Chinese Models Reshape Global AI Compute Race
Chinese AI startups such as DeepSeek, Z.ai, and Moonshot AI are delivering models rivaling Western equivalents at a fraction of the cost, driving broad adoption as they handle increasing enterprise workloads. Market data shows Chinese open-weight models now process a large share of tokens at roughly one-tenth the cost of US frontier models, with platforms like Vercel and OpenRouter reflecting a shift toward cheaper Chinese options. American and European developers praise the performance and affordability of Chinese models such as Moonshot’s Kimi K3 and Z.ai’s GLM series, fueling uptake while triggering pushback and accusations of distillation from US policymakers. Beijing rebuts illicit copying as US sanctions discussions intensify around access to advanced AI chips and training of foreign models. Major investments accelerate AI infrastructure globally, with NAVER, NVIDIA, and Brookfield planning to expand Korea’s DSX AI factory to as much as 200 megawatts and a roughly $10 billion financing package, underscoring a race to scale compute. Analysts argue cheaper Chinese models reshape the AI landscape while demand remains robust.




