Anthropic's latest Fable 5 and Mythos 5 are stuck in US export controls, and US KYC can only be used in the US. There is no need to feel sorry for Anthropic. It was completely done by oneself, and GPT's flagship model was not affected. But I also feel that for Chinese AI companies, this is like giving away a market gap for nothing. The demand for places where Claude is not needed will not disappear, it will only turn. Turning to whom? The overseas expansion of Chinese AI companies is actually accelerating. Zhipu's GLM 5.2 is currently the strongest open-source model, with outstanding architecture design and planning capabilities. The performance of the latest model has reached 80-90% of Claude Opus, and without pursuing front-end design and multimodality, the vast majority of scenarios can already be replaced. The overseas market of Zhipu has always been very hot, @ Zaiuorg should encourage more overseas users to try it out. They are not only making products for going global, but also actively seeking users. Zcode now provides 3 million free tokens for trial every day, which is really strong. Kimi, MiniMax, and Qwen are also promoting the international market. I think Qwen's strategy is the clearest, which is to open source small models to grab the ecological niche in the first place and commercialize flagship closed source models. Previously, excessive open source dismantled the moat, but now this adjustment is correct. But it's not just model companies that are moving, China's AI infrastructure is also investing and going global to solve the problem of computing power going global. In addition to investing 2 trillion yuan in computing infrastructure, China is also promoting the export of enterprise level computing power. Some state-owned enterprises' computing power center services are also migrating overseas and providing services directly locally. Package computing power and models, and sell infrastructure and applications together. Computing power is the oil of the new era, settlement is in RMB, and China wants pricing power. There is an almost identical story in history. The United States was the first to master oil refining technology and once monopolized pricing power. Later, oil producing countries formed the OPEC alliance, and the pattern was rewritten. Nowadays, computing power is following the same path, but with more complex variables. More than 60% of token costs come from electricity. The United States is more efficient and intelligent in AI chips and intelligent computing, but China is self-sufficient in the entire industry chain from power generation equipment to photovoltaic and wind power to transmission and distribution, and cannot find a second one globally. The United States has technology but lacks an energy base. China has a complete industrial chain and is packaging its computing infrastructure for export. Whoever gains pricing power depends on who can integrate the triangle of technology, energy, and infrastructure. Export controls have indeed stuck some key points in the short term. But conversely, it is also forcing Chinese AI companies to build their own complete systems. When this system matures, the pattern of computing power pricing power will undergo substantial changes. Chinese AI companies going global is not a multiple-choice question, it is a mandatory question. The question is, can they still sit at the table when the game ends.