Chinese Tech Giants Move AI Model Training Abroad

Reuters – Top Chinese firms are shifting artificial intelligence model training to offshore locations to bypass U.S. export restrictions on advanced Nvidia chips, according to a Financial Times report cited by Reuters. This move comes after the U.S. tightened controls in April, limiting sales of Nvidia's H20 chip to China. Companies are relying on lease agreements with non-Chinese data centers abroad to continue development, highlighting Beijing's push to advance AI capabilities amid escalating tech tensions.

While most firms face hurdles, DeepSeek stands out as an exception, having stockpiled Nvidia chips before the bans and training its models domestically. DeepSeek is also partnering with local manufacturers like Huawei to optimize next-generation Chinese AI chips, reducing long-term dependence on foreign technology. This domestic collaboration underscores China's strategy to build self-reliant semiconductor ecosystems.

The trend reflects broader U.S.-China rivalry, with Washington using chip restrictions as leverage in trade talks to curb Beijing's military and AI advancements. Despite these efforts, Chinese tech giants' offshore workaround demonstrates resilience, potentially accelerating their AI progress while pressuring Nvidia's market dominance in the region.