One long-form read per week:
This week, I read an article about AI. It mainly addresses
“Does AI really have a bubble?”
When AI investments truly become oversupplied, who will ultimately pay the bill?
According to estimates, cumulative investment in US AI data centers, chips, power, and network infrastructure from 2025 to 2032 could reach an astonishing $10.3 trillion. On average, that would be about 3.63% of the US GDP per year. Measured against economic scale, this even surpasses the construction booms in US history—railways, highways, electrification, and communications infrastructure.
So can AI earn enough money to justify all this? If it can’t, then who will be responsible for the money already spent?
When investments rely on cash flow,
when the bubble bursts,
first and foremost, shareholders are harmed.
But when more and more investment starts to be supported by debt, SPVs, private credit, and asset securitization,
the story changes.
Because when the next real turning point in the AI cycle arrives,
what we may need to watch won’t be just the Nasdaq.
It will be—
whether cracks appear first in the credit market.
If they do,
that crack may be the signal that this round of AI’s mega-capex supercycle has truly entered its second half.
Although this kind of risk hasn’t appeared yet,
the risks are becoming harder and harder to see.
$BTC
This week, I read an article about AI. It mainly addresses
“Does AI really have a bubble?”
When AI investments truly become oversupplied, who will ultimately pay the bill?
According to estimates, cumulative investment in US AI data centers, chips, power, and network infrastructure from 2025 to 2032 could reach an astonishing $10.3 trillion. On average, that would be about 3.63% of the US GDP per year. Measured against economic scale, this even surpasses the construction booms in US history—railways, highways, electrification, and communications infrastructure.
So can AI earn enough money to justify all this? If it can’t, then who will be responsible for the money already spent?
When investments rely on cash flow,
when the bubble bursts,
first and foremost, shareholders are harmed.
But when more and more investment starts to be supported by debt, SPVs, private credit, and asset securitization,
the story changes.
Because when the next real turning point in the AI cycle arrives,
what we may need to watch won’t be just the Nasdaq.
It will be—
whether cracks appear first in the credit market.
If they do,
that crack may be the signal that this round of AI’s mega-capex supercycle has truly entered its second half.
Although this kind of risk hasn’t appeared yet,
the risks are becoming harder and harder to see.
$BTC
