Comment by nixon_why69
21 hours ago
To add on, training + a highly paid American salary load + inference is unprofitable, if and especially if you have deepseek dropping weights and engines for anyone to run.
21 hours ago
To add on, training + a highly paid American salary load + inference is unprofitable, if and especially if you have deepseek dropping weights and engines for anyone to run.
I don't see the average person being able to run DeepSeek's most competitive models though.
H100s are renting out at a few dollars per hour. An awful lot of people could run them if they wanted.
OpenRouter's DeepSeek provider is basically this. Instead of renting H100's for dollars per hour, you pay someone that arbitrages the $/hr to $/token conversion because you can't actually saturate an H100 for hours. It's still effectively the same as running it yourself because of the ability to change providers to someone else arbitraging the same thing, and that there's always the escape hatch of switching to actually renting the H100s yourself.
They don't have to, slightly above-average people can host in Asia for way cheaper than anthropic.
Why would it be cheaper? Electricity might be (slightly, possibly), but H100s aren't