Moonshot AI, developer of the Kimi K3 model, reportedly lacks sufficient compute to both train and serve its models, indicating a significant bottleneck for Chinese AI development. The K3 model is estimated to be approximately six months behind US counterparts like Mythos and GPT-Sol in capabilities, with its production likely involving violations of US export controls.
US export controls are effectively limiting Chinese AI model development and deployment by creating compute scarcity, reinforcing the US's strategic advantage in AI infrastructure.