Global enterprises are turning to Chinese open-weight AI models as costs of US artificial intelligence services climb, according to a report by South China Morning Post on the 16th. Companies worldwide are increasingly adopting models from China's Zhipu and DeepSeek instead of OpenAI or Anthropic, driven by the narrowing performance gap between US and Chinese AI capabilities and the substantial cloud computing expenses required for US Big Tech's proprietary systems. This shift represents a growing challenge to the dominance of American AI providers in the enterprise market.
Vercel Reports 50x Surge in Chinese AI Model Usage
San Francisco-based AI web development cloud platform Vercel disclosed that daily token usage of Zhipu's GLM-5.2 model increased 50-fold since mid-June. DeepSeek's V4 Flash (a simplified version of V4 Pro) emerged as the largest single model on the platform. As of the 15th, DeepSeek models accounted for over 20% of total traffic, up from approximately 15% a month earlier. According to Vercel, open-weight models represented 29% of total token usage across its AI gateway platform, nearly tripling in proportion since April.
Chinese Open-Weight Models Cost One-Fifth of US Alternatives
The operating cost of Zhipu's GLM-5.2 model stands at approximately one-fifth the price of Anthropic's Claude Opus 4.8. Proprietary closed models require premium cloud subscriptions and charge per token (the basic unit of data processed by AI), while open-weight models allow free code downloads for execution on local hardware. The rapid performance improvements in these free download models have prompted enterprises to reassess their AI spending. Until recently, companies tolerated expensive US model fees because open-source alternatives lagged significantly, but that gap is narrowing quickly.
Coinbase and Enterprise Clients Experiment with Chinese AI Platforms
Brian Armstrong, CEO of Coinbase Global (NASDAQ: COIN), the largest cryptocurrency exchange in the United States, stated on X (Twitter) last month that the company is experimenting with setting open-weight options such as GLM 5.2 or China AI research lab Moonshot's Kimi 2.7 as default models on its AI platform. The growing adoption of Chinese open-weight models by enterprises demonstrates a shift in how companies worldwide secure AI resources.
Goldman Sachs and Apollo Analysts Assess Market Impact
Goldman Sachs noted in a research report released last week that "Chinese open-weight models are reaching a threshold of intelligent performance compared to proprietary models," adding that "this is significantly accelerating enterprise adoption." Torsten Slok, chief economist at Apollo Global Management, warned in a report last week that "if Chinese models continue to gain market share and token prices continue to decline, the projected cash flows of hyperscalers could prove overly optimistic." The rising preference for Chinese open-weight models raises questions about the future valuation of US AI research labs and cloud computing giants.
FAQ
What are open-weight AI models and how do they differ from proprietary models?
Open-weight AI models allow users to download code for free and run it on their own local hardware, while proprietary closed models require premium cloud subscriptions and charge fees per token processed. Chinese providers like Zhipu and DeepSeek offer open-weight models that cost approximately one-fifth the price of US alternatives like Anthropic's Claude Opus 4.8.
Which companies are adopting Chinese AI models according to the report?
According to South China Morning Post's report on the 16th, Vercel reported a 50-fold increase in Zhipu GLM-5.2 usage since mid-June, and Coinbase CEO Brian Armstrong stated last month that the company is experimenting with Chinese open-weight models including GLM 5.2 and Moonshot's Kimi 2.7 as default options on its AI platform.
