H20
A compliant data center GPU introduced by NVIDIA for specific markets, adapted in compute power and interconnect parameters to meet export-control thresholds.
Timeline (last 90 days)
AntAngelMed achieves inference speeds of over 200 tokens/s on H20 hardware.
NVIDIA has launched "special supply" chips such as the A800, H800, and H20 to circumvent the red lines of U.S. export controls.
Nvidia once launched compliant downgraded models such as the H20 for the Chinese market, but such products have been substantially cut in compute power, their cost-effectiveness and appeal have continued to decline, and the policy boundary may tighten at any time.
NVIDIA has launched the castrated H20 chip targeting the Chinese market, while domestic alternatives such as Huawei Ascend 910B are also accelerating to catch up.
All Facts (6)
英伟达H20芯片在2025年初遭到出口禁令限制
85%UnverifiedThe H20 has 96GB HBM3 memory, and the L20's FP8 computing performance is approximately half that of the H20.
60%UnverifiedAntAngelMed achieves inference speeds of over 200 tokens/s on H20 hardware.
50%UnverifiedNVIDIA has launched "special supply" chips such as the A800, H800, and H20 to circumvent the red lines of U.S. export controls.
50%UnverifiedNvidia once launched compliant downgraded models such as the H20 for the Chinese market, but such products have been substantially cut in compute power, their cost-effectiveness and appeal have continued to decline, and the policy boundary may tighten at any time.
50%UnverifiedNVIDIA has launched the castrated H20 chip targeting the Chinese market, while domestic alternatives such as Huawei Ascend 910B are also accelerating to catch up.
50%