[U.S. Companies Shift to Low-Cost AI Models, Open-Weight Models Accelerate Penetration] Jin10 Data, September 28 – According to the Financial Times, U.S. businesses are turning to lower-cost "open" AI models. The so-called "open-weight" models allow users to customize and run them on their own hardware; compared to using the most advanced paid versions of ChatGPT or Claude, the cost for companies to operate these models is significantly lower. Data from research platform AlphaSense shows that in August and September, mentions of "open-weight" or "open-source" models by corporate executives during earnings calls and investor conferences surged sixfold compared to the same period last year. The open-weight AI market is currently dominated by Chinese companies, but French company Mistral, as well as U.S.-based Nvidia, Reflection AI, and Thinking Machines Lab, are also developing powerful models based on similarly permissive licensing terms. Usage data from AI supply chain-related companies indicates that this shift is not just theoretical. In August, open-weight models accounted for 56% of all tokens processed by Vercel AI Gateway, compared to only 7% in December of last year.