our central estimate is around 660,000 H100-equivalents
【令人震惊的数字】走私流入中国的算力中位估算:66 万个 H100 等效——约占中国 AI 算力总量的三分之一。这个数字彻底改变了「出口管制正在有效阻断中国 AI 发展」的主流叙事。如果三分之一的算力来自走私,那么所有基于「中国无法获得先进芯片」假设的中美 AI 差距分析,都需要用这个修正系数重新计算。
our central estimate is around 660,000 H100-equivalents
【令人震惊的数字】走私流入中国的算力中位估算:66 万个 H100 等效——约占中国 AI 算力总量的三分之一。这个数字彻底改变了「出口管制正在有效阻断中国 AI 发展」的主流叙事。如果三分之一的算力来自走私,那么所有基于「中国无法获得先进芯片」假设的中美 AI 差距分析,都需要用这个修正系数重新计算。
We estimate, with 90% confidence, that between 290,000 and 1.6 million H100-equivalents of compute were smuggled through the end of 2025.
大多数人可能认为走私到中国的AI芯片数量在数万级别,但作者的估计显示实际数量可能高达数十万甚至上百万H100等效芯片,这一数量级远超公众认知,表明走私问题的严重程度被严重低估。
The biggest driver of uncertainty on the diversion side is that we don't know what fraction of diversion has been observed. The large-scale smuggling schemes detected and reported so far could represent the majority of the volume, or they might be just a small fraction of the total flows.
大多数人认为已曝光的大型走私案件代表了走私活动的主体,但作者指出这些已知的案件可能只是冰山一角,实际走私规模可能是已知的数倍,这挑战了我们对当前走私情况掌握程度的认知。
We estimate that between 290,000 and 1.6 million H100-equivalents (H100e) were smuggled to China through 2025. Our median estimate of 660,000 H100e would be roughly a third of China's total compute.
大多数人认为美国出口管制能有效遏制中国获取先进AI芯片,但作者认为这些管制实际上导致大量芯片被走私到中国,走私数量可能与中国合法获取的芯片数量相当,这意味着出口管制的效果远不如预期。
TSCi agents pass by Ntufia’s SAEMAPE officers once or twice a monthto collect the documents containing all information on the mineral bags and their tags. These SAEMAPEagents however rarely visit the actual mines. The miners take their production home in the evening, andmineral traders buy in the miners’ camp. Once traders have acquired a sufficient volume, ready for ex-port, they will call on the SAEMAPE agent to come over to label the mineral bags. SAEMAPE agents, andconsequently iTSCi, have no actual oversight on production at the level of the mines.
non reliance on ddp or traceability schemes - mechanism, and gaps in the tagging and bagging.