Core Viewpoint - NVIDIA's CFO Colette Kress believes there is no bubble in the AI sector and anticipates a significant market transformation by 2030, with data center infrastructure potentially reaching $3 trillion to $4 trillion due to increasing demand for accelerated computing [2]. Group 1: Market Outlook - The majority of NVIDIA's AI chip shipments are aimed at building new data center infrastructure rather than replacing existing equipment [2]. - By 2026, NVIDIA expects to have orders worth $500 billion for its Blackwell and Rubin GPU chips, which exceeds 3.5 trillion [2][4]. Group 2: Partnerships and Investments - NVIDIA is working on a final agreement with OpenAI, with a previous investment of $100 billion aimed at enhancing OpenAI's computing power through the purchase of NVIDIA's AI GPUs [4]. - The collaboration between NVIDIA and OpenAI is expected to be ongoing, contributing to rising stock prices and valuations for both companies, similar to trends seen with Oracle and Amazon [4]. Group 3: Production and Delivery - At the recent GTC conference, NVIDIA's CEO Jensen Huang announced that the company has GPU orders totaling $500 billion, with plans to deliver 20 million units of Blackwell and Rubin architecture GPUs over the next five quarters [4].

Core Viewpoint - The demand for high-performance computing chips and AI servers is driving the continuous iteration of liquid cooling technology, with NVIDIA leading the charge in product upgrades and requiring suppliers to accelerate the development of MLCP liquid cooling technology [1][3]. Group 1: Semiconductor Industry Performance - The semiconductor equipment sector is leading the electronic industry, with the SW electronic industry index rising by 2.96%, ranking 3rd out of 31 sectors, while the CSI 300 index fell by 0.44% [2]. - The top five sectors by performance include coal (+3.51%), power equipment (+3.07%), electronics (+2.96%), automotive (+2.95%), and machinery (+2.23%) [2]. - The top three sub-sectors within the SW electronic industry are semiconductor equipment (+9.98%), optical components (+9.08%), and integrated circuit manufacturing (+5.63%) [2]. Group 2: High-Performance Chip Power Consumption - The power consumption of NVIDIA's AI chips has increased from 400W for the A100 to 1200W for the GB200 and 1400W for the GB300, with significant performance improvements [3]. - The FP16 computing power of the GB200 reaches 5 PFLOPS, and the FP4 computing power reaches 20 PFLOPS, with NVLink bandwidth increasing to 3.6 TB/s [3]. - The total power consumption of NVIDIA's flagship DGX B200 chassis, which integrates eight B200 GPUs, reaches 14.3 kW, necessitating an additional 60 kW power and cooling capacity at the rack level for stable operation [3]. Group 3: Liquid Cooling Technology - Liquid cooling is a superior heat dissipation method compared to traditional air cooling, especially when cabinet density exceeds 20 kW, enhancing device performance and lifespan [4]. - MLCP is identified as a key form of liquid cooling technology, designed for high-performance computing devices, featuring microchannels for efficient heat transfer [4]. - MLCP supports higher TDP power and lower thermal resistance compared to traditional liquid cooling plates, although it comes with higher costs and manufacturing challenges [4]. Group 4: Potential Beneficiaries in Liquid Cooling Industry - Companies like Invec, Highlan, and Siquan New Materials are positioned to benefit from the trends in liquid cooling technology [5]. - Invec offers a comprehensive liquid cooling solution and has delivered 1.2 GW of liquid cooling products with zero leakage as of March 2025 [5]. - Highlan specializes in thermal management solutions for data centers, providing both cold plate and immersion cooling options with a PUE of ≤1.1 [5]. - Siquan New Materials is actively developing liquid cooling technologies and aims to provide customized solutions across multiple industries [5].