***
The Great Tech Exodus: China’s AI Giants Move Overseas to Secure Critical Nvidia Chips
In the global race for artificial intelligence supremacy, China’s biggest tech firms are deploying a creative, if complex, new strategy: taking their most sensitive work outside the country. A new report reveals that top Chinese companies are moving the training of their powerful AI models overseas, all in a bid to bypass strict US export controls and gain access to the world’s most advanced computing hardware: high-performance chips from Nvidia.
The chips in question are the core engine of modern AI, and the fastest ones—like Nvidia’s top-tier H100, H200, and the newer Blackwell series—are now virtually unobtainable for mainland Chinese entities due to Washington’s ongoing crackdown on advanced technology exports. These restrictions are designed to curb China’s progress in cutting-edge AI, citing national security concerns. The result? A significant gap in computing power that Chinese companies must close if they hope to compete with rivals in the US and elsewhere.
The current US regulations mean that the most powerful Nvidia chip legally available in China is the H20, a scaled-down model that is considered a full generation behind the chips that Silicon Valley companies are using today. This difference in processing ability can be massive, potentially giving users of the latest unrestricted chips up to 15 times more computational power. To compete, Chinese tech behemoths like Alibaba and ByteDance are finding ways to innovate around the regulations, not just within them.
The overseas strategy hinges on using data centers in friendly foreign jurisdictions. While the specific locations are varied, reports suggest a heavy focus on Southeast Asia and the Middle East, regions that have also become central to China’s “Digital Silk Road” infrastructure policy.
The operational tactics are fascinating. For some, it means physically transporting hard drives full of training data to foreign data centers—sometimes in places like Malaysia—where they can leverage the powerful, unrestricted Nvidia hardware. Once the intensive training is complete and the AI model has been significantly upgraded, the improved software is then brought back to China.
Other companies are relying on global cloud service providers, including Google and Microsoft, who offer access to high-end Nvidia chips through their data centers located outside of China. This allows the firms to essentially rent the cutting-edge power they cannot buy outright, all while keeping their hands clean of the direct import violation.
This great tech exodus underscores the high-stakes nature of the global AI race. For Chinese companies, securing access to the world’s best silicon is not a matter of convenience; it’s an existential imperative for remaining competitive on the world stage. As geopolitical tensions continue to reshape the tech landscape, the world is watching to see if these creative, cross-border maneuvers will be enough to keep China’s AI ambitions on track.
***