Oct 16, 2026 · 1h 8m
AI spending faces its reality test
Is There an A.I. Bubble? A Debate.
The debate asks whether unprecedented investment will produce durable economic gains or expose a fragile financial structure vulnerable to a crash.
- 1Competitive pressure can drive companies to overbuild AI infrastructure even when individual firms would prefer to slow spending.
- 2The bull case depends on real demand, productivity gains and AI-native businesses, while concentrated customers and debt raise financial risks.
- 3Both sides identify measurable tests, from enterprise spending and IPO disclosures to productivity, employment and labor-force participation data.
Don't miss
Paul Kedrosky and Azeem Azhar define the evidence that would change their minds, turning an abstract bubble debate into a set of economic tests.
The brief
Ezra Klein moderates a debate between Paul Kedrosky, who sees AI investment as an obvious bubble, and Azeem Azhar, who argues its long-term potential may justify the spending.
The argument moves from competitive overinvestment to the harder question of demand: are customers buying enough AI to support the chips, models and data centers being built?
OpenAI and Anthropic become tests of the broader thesis, as the guests debate whether failures at individual providers would destroy demand or simply shift it elsewhere.
Debt, IPO expectations and Anthropic’s customer concentration expose the financial fragility beneath the boom, while Apple appears as a beneficiary of complementary AI services.
The debate widens into productivity, employment and political stability, with AI framed both as a possible engine of growth and as a force repricing cognitive labor.
Their proposed reality checks are concrete: Azeem watches enterprise spending, while Paul wants clear gains in productivity, labor-force participation or other macroeconomic data.
What was said on this episode
43 statements · 10 positive · 27 negative · 6 neutral
A bubble occurs when asset prices detach from economic reality.
“A bubble is whenever asset prices detach from any semblance of economic reality.”
Listen at 2:28
General-purpose technology build-outs historically lead to overbuilding.
“And every one of those moments has led to overbuild.”
Listen at 2:59
AI oversupply fueled by credit will cause business failures and asset write-downs.
“The bubble is this vast oversupply of something, largely fueled now by credit, which in turn will lead to massive business failures and write-down of assets.”
Listen at 3:50
AI-related credit exposure will probably reach $10 trillion.
“it's probably going to hit $10 trillion before it's done.”
Listen at 4:10
U.S. households have 50% of net worth invested in equities.
“we're at record percentages in terms of the percentage of people's net worth that are tied to equity. 50% of people's net worth is now in equities.”
Listen at 4:55
Competitive race dynamics and abundant capital guarantee AI overcapacity.
“And that's what guarantees the overbuild.”
Listen at 7:36
Enterprise AI demand is strong and compute remains undersupplied.
“demand is very strong from American and global enterprises. Not an overbuild yet. There is vast, vast undersupply of the compute needed for AI.”
Listen at 8:54
Chip and power capacity may struggle to meet AI demand over two to three years.
“The race seems to be, can we get enough chips and power enough chips over the next two or three years to keep up with the demand?”
Listen at 9:06
GPU assets currently generate very high returns within 18–24 months.
“people are making extremely good returns in 18 to 24 months on an asset because they're able to charge essentially uber surge pricing for their GPUs.”
Listen at 12:04
Developers account for about 80% of frontier-model token use and revenue.
“something like 80% of the token usage and the revenues from the largest frontier model companies are coders, are developers.”
Listen at 15:24
More than $1 trillion will be invested in AI data centers.
“the trillion plus dollars or so of capital that will go into building data centers”
Listen at 17:37
Enterprise demand can shift to hyperscaler-hosted open-weight models.
“if people can move, enterprises can move to open weight models that are being run directly by a hyperscaler, that is just the demand moving.”
Listen at 19:31
Sustained end-customer demand is necessary for AI investment to avoid failure.
“you're seeing the end customer demand, and that's the thing without which this will always end horribly.”
Listen at 20:36
A historical bubble burst involves investment falling 50% and markets correcting at least 50%.
“we need to see the level of investment drop by 50%, and we need to see a really, really substantial, and typically it's been 50% or more, correction in whatever the relevant market is.”
Listen at 21:54
The combined prospective IPOs of OpenAI, Anthropic, and SpaceX exceed all postwar IPOs combined.
“OpenAI, Anthropic, and SpaceX, if you want to include it, combined, those three IPOs are larger than all post-World War II IPOs combined.”
Listen at 22:38
OpenAI, Anthropic, and SpaceX together exceed peak Nasdaq market capitalization.
“the combined value of those companies is larger than the value of the NASDAQ market capitalization at its peak.”
Listen at 22:53
U.S. equity valuations are at levels seen before the 1929 crash.
“the Schiller cyclically adjusted price earnings multiple, which is sort of a longer-term measure of where U.S. equities sit. We're sitting at the levels we sat just before the crash of 29.”
