“Jessica asserts that the inference phase of large language models is as computationally intensive as the training phase.”