Urgent.News

What's breaking now, across hundreds of outlets.

AI

Data centers full of U.S. chips are serving Chinese AI models–undermining Trump’s tech ambitions

The same infrastructure that secures America’s advantage in chips and cloud services gives Chinese AI models a platform to spread.

Data centers full of U.S. chips are serving Chinese AI models–undermining Trump’s tech ambitions

The U.S. Commerce Department's recent rule addressing China's remote access to AI chips located outside of the country marks a crucial step in closing a significant loophole in export controls. However, this progress is overshadowed by a paradox within Washington's AI strategy: U.S. infrastructure, primarily composed of chips designed by companies like Nvidia, is being deployed in data centers across Latin America, the Middle East, and Southeast Asia.

Consequently, these data centers are serving Chinese AI models, inadvertently bolstering their global traction.

While the U.S. aims to dominate AI from chips to models, its infrastructure inadvertently assists Chinese AI model development. Companies such as Together AI, Baseten, and Fireworks are serving Chinese models to users through infrastructure powered by U.S. chips. For instance, DeepSeek's latest model is being offered by inference providers headquartered outside China, including Amazon Bedrock. This situation accelerates Chinese models' global adoption, benefiting China's soft power and generating revenue for its AI labs.

The contradiction at the heart of the Trump administration's AI dominance ambitions lies in the fact that success in one layer of the AI stack undermines its goals in another. Although China cannot produce enough advanced AI chips, the U.S. grip on AI infrastructure is formidable, with companies like Nvidia, Google, AMD, and Amazon designing nearly all deployed data-center AI chips. Yet, the same infrastructure that bolsters America's advantage also provides a platform for Chinese AI models to spread.

The Trump administration should focus on developing competitive U.S. open-weight models and identifying government uses suited to these models for federal procurement. This approach could limit Chinese revenue from overseas deployments and reduce China's soft power gains from its models' global reach. The administration should avoid banning U.S. cloud providers from allowing non-Chinese companies to serve Chinese LLMs to avoid unnecessary costs and potential Chinese retaliation. Instead, promoting open-weight models could be a more effective strategy for U.S. AI preeminence.

Rewritten by urgent.news; may contain errors — check the original.

Also reported by 1 other outlet

Every outlet that reported this, and when →

Read the original at fortune.com →

Report a problem with this story

More in AI

More from Saturday 21 November →