Home / Technology / AI Price War: US Labs Cut Costs Amidst Chinese Competition
AI Price War: US Labs Cut Costs Amidst Chinese Competition
14 Aug
Summary
- US AI labs are significantly reducing prices for their models.
- Chinese AI companies are gaining market share with cheaper alternatives.
- Companies are curbing AI usage due to rising computational costs.
US artificial intelligence laboratories, including OpenAI and Anthropic, are now offering more affordable models to retain customers who are increasingly opting for lower-cost alternatives from Chinese competitors. This price war has seen GPT-5.6 Luna's price drop by 80% and Claude Opus 5 launched at half the price of its predecessor.
This development comes as rising AI expenses push businesses to limit usage and seek out more economical solutions. Chinese developers such as Moonshot and DeepSeek have consequently gained traction with users globally. The price cuts by US firms have decreased customer costs by nearly a quarter since mid-July.
OpenAI and Anthropic are reportedly pursuing initial public offerings at trillion-dollar valuations, intensifying the need to demonstrate returns on significant AI investments. Some corporations, like DoorDash and Airbnb, have already begun incorporating Chinese-made AI models to manage their expenditures.
These price adjustments primarily affect mid-tier products, making them more competitive with Chinese offerings. For instance, OpenAI reduced input token costs for GPT-5.6 Luna from $1 to $0.20 per million, and output token costs from $6 to $1.20 per million. Anthropic's Opus 5 is priced at $5 per million input tokens and $25 per million output tokens.
However, comparing AI model costs based solely on token prices can be misleading. More advanced models may complete tasks with fewer tokens or attempts, potentially leading to lower overall costs. Furthermore, different "effort" settings on these models can influence computing power, performance, and final expenses.
Recent benchmarks indicate mixed performance and cost-effectiveness. Anthropic's Opus 5 at "medium" effort showed comparable results to Moonshot's Kimi K3 at "max" effort. OpenAI's GPT-5.6 Luna at "max" effort was similar to DeepSeek's V4 Flash at "max" but nearly twice as expensive per task.
Despite the competitive landscape, some experts note that prices for the very top-tier models remain stable or are increasing. The current pricing changes represent a strategic move by US labs to defend their premium offerings while reducing costs in the middle market.
