Chinese AI Models Rattle Markets, Challenge U.S. Leaders

China has steadily narrowed the artificial-intelligence gap despite U.S. restrictions on advanced chips, leaning on cheaper, open-weight models that developers can download and adapt. DeepSeek’s low-cost breakthrough rattled markets in 2025, challenging assumptions that frontier AI requires ever-larger computing budgets. New systems from Beijing-based Moonshot AI and Alibaba now threaten the pricing power of OpenAI and Anthropic and raise fresh doubts over the hundreds of billions of dollars U.S. technology companies are committing to chips, data centers and energy infrastructure.

Moonshot AI unveiled the 2.8-trillion-parameter Kimi K3 on July 17, saying it trailed only OpenAI’s GPT-5.6 Sol and Anthropic’s Claude Fable 5. It priced output at $15 per million tokens, versus about $30 for Sol and $50 for Fable, and temporarily halted new subscriptions as demand strained capacity. Alibaba previewed its 2.4-trillion-parameter Qwen3.8 on July 19 and said open weights would follow. The Nasdaq fell 1.4% on July 17, while Nvidia lost 2.2%, as investors reassessed the economics of U.S. AI spending.