Oct 24, 2026 · 14 min
Investors may be looking for the AI bubble in the wrong market
The AI Bubble Isn't Where You Think It Is
The episode examines whether AI spending is destroying capital now—or building a larger, less visible risk beyond public markets.
- 1Strategists and chip executives warn that AI investment has intensified speculation and could destroy capital.
- 2Public-market valuations may not capture the most consequential AI excesses still developing elsewhere.
- 3Nvidia, Anthropic, and other industry players illustrate how infrastructure and model-building drive the investment debate.
Don't miss
The episode’s sharpest turn comes when Ed Elson argues that the most significant AI bubble may still be developing outside public markets.
The brief
Ed Elson frames the AI investment boom as a question of capital destruction, drawing on warnings from equity strategists and chip-industry executives.
The episode shifts attention beyond public markets, arguing that visible stock speculation may not be where the most consequential AI excesses are developing.
Nvidia, CoreWeave, and other infrastructure names anchor the discussion of the spending required to build and run increasingly ambitious AI systems.
Anthropic and CEO Dario Amodei bring the model-building side of the debate into view, connecting AI progress with the scale of investment behind it.
The central takeaway is deliberately unresolved: the public-market bubble may be only the early, visible edge of a larger risk.
What was said on this episode
29 statements · 2 positive · 24 negative · 1 mixed · 2 neutral
Credit markets indicated CoreWeave might soon fail.
“the credit markets were screaming loud and clear, this company might be about to implode.”
Listen at 2:26
CoreWeave borrowed more than it can probably afford.
“the company borrowed more than it can probably afford.”
Listen at 2:35
AI data centers are more expensive and difficult to build than expected.
“data centers turn out to be more expensive and difficult to build than originally anticipated.”
Listen at 3:04
China’s open-weight models cut AI token prices by half.
“Their extraordinarily cheap open-weight models have cut AI token prices in half”
Listen at 3:16
Cheap Chinese models forced OpenAI and other frontier labs to cut prices up to 80%.
“forcing frontier labs like OpenAI to reduce their prices by up to 80%.”
Listen at 3:21
AI development costs are rising while AI prices are falling.
“the cost to build AI is going up. And the price to sell it is going down.”
Listen at 3:28
Warning signs of a neocloud implosion are emerging.
“the warning signs are definitely starting to show.”
Listen at 3:47
Public markets have already priced in AI-implosion risks.
“the risks of an AI implosion are already priced in.”
Listen at 4:25
AI stock valuations have declined despite rising prices.
“valuations have actually come down.”
Listen at 4:44
Data centers account for more than 90% of Nvidia’s business.
“Data centers make up more than 90% of NVIDIA's business.”
Listen at 5:06
Some major AI customers might disappear financially.
“some of their customers might drop off the map.”
Listen at 5:16
Broadcom, Micron, and SK Hynix earnings multiples declined substantially.
“Broadcom's earnings multiple has fallen 54%. Micron's has fallen 56%. SK Hynix's has fallen 49%.”
Listen at 5:29
“CoreWeave is down 52% in the past year. Iron's down 53%. Cypher is down 55%.”
Listen at 5:46
Meta, Alphabet, Amazon, and Microsoft are neither buys nor sells.
“These are no longer buys in my view, but they're not sells either.”
Listen at 6:55
OpenAI and Anthropic are the most overvalued major AI companies.
“The two biggest names in AI are also the most overvalued. OpenAI and Anthropic.”
Listen at 7:14
OpenAI and Anthropic lose billions of dollars annually.
“They're losing billions of dollars a year.”
Listen at 7:44
OpenAI and Anthropic’s business models may never work.
“it's not clear if their business models are ever going to actually work.”
Listen at 7:46
Anthropic’s adjusted figures likely exclude revenue sharing and training costs.
“it very likely strips out revenue sharing agreements as well as their largest expense by far, which is their training costs”
Listen at 8:05
Anthropic’s $2 trillion valuation equals nearly 30 times revenue run rate.
“At a $2 trillion valuation, Anthropic trades at nearly 30 times its revenue run rate.”
Listen at 8:30
OpenAI and Anthropic command exceptionally high valuations.
“These are the kinds of prices only luxury handbag designers can get away with.”
Listen at 8:43
The main AI bubble is concentrated in private AI startups.
“we don't have an AI bubble, we have an AI startup bubble.”
Listen at 9:36
Anthropic’s valuation would halve if publicly traded.
“Its valuation would be cut in half.”
Listen at 9:45
Applying Nvidia’s price-to-sales multiple values Anthropic at roughly $1 trillion.
“the AI Lab's valuation drops to roughly $1 trillion.”
Listen at 9:53
Applying Nvidia’s multiple would halve OpenAI’s valuation.
“Do the same thing to OpenAI, and its valuation gets halved as well.”
Listen at 9:57
VCs valued Thinking Machines at four times Snap’s value despite far lower revenue.
“thinking machines was four times more valuable than Snap, despite generating a 60th of its revenue.”
Listen at 10:16
Private investors valued an unnamed AI company at $32 billion without a product.
“They priced another AI company at $32 billion, despite not having a product.”
Listen at 10:23
Anthropic will go public in one month.
“Anthropic goes public in a month.”
Listen at 10:45
Anthropic filed its prospectus and is reportedly ready for its IPO.
“The company has filed its prospectus and everything is reportedly ready to go.”
Listen at 10:47
Anthropic may delay its IPO because of market uncertainty.
“Delays IPO cites market uncertainty.”
Listen at 11:39
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.