How OpenAI Could Cause a US Recession With Steve Eisman

Anthropic

Across the shows: 18 speakers in other episodes, 9 bullish and 27 bearish calls, past 30 days

• Steve Eisman · The Real Eisman PlaybookBearish

  Anthropic's revenue growth is decelerating from 140% to 43%, and that pace is unsustainable.

  A report said Anthropic's revenue rose 43% to $16.8B for the quarter, down from 140% growth in the prior quarter. Eisman says the earlier pace was never sustainable, and he links the slowdown to the end of token maxing, when companies reviewed their budgets and cut AI spending. He thinks the deceleration is the start of the impact.

  “Right, so it jumped 43% in three months, but that's down from a growth rate in the second quarter of 140%. Well, we know that's not sustainable.”at 17:05

• Steve Eisman · The Real Eisman PlaybookBearish

  Anthropic should IPO before Q3 because token-maxing will make its reported numbers look bad.

  Eisman argues Anthropic needs to go public now because heavy token consumption ('token maxing') will make its financial numbers look increasingly weak. He frames this as a conspiracy theory that Anthropic wants to list before it reports third-quarter results, implying the financials will deteriorate and the IPO window matters.

  “I think they actually need to go public right now because the whole token maxing is gonna make their numbers start to look really bad.”at 37:26

NVIDIA

Across the shows: 11 speakers in other episodes, 11 bullish and 4 bearish calls, past 30 days

• Steve Eisman · The Real Eisman PlaybookUncertain

  Nvidia's growth is astonishing, but 70% of receivables from five customers is a concentration risk.

  Nvidia's revenue more than doubled last quarter, which Eisman calls awe-inspiring for the largest company on Earth. But the 10-Q says 70% of its accounts receivable (a proxy for revenue) comes from five customers, who are probably the hyperscalers. Eisman compares this to a 1990s software company with only two customers, which his analyst refused to take public because losing one would end it. He says the hyperscalers are themselves dependent on Anthropic and OpenAI, so the risk is concentrated in the AI labs.

  “Nvidia grew revenue this most recent quarter by over a 100%... But in the 10 Q, there's a sentence that says that 70% of Nvidia's accounts receivable, which is a proxy for revenue, is from five customers.”at 12:19

CoreWeave

Across the shows: 5 speakers in other episodes, 0 bullish and 6 bearish calls, past 30 days

• Michael Batnick · Ritholtz Wealth ManagementBearish

  CoreWeave spends $640M a year servicing debt, which is problematic given its weak cash flow.

  Batnick says CoreWeave's debt has blown out, with yields around 15%. It is now spending $640M just to service that debt, which he calls extremely problematic given its cash flow. CoreWeave is a neocloud with no advertising cash flows like Alphabet, so it has to keep borrowing to fund its data centers. Its stock is down about 60%, and insiders have been selling. The debt stress is a warning sign for the AI build-out financed by debt.

  “They're now spending $640,000,000 just to service the debt, which is extremely problematic given their cash flow situation.”at 14:33

OpenAI

Across the shows: 14 speakers in other episodes, 4 bullish and 12 bearish calls, past 30 days

• Steve Eisman · The Real Eisman PlaybookBearish

  OpenAI's reported annualized revenue figures are unreliable, since $70B dropped to $50B in a month.

  OpenAI said its annualized revenue at the end of September was $50B, after saying a month earlier it was $70B. Eisman asks where the $20B went and says the figures are not apples to apples. He also says he has no idea what annualized revenue means and that nobody does. He treats the reporting as unreliable and a sign of weakness.

  “OpenAI came out and said that their annualized revenue at the end of September was 50,000,000,000. A month ago, they said 70. What happened to 20? Where'd the 20 go? Disappeared.”at 16:28

Oracle

Across the shows: 9 speakers in other episodes, 0 bullish and 8 bearish calls, past 30 days

• Steve Eisman · The Real Eisman PlaybookBearish

  Oracle would collapse in an OpenAI-led price war, since half its backlog is OpenAI.

  Eisman's scenario is a full price war in which Anthropic's revenue growth turns negative and it cuts its compute commitments. Since half of Oracle's backlog is from OpenAI, Oracle would be badly hurt. He recalls that Oracle's stock went from $2.30 to $3.30 in two days when its backlog jumped by about $300B, then fell back after a report that the backlog was essentially all OpenAI, and is now around $1.30. He says the market doesn't trust a company that loses money and has hinged its future on that revenue.

  “Oracle goes right down the toilet because half of Oracle's backlog is just from OpenAI.”at 22:51

Space Exploration Technologies

Across the shows: 2 speakers in other episodes, 0 bullish and 2 bearish calls, past 30 days

• Josh Brown · Ritholtz Wealth ManagementNeutral

  SpaceX's record IPO traded like a normal stock despite dire predictions of collapse.

  Brown cites SpaceX as the largest IPO ever. Predictions that it would distort index funds, drain other stocks, and collapse when insiders unlocked and sold did not happen. The stock opened, fell, rose and fell back, and is now trading like a normal technology stock. Brown uses this as a precedent for how a comparably sized Anthropic IPO might be received, though he says he has no idea.

