“A hypothetical foundation model costing a trillion dollars to train is not guaranteed to be significantly more valuable than one costing $100 billion.”