They are part of nvidia's profesional offering (rendering, machine learning). They are basically on which every large model has been trained (dall-e 2, GPT3).