The AI arms race between Washington and Beijing has entered a new phase, and China is no longer playing catch-up. A wave of high-performance, dirt-cheap AI models from Chinese developers has effectively closed the technology gap with their American counterparts, giving President Xi Jinping significant new leverage over how global technology rules get written.
The cost equation that changes everything
DeepSeek, one of China’s most prominent AI labs, released its R1 and V3 models with pre-training costs estimated at roughly $5.6 million. To put that in perspective, training a frontier AI model in the US typically runs into the hundreds of millions, sometimes billions, of dollars.
DeepSeek isn’t alone. Companies like Z.ai and Alibaba have rolled out their own open-weight models, including V4 and GLM-5.2, that perform competitively with top-tier US systems. Chinese models now rank among the top globally in performance benchmarks. The gap that Washington once counted on as a strategic buffer has, for practical purposes, disappeared.
Beijing’s dual strategy: export and restrict
