I've already checked, and seen two similar capital investment waves: railroads/steam machines in ~ 1890th (and it was even much larger - some sources calculated spending at 6% of GDP) and telecom infrastructure (somewhere 1960s..1990s, spending 1.2% of GDP, just exactly like seen on AI).

And what we see?

- Must admit, some railroads become bankrupt and decades donated from taxes, but not closed, because viable infrastructure; most other suffered from giant wave of Merges&Acquisitions (standard macroeconomic term M&A).

Same wave of M&A happen on telecom market, as I hear from my friends, big business bought small companies, estimated price just on number of clients, and than through all their hardware to landfill (small networks usually built from cheap Chinese SOHO hardware, and instead of routers used PCs) and rebuild all hardware part from scratch, based on telecom-industrial grade hardware.

And really, all modern European transport and telecom infrastructure is from these two waves of huge investments.

Honesty, railways are still moderate profitable (many very significant, like England-France tunnel are unprofitable), and similarly, telecom infrastructure also not much profitable (suburban and rural are in many cases donated), and same problem was with data-centers before AI.

If capital investments into AI will end on current level, this could be considered just second wave of telecom investments - as I said, first wave was huge, but civilization survived and we now actively using fruits of these investments.

If investments into AI will exceed railways level (~5.5x telecom investments), this also will not be end of life but sure need many more profitable AI products than we have now.

Reminds me of stories about salespeople compensation in the book "The difference between God and Larry Ellison (God does not believe he is Larry Ellison)". Oracle salespeople received commissions before customers had paid, and many of them made additional "side" contracts with customers that let them cancel the deal, but were not considered (or known) for the purpose of calculating the commission.

I want to remind everyone here that Cisco in 1993 was $1, right before the dotcom boom. It peaked at $77 or 77x higher in March 2000 with a 200 P/E ratio. Nvidia is at 50 right now.

After the bubble burst, Cisco was at $17 in 2001, still 17x higher than before the boom and 2x 1997.

Is AI bubble peak? Is it 2000 or is it only 1997? The difference is immense.

The point here is that AI valuation can grow multiples more from here. And even after it bursts, it may still be bigger than in 2025.

Another analogy that might prove more apt than 17th century tulip mania is Russian railways at end of 19th / start of 20th century. All the private money going into "sovereign companies" that might be snapped at an instant by respective American/Chinese/Korean/Taiwanese government.