Listen at 23:42
The anticipated OpenAI and Anthropic IPOs are exit events rather than financing events.
“These are exit events.”
Listen at 26:58
Post-lockup shareholders will sell large quantities of IPO shares.
“You're going to be a wall of selling shareholders, showing up post-lockup, selling it to anyone who will show up and want to take the shares from them.”
Listen at 27:53
Meta represented roughly 10% of Anthropic’s revenue in 2025.
“Meta, something like 10% of their revenues.”
Listen at 30:25
Eight companies may account for 80% of Anthropic’s revenue.
“it might be eight companies that make up 80% of their revenues.”
Listen at 31:03
Apple’s business model does not require operating a frontier AI model.
“Apple has a business model that means they don't need to run a frontier model.”
Listen at 33:13
Apple benefited by relying on other companies’ AI expenditures.
“They got to free ride on everyone else's gigantic expenditures, and they come out the other side smelling like roses because they didn't actually have to make the expenditures.”
Listen at 34:46
AI infrastructure spending is increasingly financed with debt rather than cash flow.
“Spending is moving away from cash flow to debt.”
Listen at 38:21
Oracle and Microsoft are approaching declining credit quality from increased leverage.
“We're seeing that with Oracle. We're seeing that we're on the cusp of that with Microsoft.”
Listen at 38:55
The U.S. technology sector is now the largest investment-grade bond issuer.
“Tech is now the largest issuer of investment-grade bonds in the United States.”
Listen at 39:44
External financing around 30% of investment marks materially higher risk.
“My threshold for more risk was about 30%, which is about the level that we are approaching now at the end of 2026.”
Listen at 42:09
AI investment could produce Dutch disease in the United States.
“there could be a risk that that will happen in the US if it isn't already happening right now.”
Listen at 42:53
Policy stimulus is needed to prevent AI investment from starving other sectors.
“it requires... something of policy, something of a policy stimulus to make sure that those sectors get the activity that they need because I'm not persuaded that market forces on their own would do that.”
Listen at 43:06
Training will add usable construction and technical capacity within one to three years.
“in a year or two or three years, that capacity, those skills will be in the economy and able to be used.”
Listen at 44:29
AI productivity gains will gradually reduce and replace workers unless businesses grow rapidly.
“a large part of it is coming now, I think will come, unless businesses grow very, very quickly, from just a slow reduction in replacement of workers”
Listen at 46:00
Human workers overseeing AI outputs will receive higher compensation.
“those people will end up being paid more and more, better and better as a consequence of sitting at that choke point.”
Listen at 47:53
Broad AI workplace transformation will unfold over five to ten years.
“my time horizon for this runs in five, seven, ten years, not one or two, not two or three”
Listen at 48:59
AI fundamentally reprices cognitive labor.
“its fundamental function is to reprice cognitive labor.”
Listen at 49:38
AI-driven labor disruption will outpace creation of new jobs.
“the disruption will vastly outpace our ability to create new positions.”
Listen at 50:42
There is a 50–60% chance an event will drive AI regulation.
“it's like a 50-60% chance that something happens that drives regulation”
Listen at 54:10
AI regulation is inevitable.
“So no, regulation here is inevitable.”
Listen at 54:21
Industries can grow well under moderate regulation.
“industries grow quite well when there's a little bit of regulation, as the pharmaceutical industry is a great, great example.”
Listen at 54:42
About 2.5% of Americans pay for an AI subscription.
“only about two and a half percent of Americans pay for an AI subscription”
Listen at 59:14
Azeem would become concerned if enterprise AI spending growth halved for two quarters.
“if I saw a couple of quarters where the amount that American enterprises were spending on AI, was the growth rate was halving, say, two quarters in a row”
Listen at 1:03:07
AI has not yet improved U.S. productivity.
“There's no sign of AI improving productivity.”
Listen at 1:03:48
Rising labor-force participation would indicate AI is economically inclusive.
“If we saw participation rates sharply increasing, that would be something that would help to tell me, okay, AI is an inclusive technology in bringing people back into the economy.”
Listen at 1:03:50
Above-trend GDP growth would indicate AI-driven economic gains.
“If we saw GDP growth going back above trend”
Listen at 1:04:00
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.
Books & mentions
Some links are affiliate links — PodLume may earn a commission if you buy.
Middlemarch
The guests recommend this novel for its account of social change, ambition and unintended consequences.
Liberty's Dawn
It offers historical perspective on how technological transformation reshaped work, consumption and ordinary life.
Insel, the Rise and Fall of a Billionaire Utility Tycoon
Its rise-and-collapse story echoes the episode’s concerns about infrastructure, leverage and speculative investment.
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.