China has made a significant breakthrough in AI development, launching GLM-4.5, a highly competitive model that outperforms Western counterparts at a fraction of the cost. This achievement challenges the effectiveness of America's strategy to curb China's AI growth through export controls. Critics argue that these restrictions won't stop China from innovating and will instead push Chinese companies to develop their own chips, filling the supply void left by US restrictions.
(link unavailable)'s success despite US sanctions exposes flaws in America's AI strategy. The company's GLM-4.5 model matches or exceeds Western standards in coding, reasoning, and tool use, running on only eight Nvidia H20 chips. This is better performance than DeepSeek with about half the hardware. (link unavailable)'s achievement proves that sanctions won't stop Beijing's AI advancement.
The US chip export controls may be inadvertently boosting China's AI development by encouraging self-sufficiency. China's AI models, such as GLM-4.5, are not only competitive but also significantly cheaper, potentially giving them an edge in the global market. The US and China have different approaches to AI governance, with the US focusing on rule-setting and China emphasizing national sovereignty and UN-led globalism.
Experts warn that the US needs to streamline approvals, speed up reindustrialization, and rebuild large-scale computing capabilities to maintain its competitive edge. Building a Western AI supply chain with Latin America could counter China's AI Belt and Road Initiative. The Trump administration's recently announced AI Action Plan emphasizes that US strength lies in scaling supply and adoption abroad, not retreating.
