Oct 26, 2026 · 1h 3m
AI investment boom tests markets, labor and fiscal resilience
The Most Dangerous Investment Cycle In History | Paul Kedrosky
The episode connects AI’s technological disruption to concentrated revenues, massive capital spending, debt markets and the limits of government support during a downturn.
- 1AI may disrupt cognitive labor faster and more broadly than earlier general-purpose technologies.
- 2Frontier-model spending depends on winner-take-all expectations, concentrated customers and uncertain returns.
- 3As AI commoditizes routine analysis, human value may shift toward judgment about inflection points and novel problems.
Don't miss
Paul Kedrosky’s clearest distinction is that models tend toward the center of a distribution, while human judgment can identify inflection points and unusual developments.
The brief
Paul Kedrosky and Demetri Kofinas compare AI with railroads, electrification and highways, but argue its ability to substitute for cognitive work makes the disruption unusually direct.
The investment race is fueled by expectations of scale and winner-take-all rewards, even as data-center economics, token production and frontier-model returns remain uncertain.
The conversation moves from AI spending’s contribution to US growth toward its effects on valuations, borrowing costs, sovereign debt markets and fiscal capacity.
Kedrosky argues that AI can create value by removing complexity from products and services, while commoditizing routine modeling and research work.
The durable human edge, he suggests, lies less in producing standard analysis than in recognizing inflection points and pursuing important unanswered problems.
What was said on this episode
22 statements · 5 positive · 13 negative · 1 mixed · 3 neutral
AI can replace substantial portions of white-collar tasks and occupations.
“The difference with AI is that it's an incredibly broadly applicable technology that for the most part, at least in its current incarnation, is really focused on symbolic work, what we sometimes call white collar work. And in those domains, it's at least as substitutive as it is additive in the sense that it's very straightforward to see how it can replace tasks and occupations.”
Listen at 10:18
AI substitutes for human work faster than prior general-purpose technologies.
“The two things that are different, it's much more substitutive and it's happening much faster.”
Listen at 11:03
Large AI-driven layoffs and social disruption are effectively built into current adoption.
“Oh, absolutely. I think that's baked in here.”
Listen at 13:20
The current AI cycle combines loose credit, real estate, technology, and government policy risks.
“This is the first one that has all four.”
Listen at 15:56
AI capital expenditure exceeds prior investment episodes as a share of GDP.
“we're now on a, as a percentage of GDP, as a CapEx moment, we're at, we've now exceeded all prior episodes in the last 150 years.”
Listen at 19:03
Data-center spending will approach $10 trillion within four or five years.
“over the next four or five years, we'll probably get to pushing on $10 trillion in data center spending.”
Listen at 20:26
AI infrastructure requires $4–6 trillion in revenue by the early 2030s.
“You get to numbers like four to $6 trillion in revenues needed by say early 2030s.”
Listen at 20:48
AI-related layoffs could reach the low millions.
“it would be unsurprising to see numbers at least in the... you know, in the millions in terms of the low millions, in terms of actual layoffs.”
Listen at 22:01
China is pursuing industrial-scale production of AI tokens.
“China is trying to essentially become like almost an industrial manufacturer of tokens.”
Listen at 25:06
China produces AI tokens at substantially lower prices than US models.
“And they're producing these tokens at a vastly lower price.”
Listen at 25:25
Frontier AI model improvements are slowing toward an asymptote.
“what you'll see over the last three years is while models are improving, it's becoming somewhat asymptotic, meaning that on an absolute level, they're not improving as much as they once were.”
Listen at 29:26
Annual composite-benchmark gains fell from roughly 18–19% to 5–6%.
“we went from like say around 18, 19% performance year over year on composite benchmarks three years ago to more like five or 6%.”
Listen at 29:35
AI capital expenditure accounted for 25–60% of quarterly US GDP growth for seven or eight quarters.
“for at least the last seven, eight quarters, something like, and it's varied from quarter to quarter, 25% to 60% of GDP growth on a quarterly basis has been tied to AI CapEx”
Listen at 33:46
The Magnificent Seven generated nearly 45% of S&P 500 growth over five years.
“They were almost 45% of S&P 500 growth for the last five years.”
Listen at 36:13
Broad index funds are concentrated bets on technology and AI.
“if you think you're buying a diversified index fund, you're actually buying. not just a tech bet, but a concentrated AI bet.”
Listen at 36:31
More than half of US Treasury buyers are now yield-sensitive investors.
“More than half of the buyers of treasuries in buyers in debt markets now, US sovereign debt markets, are yield sensitive.”
Listen at 41:44
Highly indebted companies will prioritize reducing debt rather than borrowing after an AI downturn.
“the project for them is exactly the same as the project for the Japanese companies. who are hugely indebted in real estate after their bubble burst in the 1990s. And so they'll spend that period trying to reduce indebtedness.”
Listen at 44:06
The US no longer has fiscal capacity to spend without significant market consequences.
“The US is no longer simply able to do this and say, we have so much balance sheet space and spending space that we can do this with impunity.”
Listen at 45:59
Governments will choose austerity rather than cash handouts during deflation.
“I absolutely do, yes.”
Listen at 50:05
AI-driven complexity reduction has strongly benefited Cloudflare and tripled its stock over 18 months.
“that's created an incredible tailwind in their business. And so I think the stock has tripled or something over the last 18 months.”
Listen at 52:54
AI commoditizes basic analysis and erodes the edge from straightforward analysis.
“the commoditization of analysis has consequences for anybody who pretends to do analysis. So the notion that you can have an edge from straightforward analysis is largely falling away.”
Listen at 56:39
AI models perform well near distributions’ centers but poorly at their tails.
“models are outstanding at generating the center of the distribution and genuinely terrible at the tails.”
Listen at 1:00:48
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.