Those who are currently worried about an AI bubble and doubt whether it will repeat the dot-com bubble burst are because they are unaware of the following facts:

1. Leading AI companies have PE ratios of only 20 to 30, which is moderate or even low; during the bubble period, the PE of industry leader Cisco was 150 — and that was not selling in-demand computing power, but routers and network cables.
2. AI leading companies are continuously achieving new revenue highs, and human demand for AI will only grow. Everything is just beginning; it is still in the early stages of transforming industries and boosting productivity — the adoption rate is still very low. Can you imagine going back to the days without AI? You cannot.
3. AI has only been around for a few years; hardware infrastructure is not yet fully in place, and software breakthroughs, application explosions, and productivity boosts are still on the way. In the future, there will also be developments in robotics, space exploration, medicine, and quantum computing, all related to AI development.
The typical problem of the 2000 bubble was that many companies had no profitability, and telecom infrastructure was severely overbuilt, with fiber optics, switches, and network capacity ahead of actual demand.
And today’s situation is exactly the opposite: computing power, storage, bandwidth, inference costs, model capabilities, and application scenarios are all expanding rapidly. Many demands are present; what’s lacking are infrastructure that is not cheap enough, not abundant enough, or not good enough.
Blindly comparing to a bubble, or just being bearish because prices have risen a lot, are both the result of shallow research and missing out. Toxic thinking that YY (speculative optimism) produces.