Yes, but the hardware they use for inference like Huawei Ascend 910C is less efficient than Nvidia H100 used in US due to the difference in the process node.