The bear case argues that AI revenue growth is too slow to meet capital-expenditure obligations, growth is slowing, and open-source competition could compress margins. The bull case is that AI labs may eventually stop pretraining while demand for intelligence remains effectively unlimited and margins stay high.