> A lot of companies that serve a single family of LLMs seem to prefer nvidia though. Why is that?
replyIf you write your tools for CUDA, you’re going to prefer hardware the runs CUSA.
How is there anything more to it than this?
Cool. That's it.
replyWhat will people use to write for Jalapeño matters.
Nvidia has multi-decade heritage. Apple spent almost a decade in MLX. Snapdragon failed partly here. OpenAI announced nothing regarding to that, so this big moat that multiple companies have (nvidia the most prominent) is nil for them.