⼤摩:2026将是AI科技硬件之年

Core Viewpoint - Morgan Stanley predicts that 2026 will be a pivotal year for explosive growth in AI hardware, primarily driven by strong demand for AI server hardware [4][5]. AI Hardware Growth Drivers - The growth of AI hardware is shifting from the H100/H200 era to a new cycle driven by NVIDIA's GB200/300 and the upcoming Vera Rubin platform [5][6]. - The demand for AI server racks is expected to surge from approximately 28,000 units in 2025 to at least 60,000 units in 2026, representing over 100% growth [8]. Power and Cooling Solutions - The report highlights that the challenges of power consumption and cooling due to hardware upgrades present significant opportunities for power and cooling suppliers [10]. - The transition to 800V high-voltage direct current (HVDC) power solutions is anticipated, which will greatly enhance the value of power solutions [12][14]. - By 2027, the value of power solutions for Rubin Ultra cabinets is expected to exceed ten times that of current GB200 server cabinets [14]. Liquid Cooling Components - Liquid cooling has become a standard requirement, with the total value of cooling components for a GB300 server rack estimated at approximately $49,860, expected to increase by 17% for the Vera Rubin platform [10][15]. PCB and Interconnect Upgrades - The report emphasizes the significant impact of AI platform upgrades on printed circuit boards (PCBs) and interconnect components, with increasing requirements for layer counts and material grades [16][20]. - The number of layers in ABF substrates is projected to increase from 12 layers for H100 to 18 layers for Vera Rubin [17]. - The demand for high-layer HDI boards and high-grade CCL materials is expected to create structural growth opportunities for PCB and upstream material suppliers [20][22].