China’s AI Advantage Is Bigger Than You Think

Summary

AccelXR, author of Delphi Digital's 'Evaluating the AI Labs of China', joins Tommy to argue that China's AI progress is driven by real innovation, not just distillation. They discuss Chinese labs' reinforcement learning and mixture-of-experts work, hardware/software co-design with domestic chips, China's electricity advantage, and three scenarios for the US-China AI race. The conversation covers talent flows, open-source dynamics, and implications for Alibaba, Chinese AI cloud pricing, and US labs.

• China's AI labs are innovating beyond distillation, especially in reinforcement learning and mixture-of-experts.
• Chinese labs are writing low-level GPU code and co-designing with Cambricon and Ascend.
• Electricity may become the key AI bottleneck, and China has a large grid-capacity advantage.
• Three scenarios for US-China AI: functioning bridge, complete breakdown, or US embracing open source.
• Chinese AI researcher returnee rates and domestic PhD pipeline are rising.
• Chinese AI cloud pricing and domestic chip training are key signals to watch.
• US labs still hold a capability lead, but China's training sovereignty is the flipping point.