If GLM-5.3 was truly trained and ran on Chinese chips, we'll have to ask what the business model of Nvidia itself is soon :)