According to the latest AI Server industry research from TrendForce, cloud service providers (CSPs) are accelerating the deployment of related infrastructure to meet the surging demand for AI applications. TrendForce estimates that global data center power demand capacity will reach 161GW in 2026, an annual increase of approximately 31%, with AI Servers being the primary growth driver and accounting for an estimated 33.4% of total demand capacity.
TrendForce categorizes data center power demand capacity into four main sources: General Servers, AI Servers, other IT equipment, and non-IT equipment such as cooling and power supply systems. Observing recent changes in global data center power demand capacity, it reached 122.9GW in 2025 and is estimated to rise to 161GW in 2026. With CSPs continuing to expand capital expenditures and overall Server shipments not declining, it is projected that the growth rate of data center power demand capacity will remain above 30% in 2027.
TrendForce notes that although improvements in data center electromechanical efficiency have reduced the load driven by per-unit IT power consumption, the rapid deployment of IT equipment led by AI Servers, coupled with the simultaneous growth in total demand capacity for non-IT equipment, is prompting data centers to begin planning the adoption of High-Voltage Direct Current (HVDC) power supply to optimize the efficiency of the overall power distribution chain.
Looking at the composition of total data center power demand capacity, the share of AI Servers has grown from nearly 25% in 2025 to approximately 33.4% in 2026, and is expected to potentially surpass 40% in 2027. In contrast, the share of General Servers is projected to decline from around 40% in the past to 25.6% by 2027.
Global data center power supply-demand gap estimated to reach 268GW by 2030
While data center power demand continues to grow, the pace of grid power supply expansion has not kept up. According to TrendForce estimates, before 2025, grid deliverability was still able to meet the growth in data center power demand capacity. However, starting from 2026, supply and demand begin to diverge, with the gap widening significantly after 2028. By 2030, global data center power demand capacity is estimated to reach 490.7GW. Assuming other power demand is met, the capacity available to data centers from the grid would be 222.6GW, resulting in a supply-demand gap of 268GW.
TrendForce's analysis suggests this supply-demand gap should be understood on two levels. First, data center power demand can be met not only by grid supply but also through behind-the-meter self-generation by operators. However, this portion is not included in the grid capacity available for data centers, so this estimation method does not reflect its role in filling the gap, causing the book gap to be higher than the actual shortage. Second, delays in grid interconnection and transmission and distribution construction can lead to actual power supply shortages.
Taking the United States, which currently has the highest data center deployment, as an example, it is estimated that before 2028, a significant portion of data center power consumption is supplied by operators' self-generation and existing facilities, allowing most nationally scheduled grid projects to come online on time. However, starting in 2028, grid delivery delays are expected to become more pronounced, becoming a major factor in the widening supply-demand capacity gap for data centers.
Based on current grid data, it is estimated that by 2030, the data center power supply-demand capacity gap in the United States will exceed 170GW. Whether this gap can be narrowed in the future through behind-the-meter self-generation supply and improvements in grid supply conditions remains to be seen.