NVIDIA A100 Tensor Core GPU from China
Data center GPU with 80GB HBM3 memory designed for AI training and HPC simulations. Classified in HTS 8542.31.00.40 as a graphics processor optimized for matrix math and deep learning workloads. Features 6,912 CUDA cores and multi-instance GPU partitioning technology.
Duty Rate — China → United States
50%
Rate breakdown
9903.91.0550%Effective with respect to entries on or after January 1, 2025, articles the product of China, as provided for in subdivision (f) of U.S. note 31 to this subchapter
9903.03.030%Articles the product of any country, as provided for in subdivision (aa)(ii) of U.S. note 2 to this subchapter
Import Tips
• Export license may be required for high-compute GPUs; check BIS Entity List for suppliers
• Declare tensor core count and FP64 performance specs to justify graphics processor classification
• Segregate data center vs consumer GPUs in manifests to avoid averaging duty rates