Nvidia(US:NVDA) Jin Rong Jie·2025-12-25 00:10Core Insights - NVIDIA's adoption of liquid cooling technology for its B100/H200 AI servers marks a significant shift, elevating liquid cooling from an optional technology to a necessary configuration for AI servers, driven by surging computational power demands and energy efficiency regulations [1][2] Group 1: Liquid Cooling Technology Adoption - NVIDIA's GB300 series AI server cabinets will begin mass production by the end of 2025, with the Vera Rubin platform expected to launch in the second half of 2026, further enhancing the value of liquid cooling [1] - The power consumption of AI servers has increased from 3000W to over 10000W, making traditional air cooling inadequate; Goldman Sachs projects that the penetration rate of liquid cooling in AI servers will rise from 15% in 2024 to 45% by 2025 [1] - The Ministry of Industry and Information Technology mandates that new large data centers maintain a Power Usage Effectiveness (PUE) of below 1.2, with liquid cooling being the only solution capable of consistently meeting this requirement [1] Group 2: Market Growth and Financial Implications - The value of liquid cooling components in NVIDIA's GB300 NVL72 rack-level AI systems is estimated at $49,860, representing a nearly 20% increase from previous generations [1] - The Chinese liquid cooling server market is projected to reach $3.39 billion by 2025, with a year-on-year growth rate of 42.6%; the compound annual growth rate from 2025 to 2029 is expected to remain at 48%, with the market potentially exceeding $16.2 billion by 2028 [1] - The widespread adoption of liquid cooling by NVIDIA strengthens the growth certainty of the liquid cooling sector, benefiting core equipment and component manufacturers directly [2]