Chinese AI Giants Dodge US Restrictions by Training Models in Southeast Asia

Chinese tech giants have found clever ways to sidestep U.S. restrictions on advanced AI chips. Companies like Alibaba and ByteDance are moving their large language model training to data centers in Southeast Asia, particularly Singapore. This strategy helps them access the cutting-edge Nvidia chips they need while remaining within legal boundaries.

These offshore data centers operate fully within current export rules. Singapore-based operators confirm this approach is “an obvious choice” for Chinese firms looking to continue AI development despite U.S. chip supply limitations. The strategy has proven effective, allowing Chinese companies to build competitive AI systems despite domestic restrictions.

Singapore has become the legal backdoor for Chinese AI giants seeking advanced chips beyond U.S. restrictions

The results speak for themselves. Alibaba’s Qwen and ByteDance’s Doubao models now rank among the world’s top-performing AI systems. These Chinese models match the capabilities of U.S.-developed systems while often maintaining cost advantages. Developers worldwide have even adopted Qwen as a free alternative to American models. Current benchmarks confirm that Chinese AI models are now closely matched with their U.S. counterparts in performance and capabilities.

Regulatory gaps have made this possible. When the Trump administration removed the Biden-era “diffusion rule,” it eliminated restrictions on offshore AI chip usage for Chinese companies. Current export controls focus mainly on domestic chip production within China rather than offshore training operations, creating a workable loophole.

Chinese tech firms haven’t limited themselves to Southeast Asia. They’re expanding their data center presence in Middle Eastern regions too. The U.S. ban on Nvidia’s H20 chip for the Chinese market in April 2025 will likely accelerate this offshore training strategy. This multi-region approach provides redundancy and guarantees continued AI development regardless of changing regulations. With 92% of global leaders planning to increase AI investments over the next three years, Chinese companies are ensuring they won’t be left behind in the competitive AI landscape.

There are limitations, however. Chinese companies can’t move private customer data outside China’s borders due to their own government’s data sovereignty rules. This means customization of AI models using client-specific data must still occur within China.

Despite U.S. efforts to restrict China’s access to advanced AI technology, these offshore strategies have allowed Chinese firms to maintain development momentum and produce globally competitive AI models while technically complying with international regulations.