Anthony Schiavo

Cutting edge models AI are getting more energy intensive, but non-cutting edge models are not. It's very likely that mass deployment (if it occurs) will be of energy-optimized models. This, plus chip improvements and datacenter design improvements, will keep energy demand down.

There has been an explosive increase in demand for data center capacity in the last five years, but it's overwhelmingly from crypto, mostly Bitcoin. Bitcoin's ~150 TWh of demand represents maybe 40% of global demand and a overwhelming percentage of demand growth.

First, data center energy demand was mostly flat from 2010 to 2019 - this is a function of dramatically improved energy efficiency and scale. Per the IEA:

me when i spend billions on AI that can't get the background of images right

mainstream media organization with maximum left bias

Am I to late to this @mikeduncan? Greetings from Boston

Alibaba: greenwash if you want, we don't give a fuck

Thanks to @BNNbloomberg and @AndrewBellBNN for having me on to talk EVs and li-ion. #lithiumtwitter don't @ me. #luxtake bit.ly/3k2p73N

In our new report, our experts outline the materials innovations needed to help increase production, make masks suitable for a consumer population, and improve sustainability. Download the exec summary to learn more. #LuxTake #research #facemasks bit.ly/35cULp1