The proportion of high-heat-flux servers has gradually increased with the increasing number and scale of data centers. Traditional air cooling cannot handle the heat dissipation and energy saving needs. In this regard, the liquid cooling technology with a large latent heat and high heat transfer coefficient has emerged. Research on liquid cooling systems primarily focuses on system efficiency enhancement, system comparison, and overall thermal management. Among them, the working fluid, as an important medium for heat exchange, is crucial for the liquid system performance. In this paper, selection principles of phase-change working fluid are proposed. The investigations and application progress of six types of phase-change working fluid (natural working medium, HFCs, HFOs, PFCs, HFEs, mixed working medium) are reviewed, summarized, and analyzed according to the contact modes (direct and indirect). The prospects for future research on phase-change liquid cooling fluids are presented.

The power usage limit becomes lower with the green development of data centers. The small liquid cooling device is emerging as a new cooling option for small data centers with high energy density, which helps recover the waste heat. In this paper, the energy efficiency characteristics of three types of water-source heat pump systems under different surplus heat conditions are compared and analyzed in the surplus heat load range of small data centers. The energy efficiency of the water-source heat pump system is distributed at 5-7, 5-6, and 4.5-6.5 on the source side, transmission and distribution side, and user side, respectively. When the user-side heat load is retained at 0 kW to 1750 kW and 2500 kW to 3200 kW, the energy efficiency of the source-side enhanced water-source heat pump system is highest. When the user-side heating load demand is 1750 kW to 2500 kW, the energy efficiency of the source-side enhanced water-source heat pump system is near that of the transmission and distribution-side enhanced water-source heat pump system. Finally, the coupling matching formula Q_N=kQ_S is obtained between the user-side heat load demand and data center surplus heat load, in which k ranges from 1.1962 to 1.3642.

The low-carbon transformation of data centers is highly significant for achieving carbon peaking and carbon neutrality. This study compared and analyzed the overall situation of data centers in China. Three variables—energy efficiency improvement rate, proportion of non-fossil energy—and negative emission technology intensity were introduced based on the CO₂ emission and intensity targets of China in key years, and the total CO₂ emissions of the data centers were projected via scenario analysis. The results demonstrated that the power consumption of the data centers increased gradually; the carbon emissions first increased and then decreased, and the power usage effectiveness (PUE) of the data centers decreased gradually. The carbon peak time of the three scenarios is 2030, and the expected times to achieve carbon neutrality are 2059, 2057, and 2055 in the three scenarios. In light of the goal to achieve carbon neutrality by 2060, the data center industry should further improve the energy efficiency utilization rate, increase the proportion of non-fossil energy, strengthen the technological innovation of carbon capture and storage, and enhance the level of carbon sinks.

Waste-heat recovery and utilization represent key technological pathways for promoting the low-carbon development of data centers. A comprehensive analysis of annual energy consumption and carbon emissions, coupled with a rational evaluation of system-level efficiency, is essential for advancing the integrated utilization of data-center resources. This study focuses on an integrated cooling and waste-heat recovery system in data centers and proposes two evaluation indicators from the perspectives of energy and carbon emissions: general exergy efficiency, η_GEX and general carbon efficiency, η_GOC. The proposed indicators are validated using simulation data. The results indicate that the general exergy efficiency metric effectively evaluates data center integrated systems. Specifically, the system in Lhasa achieves the highest general exergy efficiency of 29.27%, whereas the lowest general exergy efficiency (23.65 %) is observed in Harbin. The general carbon efficiency indicator enables a quantitative assessment of the carbon-reduction potential achieved by replacing conventional heating systems integrated with data centers. Depending on the type of displaced traditional heating technology used, the general carbon efficiency in Lhasa ranges from 3.75 to 4.45. Overall, the proposed evaluation framework provides a robust and rational basis for assessing the operational efficiency and carbon-mitigation potential of data center integrated systems, offering valuable guidance for their practical deployment and retrofitting in modern data centers.