Oct 14, 2026 · 1h 12m
AI supply wave could pressure crowded markets
The AI Bubble Will Pop from Equity Supply Says This $80 Billion Asset Manager | Ben Inker | GMO
The episode connects future IPO issuance, AI overbuilding, credit stress, and concentrated valuations to a potentially severe market correction.
- 1New equity from AI-linked companies could pressure markets as insider unlocks and IPOs increase supply.
- 2Overbuilding may create stress across data-center finance, private credit, and hyperscaler infrastructure even if AI demand persists.
- 3Resources, value stocks, and emerging markets offer potential diversification when mega-cap technology valuations and profits face scrutiny.
Don't miss
Inker’s central warning is that an AI downturn could be driven by simultaneous equity issuance and overbuilt compute, rather than a simple collapse in demand.
The brief
Ben Inker of GMO argues that transformational technologies often produce speculative bubbles, and AI may be vulnerable to cheap, interchangeable services colliding with enormous investment.
The potential IPO wave matters because new shares and insider unlocks could create unusual equity supply, echoing issuance patterns before the market difficulties of 2000 and 2021.
The risk extends beyond stocks: data-center borrowing, private credit, and bank financing could suffer if AI projects become uneconomic or compute becomes oversupplied.
Inker also questions whether mega-cap earnings and margins can remain exceptional, while warning that memory-chip profits and other cyclical earnings may already reflect peak conditions.
His portfolio response favors diversification through resources, value stocks, and emerging markets, where profits have improved without the same degree of valuation expansion.
What was said on this episode
46 statements · 7 positive · 34 negative · 5 neutral
Compute oversupply would cause defaults, bankruptcies, and poor hyperscaler returns.
“Massive defaults on the credit side. We will see companies on the equity side go bust. We will see a problem for the hyperscalers where their return on investment turned out to be lousy.”
Listen at 0:29
An AI compute oversupply shock could reduce the market 35–50%.
“That is a minus 35 to 50% in the market.”
Listen at 0:48
Transformational technologies such as AI normally generate speculative bubbles.
“Why do we think AI is a bubble? I'd say kind of at the highest level, it is because it is a transformational technological change. Transformational technological changes normally create bubbles.”
Listen at 1:58
Most investment in transformational technology eventually earns poor returns.
“most of the money that gets invested doesn't have a great ROI.”
Listen at 2:42
Excess capital eventually causes supply to exceed demand.
“eventually, as everybody who participated in the building of that supply seems to be getting rich and more and more people want to be involved, you eventually hit a point where now supply. is bigger than demand.”
Listen at 3:32
AI compute and model provision are more likely competitive than winner-take-all businesses.
“what we see in both cases is those look more likely to be competitive businesses than winner-take-all businesses.”
Listen at 4:47
Broad AI adoption requires low prices, which makes high profitability difficult.
“If we are going to get really broad use of artificial intelligence, it needs to be cheap. And the problem with it being cheap is it is hard for something to be cheap and be wonderfully profitable.”
Listen at 5:55
AI is likely to prove to have been a bubble.
“we think this is likely to... prove to have been a bubble.”
Listen at 6:08
SpaceX, Anthropic, and OpenAI are expected to become historically large public offerings.
“SpaceX soon to be Anthropic and then soon to be OpenAI coming public as the largest IPOs in history.”
Listen at 7:54
About $2 trillion of additional SpaceX equity will become available and partly be sold within 12 months.
“over the next 12 months, the other $2 trillion worth of SpaceX becomes available to sell. And a decent chunk of that will be sold.”
Listen at 8:21
Nearly all SpaceX holders will have selling opportunities over roughly 12 months.
“It is going to be the case that over the 12 months or so, almost everybody who owns it is going to have the opportunity to sell.”
Listen at 10:31
SpaceX, OpenAI, and Anthropic IPOs could equal 6–7.5% of U.S. market capitalization.
“in aggregate, we are talking about assuming OpenAI and Anthropic do come public, maybe six, seven, seven and a half percent of the aggregate market cap of the U.S. which is a greater increment of market cap via IPO than we have ever seen in a 12-month period.”
Listen at 10:44
Each percentage point of IPO supply is associated with roughly 4% worse market performance.
“In general, the market does about 4% worse per percent of market cap that comes in in an IPO in a given year.”
Listen at 11:47
Six percent IPO supply would imply market performance 24% worse than normal.
“if we are seeing 6% coming in, you know, that's 24% worse than normal.”
Listen at 12:00
AI bubble unwinding would most affect stocks that benefited from AI.
“if AI is a bubble that bursts, it is the stocks that benefited from AI going up.”
Listen at 13:24
AI beneficiaries face the greatest downside if AI enthusiasm reverses.
