新一代浸没式液冷解决方案

Search documents
研报掘金丨天风证券:维持亨通光电“买入”评级,AI算力时代光通信布局领先
Ge Long Hui A P P· 2025-09-18 07:55
格隆汇9月18日|天风证券研报指出,亨通光电上半年实现归母净利润16.13亿元,同比增长0.24%;Q2 实现归母净利润10.56亿元(YOY-3.63%),业绩基本符合预期。海洋项目持续拓展斩获,在手订单 多。截止2025H1,公司拥有海底电缆、海洋工程及陆缆产品等能源互联领域在手订单金额约200亿元, 或为后续持续快速增长做强有力支撑。亨通光电AI先进光纤材料研发制造中心扩产项目已于近期正式 投入建设,计划2026年2月竣工。该项目瞄准未来发展趋势,布局扩产,满足对多模、多芯、空芯等特 种光纤持续增长的市场需求。液冷方面,2025年,亨通与行业合作伙伴联合打造的高密度计算场景--新 一代浸没式液冷解决方案发布,契合数据中心未来散热需求。维持"买入"评级。 ...
超微电脑5个月内大涨50%,全球AI数据中心液冷市场迎来新热潮
Tai Mei Ti A P P· 2025-05-19 15:34
Core Viewpoint - The article discusses the advancements in liquid cooling technology for data centers, particularly highlighting Supermicro's new DLC-2 technology and its implications for energy efficiency and operational costs in the context of rising AI computing demands [2][5][9]. Group 1: Supermicro and DLC-2 Technology - Supermicro announced the launch of its next-generation DLC-2 technology, which has led to a fourfold revenue increase over the past three years, driven by the maturity of liquid cooling technology [2]. - The new data center construction modular solution (DCBBS) integrates the DLC-2 technology, achieving a 40% improvement in energy and water savings, a heat capture rate of 98%, and a noise level as low as 50db, while also reducing capital expenditures by up to 20% [2][9]. - Supermicro's stock has experienced significant volatility, with a 300% increase in 2023 due to AI demand, followed by an 80% decline in 2024 due to financial reporting delays and governance issues [7][12]. Group 2: AI Computing Power and Market Trends - NVIDIA's CEO Jensen Huang emphasized the critical role of computing power in AI development, noting that AI computing performance improves by a factor of one million every decade [4]. - The IDC report predicts that China's AI computing power market will reach $25.9 billion by 2025, growing 36.2% from 2024, while total energy consumption in Chinese data centers is expected to exceed 400 billion kWh, accounting for over 4% of total national consumption [4]. Group 3: Liquid Cooling Technology Overview - Liquid cooling technology is categorized into non-contact and contact types, with non-contact cooling being the most widely used due to its maturity [5]. - The average energy consumption for cooling systems in air-cooled data centers exceeds 50%, with higher ratios in AI computing centers [5]. - The emergence of high-density computing scenarios has led to the development of new immersion cooling solutions, which are expected to see widespread adoption as server power densities increase [5][6]. Group 4: Financial Performance and Future Outlook - MiaoSuan Technology, founded in 2023, aims for a revenue target of 300 million yuan in 2025, leveraging its immersion cooling technology [6][7]. - Supermicro's recent partnership with DataVolt in Saudi Arabia indicates ongoing interest in high-performance computing solutions despite past financial challenges [7]. - The operational efficiency of liquid cooling systems can lower Power Usage Effectiveness (PUE) to as low as 1.15, with potential for further reductions in specific areas [9].