we think Surge is also active internationally, selling to Chinese labs like Moonshot and Z.ai. Indeed, we believe access to these RL environments played a huge part in increasing the capabilities for Kimi K2 Thinking and GLM-4.6. When companies advertise increased capabilities in PowerPoint deck or Excel sheet creation, that is directly downstream of additional RL training. Chinese VC firms are actively trying to stand up Chinese data foundry competitors so they can fully serve the local ecosystem at cheaper prices than western ones. Most Chinese labs are still early to scaling RL: Qwen is currently spending around 5% of their pre-training compute on post-training.