2 ms·

Adding more 9s is costly, and AI training is very suitable to be throttled and/or interrupted. I'm not talking about days or weeks of downtime, but these things

Adding more 9s is costly, and AI training is very suitable to be throttled and/or interrupted. I'm not talking about days or weeks of downtime, but these things are definitely being considered. Source: I'm working at a Google datacenter. See e.g. this post from Urs Hölze, one of the fathers of hyperscale computing: https://www.linkedin.com/posts/urs-h%C3%B6lzle_rethinking-load-growth-assessing-the-potential-activity-7296276501052760064-r1qa https://www.linkedin.com/posts/urs-h%C3%B6lzle_rethinking-lo...

myrmidon 1y ago
Hm... I'm still not really buying the "turn datacenter off during peak electricity demand" scenario at all, because the ratios just don't seem credible to me: Assuming ~$10M of capex (to buy the datacenter) per MW of electrical power (required by the datacenter), and hardware that is obsolete after 5 years (or even 10!), turning that datacenter off for an hour just to save like ~50$/MWh (or whatever spot price is) seems extremely counterproductive, because your hardware running for that additional hour is worth multiple times that (you spent like >$100 per operating hour on the hardware alone assuming 10 year lifetime). It seems much more attractive (and credible) to just install more batteries (or even a gas turbine), instead of chasing demand-side-regulation pretensions. edit: thx for the link though, that is a very interesting study/data even if I disagree with that conclusion!