Chinese artificial intelligence (AI) models are rapidly gaining ground globally, including in the U.S., because they are significantly more affordable and increasingly efficient than American alternatives. Companies like cryptocurrency exchange Coinbase are switching to Chinese AI models to cut costs. This trend is causing concern among U.S. tech executives and investors, as it threatens American AI leadership and could lead to substantial capital destruction in the U.S. stock market.
While U.S. companies like OpenAI still dominate in consumer chatbots, Chinese models are proving nearly as good, especially for specialized tasks like coding and research. Mozilla's CTO, Raffi Krikorian, switched to Chinese AI startup Moonshot's Kimi K3, finding it "snappier" and more cost-effective than expensive U.S. models like Anthropic's Claude Fable. He also uses Z.ai's GLM-5.2 for routine tasks. Another user noted DeepSeek is "overall the same quality but 17 times cheaper," making it attractive in regions with limited financial resources.
Goldman Sachs indicated that Chinese AI models are reaching a "critical stage" for widespread adoption, driven by the demand for cost-effective solutions for complex "agentic" AI tasks. This shift is particularly problematic for U.S. companies given that AI prices are calculated per million tokens, and the use of AI agents dramatically magnifies cost differences. Jefferies' Christopher Wood warned that financial markets have not fully priced in China's rapid progress in AI, highlighting that top Chinese AI models processed 36.39 trillion tokens on the OpenRouter platform in a week, compared to 7.39 trillion tokens for leading U.S. models during the same period.
The U.S. government's restrictions on Chinese access to American chips and technology, intended to slow their progress, may inadvertently be creating opportunities for Chinese competitors. For instance, restrictions on Anthropic's Fable and Mythos models created an opening for Z.ai's GLM-5.2. This could lead to a "technological Cold War" where countries choose between American or Chinese AI systems, potentially allowing Beijing to influence global standards and disseminate information reflecting its preferred views.
