Comment by BoorishBears

Why do you think an inference provider competing for the same compute as OpenAI and Anthropic gives up those margins rather than giving a modest discount over the frontier for near-frontier performance?