That crash was due to laying out too much capital in the form of fiber that wasn’t being used. The difference today is that every GPU added is immediately maxed out, and if it’s not it’s because the power plant for the area is maxed out.
That crash was due to laying out too much capital in the form of fiber that wasn’t being used. The difference today is that every GPU added is immediately maxed out, and if it’s not it’s because the power plant for the area is maxed out.
And the GPUs are being maxed out because money seeking a high return is being thrown at LLMs in the hope that they turn into AGI, because there's no other market segment promising such high returns... and being believed.
I suspect the bubble is self-sustaining, because everyone making investment decisions is confident that they won't be blamed if (IMHO, when) the bubble collapses, because "everyone was investing in AI" while simultaneously, even those who might be skeptical are terrified of missing the boat if LLMs really do live up to their hype, or at least convinced they can ride the bubble and find a greater fool before it pops.