Homepage AI US computing infrastructure accelerates global spread of Chinese AI models

US computing infrastructure accelerates global spread of Chinese AI models

Moonshot AI, Moonshot, AI, Artificial Intelligence, logo, Chinese, AI Tiger, LLM
Samuel Boivin / Shutterstock.com

While Washington tightens export controls on AI chips, American-owned global data centers continue hosting Chinese open-weight models, undermining the Trump administration’s technological dominance goals.

The US Commerce Department recently issued a new rule blocking Chinese companies from remotely accessing advanced AI chips located outside of China. According to Fortune, this regulatory action aims to close a significant export control loophole highlighted by the recent Nvidia-Megaspeed scandal. While the move helps American AI labs maintain their competitive edge, a deeper contradiction continues to undermine Washington’s technological ambitions.

The Trump administration wants the United States to dominate the global artificial intelligence sector from hardware production to software models. A central component of this strategy involves exporting high-end, American-designed chips to data centers across Latin America, the Middle East, and Southeast Asia. However, this aggressive infrastructure expansion inadvertently provides a powerful platform for Chinese artificial intelligence competitors.

American-owned data centers regularly rent their computing capacity to global inference providers seeking reliable hosting solutions. These providers then utilize the American hardware to serve Chinese large language models to users around the world. As a result, the global proliferation of American compute infrastructure is directly helping Chinese model developers gain international traction.

American hardware fuels revenue and soft power for Beijing

Many observers mistakenly assume that the growing user base for Chinese AI models relies exclusively on domestic Chinese microchips. In reality, China currently lacks the manufacturing capacity to produce enough advanced chips to satisfy even its own internal market demand. Instead, American technology giants like Nvidia, Google, and Amazon supply nearly all the AI chips deployed in modern global data centers.

This formidable grip on global cloud services paradoxically allows Chinese open-weight models to reach broader international audiences. American inference providers such as Together AI and Amazon Bedrock actively host top-tier Chinese models for their global enterprise customers. Nvidia even provides dedicated resources to help developers optimize and deploy various open-weight Chinese models on its hardware ecosystem.

Serving these models on American infrastructure accelerates their global adoption and significantly boosts Beijing’s international soft power. Chinese model developers also generate direct financial returns when foreign enterprises scale their usage on these platforms. For example, Moonshot AI reportedly demands a significant revenue share from American cloud providers hosting its popular Kimi K3 model.

Washington urged to compete rather than ban foreign models

The current dynamic largely depends on the stark market division between proprietary models and open-weight alternatives. While American developers dominate the proprietary space, global users often prefer Chinese models when selecting customizable, open-weight solutions. Because users naturally gravitate toward the most capable open-weight options, American infrastructure disproportionately benefits Chinese developers in this specific category.

Despite this strategic challenge, experts warn the Trump administration against banning American cloud providers from hosting Chinese models. Implementing such restrictive measures would severely hinder allied nations seeking to adopt AI and could trigger damaging economic retaliation. Broad prohibitions would ultimately impose excessive costs on global businesses without delivering proportionate national security benefits to the United States.

Instead of pursuing outright bans, policy analysts argue that the United States must aggressively compete in the open-weight market. The federal government should encourage American tech companies to develop cost-efficient open-weight models that can outperform their Chinese counterparts. By directing federal procurement toward domestic open-weight systems, Washington can organically shift global computing infrastructure back toward American software solutions.

Ads by MGDK