As global investors eagerly await the unfolding of China’s A-shares market on Tuesday, the tech sector is quietly preparing for 2025. On Monday, CES (Consumer Electronics Show) officially announced that Nvidia's founder and CEO, Jensen Huang, will deliver a keynote address on January 6, the eve of the Las Vegas event.
What can we expect from the upcoming CES? There’s a strong possibility that Huang will unveil the RTX 50 series graphics cards, featuring the new Blackwell architecture, including the highly anticipated 5090 model. According to insider information, the RTX 50 series is set to be a major focus for Nvidia at CES. Rumors suggest that the RTX 5090 may debut alongside the RTX 5080, both of which are expected to have significantly increased power consumption. The 5090 is reported to possess 21,760 CUDA cores, with power usage reaching up to 600 watts, 150 watts more than its predecessor, the 4090. Furthermore, both graphics cards are rumored to utilize GDDR7 memory, with the flagship model boasting 32GB.
Considering the typical product update cycles in consumer electronics, it has been two years since Nvidia first launched the RTX 40 series, which debuted in September 2022. Earlier this year at CES, Nvidia refreshed its 40 series with the introduction of Super models, including the RTX 4070 Super, RTX 4070 Ti Super, and RTX 4080 Super. As the leader of a $3 trillion "AI chip empire," Huang is expected to emphasize advancements in AI hardware and software. However, for investors, straightforward GPU shipment data often drives ambition for future growth.
Amid this backdrop, Supermicro, a key player in Nvidia's ecosystem, reported strong shipment figures, seeing its stock rise by more than 15% on Monday. The company announced significant progress in its liquid cooling solutions, highlighting that it has shipped over 100,000 GPUs within a single quarter.
Supermicro is launching a comprehensive liquid cooling solution for data centers that includes cooling distribution units, cold plates, cooling towers, and end-to-end management software. Their new technology is projected to increase computing density by four times, allowing each server rack to accommodate up to 96 Nvidia B200 GPUs. This infrastructure innovation is said to reduce energy consumption by up to 40% and save 80% of physical space. The system supports hot water cooling up to 45 degrees Celsius, enabling the use of waste heat from AI chips for regional heating and greenhouse applications.
Recently, Supermicro has deployed over 100,000 liquid cooling GPUs for several major AI operations and cloud service providers. Given the demands of AI and high-performance computing workloads—each server operating close to 12kW and generating over 100kW of heat per AI rack—energy-efficient and space-saving cooling systems are crucial for large cloud service providers.
Since August, Supermicro has faced turbulence following a negative report from Hindenburg Research regarding its revenue recognition policies, which has consequently sparked a federal investigation. The company previously reached a settlement with the U.S. SEC in 2020, paying $17.5 million in penalties.
Needham analyst Quinn Bolton notes that Supermicro’s pioneering liquid cooling technology could help the company expand its market share further. Currently, the company's stock price reflects the board’s evaluation of the internal control risks involved.
Ranking
