Efficient Waste Heat Utilization Series

In today’s society, access to information has become a driving force for global development. The rapid growth of the artificial intelligence (AI) industry, fueled by the emergence of powerful graphical processing units (GPUs), has further accelerated the demand for data centers. These GPUs, designed to handle massive parallel computations, have become essential for AI applications such as machine learning and deep learning. As a result, the widespread use of cloud data services has transformed the data center industry, making data access more convenient. Massive data storage in the cloud is supported by data centers of various scales. However, operational costs for data centers are increasing, with a significant portion stemming from energy demand for cooling infrastructure.

Traditional data center designs often rely on energy-intensive air-cooling processes and excessively engineered indoor air conditioning. By improving air handling designs, we can start to enhance data center cooling methods. Market-leading air-cooling technologies reduce server exhaust and intake mixing by rearranging server racks (“hot aisle”-“cold aisle” configurations) and use detailed computational fluid dynamics analyses in raised-floor designs to optimize airflow and minimize pressure drops. Through large-scale data center redesigns and natural cooling integration, Google has lowered its average PUE to 1.10. However, small and medium-sized data centers still face significant challenges in reducing cooling energy consumption.

With the emergence of high-density data centers, the market share of next-generation liquid-cooling technologies is gradually increasing. Liquid has a heat transfer capacity three orders of magnitude higher than air, making liquid cooling devices better suited for cooling information and high-energy-density electronic devices, such as GPUs. Despite higher initial investment costs, liquid-cooling devices on the market reportedly reduce energy consumption by 30-50% on average. Unfortunately, waste heat extracted from data equipment as a medium-to-low temperature heat flow is low-grade energy, making it difficult to reuse in most scenarios.

One solution is adsorption cooling, which utilizes waste heat from data centers to drive cooling systems. This technology can significantly reduce energy consumption and operational costs while maintaining the required cooling capacity for data centers. Adsorption chillers can operate at low-grade heat sources, making them ideal for use with data center waste heat.

Boulton heat pumps, based on unique materials and advanced thermal cycle designs, can be integrated with air-cooling and liquid-cooling devices in the market, directly utilizing waste heat from data equipment to drive innovative cooling systems, including adsorption chillers. Without the need for additional power input, waste heat is converted into heating or cooling streams for data center infrastructure beyond server racks. Boulton heat pumps can further improve the energy efficiency of existing air-cooling and liquid-cooling devices by up to 30%, collectively reducing energy consumption for data center cooling processes by up to 80% (20-30% of a data center’s total power consumption). This integration of adsorption cooling technology with data center cooling systems provides a sustainable solution for reducing energy consumption and costs, promoting a greener future for the industry.