity will come online between 2026 and 2030, generating US$1.2 trillion in real estate asset value. Users may also incur additional expenditures of US$1 to 2 trillion to equip their data centers with IT equipment.

Regional Landscape: The Americas Lead, While Asia-Pacific and EMEA Expand Steadily

It is projected that the data center industry will add 97 GW of capacity between 2025 and 2030, doubling in size over the five-year period. By 2030, global data center capacity could reach 200 GW. This rapid growth is primarily driven by the expansion of hyperscale cloud services and increasing demand for AI.

The Americas represent the largest data center regional market, accounting for approximately 50% of global capacity. At the same time, it is the fastest-growing of the three major global regions. By 2030, the region’s supply is projected to achieve a compound annual growth rate (CAGR) of 17%, thereby maintaining its position as the dominant data center hub. The United States drives the bulk of activity in the region, accounting for roughly 90% of total Americas capacity.

By 2030, data center capacity in the Asia-Pacific (APAC) region is projected to expand from 32 GW to 57 GW, representing a compound annual growth rate (CAGR) of 12%. Among these, colocation data centers are expected to lead the growth at a CAGR of 19%, while on-premises data center capacity is forecast to decline by 6% as enterprises continue their migration to the cloud.

Europe, the Middle East, and Africa (EMEA) is projected to achieve a compound annual growth rate (CAGR) of 10%, driven by government support for AI infrastructure and growing demand for sovereign AI clouds that comply with data privacy regulations. The region is expected to add 13 GW of new capacity, with growth concentrated in mature European hub markets and emerging Middle Eastern markets advancing digital transformation.

2027: A Critical Inflection Point—AI Inference Workloads Will Surpass Training

Despite the rapid growth in daily active AI users, AI will account for only about one-quarter of data center workloads in 2025, with demand primarily driven by training tasks. However, a major shift is expected in 2027, when inference workloads are projected to surpass training and become the primary driver of AI demand.

AI models typically represent a one-time or recurring investment. However, once the model is deployed, inference-driven applications continue to generate revenue over time. Going forward, each deployment of an AI model will create ongoing inference demand, which is expected to grow as adoption increases.

However, this growth hinges on the emergence and rapid adoption of reasoning applications that have not yet achieved large-scale deployment. To meet the demand for reasoning while minimizing latency and delivering efficient user service, these applications must be deployed in a geographically distributed manner, thereby driving the development of regional deployments and edge-embedded systems.

Grid Bottlenecks Drive Transformation: Data Centers Shift to Self-Powered Operations and Energy Storage

With average lead times for grid connections in major data center markets now exceeding four years, data center operators are expected to ramp up behind-the-meter power supply strategies and explore the deployment of energy storage systems.

In the United States, natural gas is expected to play a significant role in alleviating grid constraints—both as a temporary transitional power source and, increasingly, as a permanent on-site generation solution. The surge in global turbine orders underscores this trend. However, it is worth noting that some large data centers remain cautious about natural-gas-based solutions, as they believe such approaches do not align with sustainability goals.

As a solution, natural gas plays a less prominent role in the EMEA and APAC regions, where the adoption of renewable energy sources such as solar and wind power is on the rise. For instance, in EMEA, projects that integrate renewable energy with dedicated transmission lines can reduce tenants’ electricity costs by up to 40% compared with grid-supplied power.

In the Asia-Pacific, Europe, Middle East, and Africa regions, natural gas plays a relatively minor role as a solution. In these markets, the adoption of renewable energy sources such as solar and wind power is on the rise. For instance, in the EMEA region, projects that combine renewable energy with dedicated transmission lines can reduce customers’ electricity costs by as much as 40% compared with grid-supplied power.

Due to delays in grid connection for utility-scale power grids, some data center operators are shifting from power purchase agreements (PPAs) to direct investment in building their own power-generation facilities. Moreover, numerous markets—such as Ireland and Texas—have adopted “bring your own power” policies, further accelerating this trend.

Costs Continue to Rise: Construction Costs Reach $11.3 Million per Megawatt

The data center industry is expanding at an astonishing pace, leading to extended delivery cycles, a shortage of skilled technical personnel, and rising development costs.

Between 2020 and 2025, the global average construction cost for data centers increased from USD 7.7 million per megawatt (MW) to USD 10.7 million per MW, representing a compound annual growth rate of 7%. JLL Research forecasts that by 2026, the global average cost will rise by another 6%, reaching USD 11.3 million per MW.

Speed to power is the primary factor in site selection, followed by community support, regulatory delays, and proximity to customers. However, as data center projects scale up, the relative importance of construction cost differences in the site-selection decision may increase.

Note: The figures in the chart reflect only the construction costs for the data center’s shell and core. Users typically bear the costs of technical fit-out, while AI infrastructure can cost as much as US$25 million per megawatt.

Investment Supercycle: Total Data Center Spending Approaches $3 Trillion

Over the next five years, the global data center industry is projected to grow at a compound annual growth rate (CAGR) of 14%, potentially adding 100 GW of new capacity across hyperscale, colocation, and on-premises facilities. This would translate into $1.2 trillion in real estate asset value and require approximately $870 billion in additional debt financing.

These figures do not include the $1 to $2 trillion that users will spend on configuring GPU and networking infrastructure. This implies that total data-center spending over the next five years could approach $3 trillion. Taken together, it is safe to say that we are currently in a supercycle of infrastructure investment.

The scale of new data center projects is steadily increasing, driving up construction costs. Consequently, the industry will remain in a state of consolidation due to high development expenses and the increasingly complex technologies required for the design, construction, and operation of modern data centers.

These continually rising entry barriers are curbing speculative activity in development projects and accelerating the advancement of viable projects backed by reputable developers. For such firms, the debt financing market will remain open.

Looking Ahead

Currently, the data center industry is experiencing the largest infrastructure investment supercycle in its history. The high degree of interconnectivity among data centers is driving AI-powered expansion that is reshaping multiple sectors, including energy, technology, and real estate.

The shift from AI training to inference will reallocate workloads from centralized clusters to distributed regional hubs, fundamentally transforming capacity planning and geographic deployment strategies.

Energy infrastructure has become a critical bottleneck constraining expansion. Grid constraints are now threatening the industry’s growth trajectory, making behind-the-meter generation and integrated battery energy storage solutions essential pathways for sustainable scaling.

Investors and developers must strike a balance between speed to market and capital efficiency, all while navigating supply-chain constraints and evolving demand patterns. Industry leaders must translate these competing forces into a sustainable competitive advantage. In this investment supercycle, the true winners will be companies that can anticipate demand inflection points ahead of time while remaining agile enough to adapt to the continuous evolution of AI models and use cases.