Chinese low-cost open models pressure cloud giants' AI returns, Jefferies and Bain warn
TechNews reports that analyses from Jefferies and Bain argue the rise of low-cost Chinese and open-source AI models is driving down token prices, threatening US cloud providers' ability to recoup trillions of dollars in data center investment. Jefferies estimates roughly half of the about $2.3 trillion in contracted backlog held by Microsoft, Alphabet, Amazon and Oracle comes from OpenAI and Anthropic, while Chinese models now process 76.12 trillion tokens a week versus 4.37 trillion before late April, against just 14.2 trillion for the top four US providers; it estimates Moonshot AI's open-source Kimi K3 reaches 95% of Anthropic's flagship Claude Fable 5. Bain says the AI industry must generate $6 trillion in annual revenue by 2031, but existing AI application revenue is only about $1.8 trillion, leaving a $2.7 trillion gap even after adding $1.5 trillion from search ads, autonomous driving and AI simulation.
If cheap models keep compressing token pricing, cloud capex payback and the debt-financed AI infrastructure buildout both come under pressure.