英伟达悄悄修改数据中心铜用量表述，曾被指“出现单位乌龙”

Group 1 - The core controversy revolves around a significant error in the estimation of copper demand related to AI data centers, initially reported by NVIDIA, which has since been corrected [1][2] - NVIDIA's original claim suggested that a 1 GW data center could require up to 500,000 tons of copper, but this was later clarified to approximately 200 tons based on correct calculations [2][3] - The revised estimates indicate that even with the projected growth of global data centers to about 219 GW by 2030, the copper demand would only be around 44,000 tons, representing less than 0.2% of the current global annual copper production [3] Group 2 - Analysts believe that the main issue is not about copper shortages but rather the engineering limits of low-voltage power architectures in high-power data centers [4] - The original intent of NVIDIA's article was to highlight the need for transitioning to new power supply solutions, such as 800V high-voltage direct current (HVDC), as traditional systems struggle with increased current and efficiency [4] - Misinterpretations regarding AI's impact on copper resources stemmed from misunderstandings of technical context and unit expressions [4]