Resin Materials for AI Data Center Components

Data centers are being developed around the world as critical infrastructure supporting cloud services, video streaming, generative AI, and more. At the same time, growing computing demand for AI is causing data center power consumption to increase rapidly. The IEA projects that global data center electricity consumption was approximately 415 TWh in 2024 and will reach approximately 945 TWh by 2030.^*1　
The expansion of high-speed, high-density computing facilities for AI is cited as a major factor. As a result, data centers are increasingly being positioned not merely as ICT facilities, but as social infrastructure that must be considered together with power infrastructure.
National policies are also advancing in response to this growing demand. The focus is shifting from simply promoting the installation of data centers to ensuring the stable acceptance of large amounts of power, efficient energy use, and sustainable operation.​^​​
In the EU, under the Energy Efficiency Directive, reporting obligations related to energy performance and water footprint have been introduced for data centers above a certain scale. The European Commission is also considering a common database, a rating system, and minimum performance standards.^*2　
In other words, future data centers will be evaluated not only by their computing capability, but also by how effectively they manage power, water, and environmental impact. ^​​​​

*1：Energy and AI – Analysis – IEA*2：Energy performance of data centres – Energy – European Commission​​​​

In conventional data centers, it was common to install large numbers of relatively standard servers and manage the entire room with HVAC equipment. In recent years, however, the introduction of AI servers equipped with large numbers of GPUs has significantly increased both power density and heat generation per rack.
The IEA identifies accelerated computing servers for AI as accounting for a large share of the increase in data center power demand through 2030, while also noting that the resulting demand for cooling and other infrastructure cannot be ignored.^*3　
Data center evolution is therefore driven less by simply adding more servers and more toward increasing the amount of power and heat handled by each rack and system.​
For this reason, equipment inside data centers is shifting from a simple arrangement of IT equipment to a configuration that combines systems for stable power delivery, redundant systems to avoid downtime during outages or voltage dips, and mechanisms for efficiently removing generated heat.​
The DOE design guide also frames data center design as the integrated optimization of IT equipment, HVAC, cooling, and electrical systems. It presents a comprehensive design approach that includes electrical systems such as PDUs and UPSs, airflow management, and liquid cooling equipment.^*4

*3：Energy and AI – Analysis – IEA*4：Best Practices Guide for Energy-Efficient Data Center Design

In data centers, preventing power interruption is even more critical than in general building facilities. Therefore, power supply design based on redundant configurations such as 2N and N+1 is essential. Such designs not only convert and distribute incoming power efficiently to each facility, but also enable continued operation during faults or outages.
The DOE design guide describes a data center power system as a configuration that combines utility power, backup generators, UPSs, PDUs, and related equipment. It also presents an approach that incorporates multiple UPSs and PDUs to ensure redundancy.^*5　
Furthermore, because UPS efficiency decreases at low load factors, using multiple smaller UPSs or modular UPSs can be effective in maintaining appropriate load factors.^*5,6　ENERGY STAR also notes that reducing UPS losses and appropriately managing load factors directly contribute to energy savings. For some models, the use of eco mode can also be expected to reduce electricity costs.^*6,7　
Therefore, in future data centers, important design considerations will include not only how much power is used, but also how power is distributed within the facility, how losses are reduced, and how backup systems are configured.
In response to the rapid expansion of AI data centers, a 2024 DOE recommendation also states that new backup power strategies and flexible operating methods should be supported for data centers.^*8　It indicates the need to consider not only conventional emergency power, but also battery storage, alternative power sources, long-duration storage, and the use of grid services. It also notes that virtual power plants and microgrids can serve as starting points.^*8　
This trend shows that future data centers are shifting from mere power-consuming facilities to infrastructure that must perform advanced power management while supporting high loads.

*5：Best Practices Guide for Energy-Efficient Data Center Design*6：Reduce Energy Loss from Uninterruptible Power Supply Systems*7：Uninterruptible Power Supplies*8：Powering AI and Data Center Infrastructure: Recommendations

In data centers, most of the power consumed ultimately becomes heat, making cooling a central part of operations.
The DOE design guide treats airflow management as an important measure for preventing the mixing of supplied cold air with server exhaust heat and for delivering the required amount of cooling to the required locations.^*9　
In particular, hot-aisle/cold-aisle configurations, which clearly separate the intake and exhaust sides of racks, are a fundamental concept in high-density data centers. Proper design can lead to lower operating costs, reduced risk of processing interruptions, and improved power density.
In air-cooled data centers, even slight air leakage or disturbance in the airflow path can significantly affect cooling efficiency. The DOE and ENERGY STAR present the prevention of cold-air and exhaust-air mixing, the suppression of unnecessary bypass airflow, and the management of openings inside and outside racks as effective energy-saving measures.^*10　
ENERGY STAR notes that properly blanking off empty rack space can, in some cases, deliver energy savings of 1–2% per rack. DOE case examples have also confirmed that improved airflow management can increase cooling capacity and reduce fan power.^*11　
Therefore, important design elements in data centers include not only large-scale cooling equipment, but also precise airflow control around racks and the suppression of air leakage paths.​
In high-density servers, the power consumption of internal server fans cannot be ignored. The DOE design guide explains that as server inlet temperature rises, internal fan speed increases. For example, a rise from 77°F to 91°F can increase fan speed by about 30% and more than double fan power.
For this reason, rather than simply raising room temperature, it is important to stabilize the thermal environment at the server inlet. For high-heat-generating GPU servers, the DOE also recommends considering direct liquid cooling. Going forward, mechanisms that transport heat using liquids, rather than only moving air, are expected to become increasingly important.

*9：Best Practices Guide for Energy-Efficient Data Center Design*10：Manage Airflow for Cooling Efficiency | ENERGY STAR*11：DOE AM Tool–User’s Manual v3.1 (7-25-2023).pdf

The environmental impact of data centers cannot be fully assessed by power consumption alone.
The DOE states that in data centers using water cooling or cooling towers, water usage is influenced not only by the heat generated by IT equipment, but also by the efficiency of each stage of heat removal. It presents WUE, or Water Usage Effectiveness, as an evaluation metric.^*12　
In the EU as well, a reporting system has been established to collect and publish data on water footprint in addition to energy performance.^*13　Going forward, data centers will be evaluated not only by how much computing they can perform, but also by how efficiently they can operate with minimal power loss and water use.
For this reason, future data centers will require not only improved server performance, but also efficient power delivery, stable operation during outages, and optimized air and liquid flow to dissipate heat effectively.​
In other words, we believe we have entered an era in which the competitiveness of a data center is determined not by IT equipment alone, but by the design capability of the surrounding infrastructure: how electricity, heat, and water are managed.

*12：Best Practices Guide for Energy-Efficient Data Center Design*13：Energy performance of data centres – Energy – European Commission