US and Chinese AI Giants Escalate Price War Over Large Language Models
Global AI companies, including OpenAI, Google, and Anthropic, are reportedly slashing API prices and upgrading mid-tier models to counter low-cost competition from Chinese rivals like DeepSeek and Moonshot AI's Kimi. For example, OpenAI has significantly reduced the token prices for its entry-level and mid-tier models, such as the ChatGPT 5.6 Luna and Terra. This price war makes advanced AI capabilities much cheaper for developers and businesses, but it threatens the profit margins of major AI firms facing massive computing costs. It highlights a strategic divergence where US firms invest heavily in frontier capabilities while Chinese developers focus on highly cost-effective engineering. While OpenAI slashed the price of its entry-level Luna model by 80% to compete with DeepSeek V4, it maintained or even increased prices for its premium flagship models like Sol Fast. AI companies are betting on the Jevons Paradox, hoping that lower token costs will drive exponential growth in usage to offset the high infrastructure expenses.
## BACKGROUND
DeepSeek is a Chinese AI company funded by the hedge fund High-Flyer, known for releasing high-performing, open-weight large language models at a fraction of the training costs of US rivals. Moonshot AI is another prominent Chinese AI startup that developed the Kimi chatbot, which gained recognition for supporting long context windows.