“the companies that have benefited from AI are the companies that have the most at risk.”
Listen at 13:57
Downside from an AI reversal depends substantially on valuation.
“the impact should certainly depend on how expensive you are”
Listen at 15:08
AI data-center investment will eventually be excessive and produce large write-offs.
“in the end, we will wind up investing too much in these data centers and there will be giant write-offs.”
Listen at 18:59
Investing in data centers currently carries substantial risk.
“To be investing in data centers today, I think it's a slightly scary thing to do.”
Listen at 20:15
Investors should prefer many alternatives to the AI sector currently.
“we think there are lots of better places to be investing today than in the AI space.”
Listen at 20:26
Investors should not concentrate their portfolio in AI alone.
“you do not want your portfolio to be a bet on a single thing.”
Listen at 23:45
Passive investing grew from under 10% of the market to approximately half.
“the rise of passive has been huge. It went from under 10% of the market in the 80s to maybe half now.”
Listen at 24:36
AI-related debt is increasingly constrained by lenders’ need for diversification.
“we are starting to hit the situation where the need for the diversification is biting in AI-related debt.”
Listen at 32:25
Planned AI investment could create a severe concentration problem in corporate debt.
“if we get all of the investment that is slated to come, this will become a profound problem”
Listen at 34:28
AI borrowing is starving non-AI long-duration investments of capital.
“long dated investments that are not directly powering data centers is all being starved.”
Listen at 36:15
Providing capital to homebuilders can be profitable when housing investment is constrained.
“supplying capital to the companies building houses actually turns out pretty well.”
Listen at 36:55
Government AI bailouts could destabilize the government bond market.
“I don't know where the money comes from to do this without destabilizing the government bond market”
Listen at 38:41
The Federal Reserve could substantially lower capital costs during a severe recession.
“the Federal Reserve, from where rates are now, or at least where rates are going to be in 2027, had plenty of room to lower the cost of capital if the economy comes to a screeching halt.”
Listen at 40:23
Microsoft is not expected to default on its debt.
“I am not expecting Microsoft is going to have to default on its debt.”
Listen at 41:19
Public credit markets currently show little evidence of distress.
“right now we don't see a lot of evidence of distress in the public credit markets.”
Listen at 42:01
High diesel prices raise transportation and agricultural costs.
“The price of diesel is really high everywhere. That really increases the cost of getting goods from place to place. It increases the cost of agriculture.”
Listen at 43:48
Investors should overweight resource and high-quality stocks absent other information.
“groups of stocks that I would be over allocated to if I knew nothing else. One would be the resource space. The other would be high quality stocks”
Listen at 47:39
Recent AI-related profits in Korea and Taiwan are not fully sustainable.
“we do not believe that those profits are fully sustainable.”
Listen at 51:40
Reduced U.S. competition has increased profits in less competitive industries.
“The U.S. has become a much less competitive place in most industries and the places where there's less competition. It's been great for profits.”
Listen at 54:11
Data-center services are difficult to differentiate substantially.
“data centers is inherently a undifferentiated business. I mean, not entirely undifferentiated, but a difficult to profoundly differentiate business.”
Listen at 55:10
AI model providers currently have essentially no customer lock-in.
“The lock-in is completely non-existent.”
Listen at 55:59
AI model delivery is currently highly competitive.
“that also feels like a pretty competitive industry.”
Listen at 56:11
The largest U.S. companies are positioned to keep dominating their industries.
“those companies are very well positioned to continue dominating the industries that they currently dominate.”
Listen at 57:45
The largest stocks generally underperform over long historical periods.
“in general, the largest stocks underperform.”
Listen at 58:12
Historically, peak-profit memory-chip makers are poor purchases.
“Historically, this would be a bad time to buy a company like the Memory Makers.”
Listen at 1:02:18
Falling memory-chip margins could cause memory-chip revenue to decline.
“if their margins come down, their margins are so high that their revenue will be at risk.”
Listen at 1:03:13
Memory-chip companies will soon become unattractive investments.
“My guess is we're going to see the other side soon enough that these companies won't be a great investment.”
Listen at 1:04:07
Approximately half of current AI compute spending may fund model training.
“if half of the spend right now is on training of new AI models.”
Listen at 1:06:22
The AI industry may have overspent on model training.
“we spent way too much money on training AI models.”
Listen at 1:06:32
AI compute oversupply would trigger widespread credit defaults and company failures.
“we will have massive defaults on the credit side. We will see companies on the equity side go bust.”
Listen at 1:08:07
Poor data-center returns would hurt Microsoft, while CoreWeave could fail and Supermicro might fail.
“Microsoft will go down if their return on investment on the data centers was lousy. but Microsoft will be okay. CoreWeave will not. Supermicro may not be.”
Listen at 1:08:51
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.