Sovereign AI is no longer an option but a mandatory one. Fable5 and Mythos5 were just launched three days ago when they were ordered globally to be cut off by the US government due to a supposedly "narrow jailbreak" issue. Anthropic has publicly stated that they do not agree with this approach, believing that the evidence is not sufficient to suddenly remove commercial models, but they still have to comply. This sends two cruel signals: 1) The US government can intervene in cutting-edge commercial AI at any time, even if you are a paying user, even if the model is already open, and even if the company itself opposes it. 2) Frontier AI has been clearly defined as a strategic resource, standing in the same control logic as high-end chips and nuclear technology. Export control is no longer just a matter of hardware, but extends to model weights and capabilities themselves. The signal is very clear: the strongest AI cannot be easily used by "foreigners" In the past few years, sovereign AI has been more of a strategic discussion and policy slogan for some countries. Now, this event has significantly increased the 'necessity'. When countries discover that even American companies like Anthropic, which place a relatively strong emphasis on security, cannot guarantee the sustained availability of their strongest models, any rational government will reassess the risk of relying on external frontier models. This will have several direct consequences: 1) Autonomous acceleration of computing power: Countries will spend more money to build their own intelligent computing clusters. Europe is already calling for building its own cutting-edge modeling capabilities, while Middle Eastern sovereign foundations will continue to invest heavily in hardware to build data centers. China will continue to adopt a dual approach of domestic chips, algorithm optimization, and national computing power networks. 2) Model autonomous acceleration: Closed source frontier models are becoming increasingly unreliable, and open-source weighting+local fine-tuning+synthetic data will become the more mainstream path. China has already demonstrated an advantage in global downloads of open-source models, and such events will only accelerate this trend. 3) Large scale accumulation of government funds: The private capital expenditures of major technology companies are already very strong (several giants are expected to continue to maintain high-intensity investment in 2026), and now governments around the world will add another layer of sovereign funds. The EU's € 200 billion plan, the UK's sovereign AI fund, and China's national level intelligent computing projects will all receive more political and budget support due to such events. This is not an isolated incident, but the beginning of a new normal Looking at the timeline, the export control of chips by the United States has been going on for many years, and now expanding to the model level is a natural evolution. In the future, there may be more similar operations: a model may be restricted from access due to "security risks", and a company may be required not to provide services to specific countries. In this environment, sovereign AI is no longer a romantic strategic narrative, but a cold survival strategy: Europe will continue to struggle between "technological sovereignty" and "regulatory sovereignty"; The Middle East will use money to buy hardware and models; Southeast Asian and global southern countries are more likely to embrace open source+multi-source strategies; And China is highly likely to become one of the countries with the most resolute and systematic investment in sovereign AI construction. The United States aims to maintain its AI advantage through export controls, which is likely to further promote the multipolarization of global AI. Of course, many people may say that countries may not be able to achieve the capabilities of the US big model, but it is not important. If countries really want to promote sovereign AI, the core must remain available and controllable at a certain level, and the core cannot be suddenly choked. Nvidia CEO Huang suddenly realized that the sovereign AI, which had been shouting for a long time, had added a significant catalyst, making him even happier. This is a good thing for the AI computing power industry chain, which means that in addition to the capital expenditures of large technology companies, governments around the world will stack a large amount of public funds and policy support in the future, jointly pushing up the total expenditure on global AI infrastructure and sovereign AI construction. Of course, the starting point is different: 1) Big tech companies pursue model leadership, technological advancement, commercial returns, and market share, and they will continue to invest money. 2) Governments of various countries pursue national security, sovereignty, and strategic autonomy. They are concerned about future bottlenecks, so they are willing to spend money on building local computing power, models, and ecosystems, even if the short-term returns are not high. In history, after the signing of the Treaty on the Non Proliferation of Nuclear Weapons in 1968, five more countries actually possessed the capability of nuclear weapons. When a country uses the most advanced technology as a weapon, the most rational response of other countries is to quickly produce the weapon, or at least create a version that can protect itself.