This thread isn't about who wins, it's about the implication that it's too risky to build anything that depends on inference because AI companies are operating at a loss.