China’s Low-Cost AI Models Force Corporate America to Rethink Sky-High AI Bills
U.S. companies are cutting AI costs by switching to cheaper Chinese models, reshaping the market and challenging premium providers like OpenAI and Anthropic.Summary
Corporate America has begun cutting AI expenses by replacing expensive U.S. models with cheaper Chinese alternatives, a shift driven by stark cost differences in data‑center construction, electricity, and hardware. A 400‑megawatt plant in China costs about $2.4 billion versus $4 billion in the U.S., and Chinese operators pay roughly 25 % less for comparable racks, saving hundreds of millions on power alone. These savings have led firms such as DoorDash, Airbnb, Siemens, and Lindy to adopt models like DeepSeek‑V3, Kimi‑K2, and GLM‑5.2, which deliver performance within a few percentage points of premium U.S. systems while costing 60‑90 % less. The trend is accelerating as Chinese models reach 30 % of token share on OpenRouter and 15 % of global usage by late 2025, while U.S. leaders like OpenAI and Anthropic face valuation pressure and must compete on price as well as capability. Geopolitical tensions add uncertainty—security concerns and potential U.S. bans loom—but open‑weight releases from both sides encourage experimentation over lock‑in, and many companies now mix models to balance cost, performance, and risk.
(Source：Webpronews)