Hacker News new | ask | show | jobs
by throwfaraway4 1 day ago
I think this ignores the potential impact of local AI across 2+B devices that no longer need SOTA models for the majority of use-cases. What does that do to datacenter demand?
1 comments

It’s not like Nvidia can’t get into local LLM market.

As a local LLM enthusiast, I highly doubt local LLMs will matter for many more years. There will always be extremely high demand for the smartest cloud models.

Local LLMs will never be as efficient in $/token/intelligence.