Is NVDA completely blocking its escape route? China's AI industry is about to undergo a major reshuffle, and no one can escape these 5 major changes: 1. A real opportunity for domestic AI chips: Previously, everyone could rent NVIDIA GPUs, and even if domestic chips had slightly lower performance, not everyone was willing to bother with them. Now, with overseas computing power channels tightening, domestic chips such as Huawei Ascend, Cambricon, Hygon, and Kunlun Chip may receive more orders. What was once a backup plan may now be indispensable. 2. Alibaba Cloud's AI business may be impacted: One of Alibaba Cloud's biggest advantages is its ability to provide customers with NVIDIA GPU computing power. Customers pay to rent cards to train models, while Alibaba provides the computing power and earns money from cloud services. If overseas GPU procurement and leasing are hindered, customers who want to rent cards cannot obtain them, and the growth rate of its AI cloud business may slow down. Alibaba has its own Pingtouge chips, but customers may not be willing to replace NVIDIA with domestic cards. 3. ByteDance may face a larger computing power gap: Whether ByteDance can maintain its leading position largely depends on whether the large model behind it can continue to be upgraded. ByteDance has historically relied heavily on overseas data centers to rent NVIDIA GPUs. If this route is blocked, training next-generation large-scale models could become slower and more expensive. Going forward, ByteDance will compete not only on algorithms but also on its ability to acquire a sufficient number of high-end chips. 4. AI Startups May Face a Survival Test Companies like Lunar Dark Side, Zhipu, and MiniMax previously avoided spending billions on data centers, directly renting overseas GPUs to train models and using algorithms and talent to catch up with OpenAI and Anthropic. Now, if even renting overseas GPUs becomes increasingly difficult, companies without the funds to purchase chips or the ability to build their own data centers will be at a significant disadvantage. The AI industry may undergo a reshuffling, with large companies becoming stronger and smaller companies struggling to survive. 5. DeepSeek and the Domestic Computing Power Supply Chain May Benefit Against the Trend DeepSeek's early adaptation to Huawei Ascend chips has driven the migration of models from the NVIDIA CUDA ecosystem to the domestic CANN ecosystem. Although the migration cost is not low, if NVIDIA computing power becomes increasingly difficult to obtain, companies that complete domestic adaptation early may gain a competitive advantage. At the same time, the domestic GPU, AI server, data center, power equipment, and liquid cooling industry chains may also usher in new growth opportunities. The landscape of China's AI sector is about to change.