I'm not sure when the AI bubble peaks, but I think its wrong to tie it to a specific timeframe during the dotcom bubble. I think what is relevant is when the capex spending starts assuming growth rates that are too optimistic. There isn't a rule that says it must be a certain amount of calendar time. In the case of CSCO the company peaked a bit before growth rates in internet usage started to slow. I think things here will probably unwind when growth rates slow, and they recognize they were too optimistic about future demand. As of now they don't even seem to be considering costs. I think Zuckenberg was on record saying something to the effect that investment levels don't make sense, but the downside of missing out is too great ie corporate FOMO.