Hajra Naz
Top Chinese tech companies are increasingly training their artificial intelligence models outside the country to secure access to Nvidia chips and sidestep tightening U.S. restrictions on advanced semiconductor exports, the Financial Times reported on Thursday.
According to the report, major players, including Alibaba and ByteDance, have begun shifting parts of their latest large language model (LLM) training to data centres in Southeast Asia. Two people with direct knowledge of current operations told the newspaper that offshore training has grown steadily in recent months, especially after Washington expanded its chip export controls in April, blocking the sale of Nvidia’s H20 and other high-end AI processors to China.
The move marks a significant pivot for Chinese AI firms, which have relied heavily on Nvidia’s GPUs to develop cutting-edge models. With access inside China now increasingly limited, the companies are turning to leased overseas data centres owned by non-Chinese operators. These arrangements allow firms to use restricted Nvidia hardware without violating U.S. rules that target domestic access to advanced chips.
One notable exception is DeepSeek, a fast-rising AI start-up that reportedly stockpiled a large number of Nvidia chips before the latest U.S. export bans. Because of this inventory, DeepSeek has been able to continue training its models inside China. The company is also working closely with Huawei and other local chipmakers to improve domestic AI processors and reduce reliance on foreign technology, the FT said.
The broader shift highlights the growing pressure on China’s technology sector as it tries to keep pace in the global AI race. Washington’s export controls have made it difficult for Chinese companies to access state-of-the-art GPUs, forcing them to explore alternative strategies—from offshore training to accelerating domestic chip development.
Analysts say these workarounds may offer short-term relief, but long-term dependence on overseas facilities raises concerns around cost, data governance, and operational security. At the same time, Beijing is pushing aggressively to strengthen its semiconductor ecosystem, viewing AI capability as a strategic priority.
