The data center buildout echoes the late 1990s fiber frenzy. Companies strung cable across continents for bandwidth nobody needed, on the assumption that demand would only go up and that the underlying technology would not change fast enough to matter. Both assumptions proved wrong. The chips will get faster. The models will learn to do more with less. The Excel chart pointing up and to the right rarely survives contact with Moore’s Law.
These are physical products, as much as they might seem like digital ones. You can’t put up walls faster than a query on ChatGPT. Neither can you get energy sources revved up on demand. Physics is physics, and atoms are atoms. It just doesn’t make for a good announcement.
I worked for Qwest in 2002-2002. This is exactly like that.
Yeah, there seems to be an awful lot of history repeating itself here…