  “It opened up, it fell, then it went up again, then it fell back, but it's acting like a normal stock, a normal technology stock.”at 38:43

Lennar

Across the shows: 1 speaker in other episodes, 0 bullish and 1 bearish calls, past 30 days

• Michael Batnick · Ritholtz Wealth ManagementBearish

  Lennar and other homebuilders are falling as the 2022 incentive-driven sales frenzy ends.

  Batnick says homebuilders like Lennar, Pulte and KB were on fire in 2022 only because they offered heavy concessions to buyers in a frozen market, which was a free-for-all. He argues that dynamic is over: buyers are accepting that high mortgage rates are the new normal, and new inventory is slowly coming to market. He says housing stocks are being destroyed, and Lennar is falling every day, though it had a bounce that day.

  “And that dynamic is over... the housing stocks are getting destroyed. Lennar, I think, I mean, it's straight down every single day.”at 42:23

Frontier AI labs & model race

• Steve Eisman · The Real Eisman Playbook

  AI models have no moats; users switch freely between Claude, Gemini and ChatGPT.

  Eisman says the AI business has no moat at all. Unlike Google, where users stayed for years, people move between Claude, Gemini and ChatGPT as the best model changes. He argues this is why the labs are spending so heavily on lobbying and doom messaging, since open-weight alternatives erode whatever pricing power they have.

  “And the basic problem with the entire AI enterprise is that there are no moats. None. Right. Absolutely none... One model is good one day and then people switch to the next.”at 2:02

• Josh Brown · Ritholtz Wealth Management

  Enterprise standardization on one LLM is a sneaky moat, since staff stay on what they know.

  Brown's firm standardized on Claude Enterprise, and he moved his ChatGPT work to Claude because the firm already pays for it. He thinks the enterprise LLM is a sneaky way to build a moat: once Fortune 500 workers are used to a model, they are unlikely to switch. Eisman disagrees, arguing that they would switch to an open-weight model if it were much cheaper.

  “I sort of think the enterprise LLM is a sneaky way to build a moat because you get all the workers of the Fortune five hundred accustomed to using it, and then why would they go use another?”at 3:28

AI regulation & safety

• Steve Eisman · The Real Eisman Playbook

  Claims that LLMs will destroy Earth are a lie, used to win regulation that shuts out rivals.

  Eisman argues the AI labs' warnings that LLMs could end the world are self-serving. He thinks the leaders expect that once they are big enough, regulation will be the lever to keep Chinese open-weight models out and lock in a duopoly, since they fear regulation they can't manipulate. He calls the doom claims 'a lie'. He also says the models are not sentient and only predict the next word.

  “some of these guys have made that LLMs are going to destroy planet earth is a lie. Yeah. Plain and simple. It's just a lie.”at 4:14

• Steve Eisman · The Real Eisman Playbook

  Government stakes in AI labs would create too-big-to-fail expectations and bailouts; he opposes them.

  Asked about government stakes in AI labs, Eisman says he thinks the idea would be awful. Once the government owns a stake, investors would assume these firms are too big to fail and would be bailed out, which creates moral hazard. He does not want the government involved.

  “No, because then everybody would assume that they'd be too big to fail and they'd get bailed out. I don't want the government involved.”at 35:17

US growth & recession risk

• Steve Eisman · The Real Eisman Playbook

  An OpenAI bankruptcy would tip the US into recession, unlike 2008 because banks are fine.

  Eisman draws a contrast with 2008. A GM bankruptcy would cost jobs and suppliers and cause a regional recession. A JPMorgan failure would be systemic, which is why the government bailed out banks. An OpenAI bankruptcy would be different: the banks are fine, but the US economy would still go into recession immediately, because the AI labs made commitments to 'basically everybody who would take their pen' and the concentration of AI spending is so large. Eisman says this is hypothetical, not a prediction.

  “OpenAI went bankrupt tomorrow and I'm not making that prediction, but let's just say that it did. I think The US economy would go into a recession immediately.”at 11:08

• Steve Eisman · The Real Eisman Playbook

  An AI collapse could cause a bad recession, but not a financial crisis on the same scale.

  Eisman argues that the next downturn won't be a great financial crisis because institutions are no longer levered 40 to 1 as they were before 2008. He still says a bad recession is possible if AI really falls apart, which ties the macro risk directly to the AI trade. He does not put a probability or timeframe on it.

  “It's not like that anymore, but could we have a bad recession if AI really falls apart? You bet... But it's not on the same order of magnitude.”at 47:14

AI infrastructure & data centers

• Steve Eisman · The Real Eisman Playbook

  Local pushback will slow the data center buildout a lot, which could support existing compute.

  Eisman sees widespread local opposition to data centers across the country, and expects the buildout to slow significantly. He notes this is hard to quantify. He thinks slower buildout could leave the compute that exists in greater demand, since you can't place new compute without a data center, though he is unsure whether it would be bullish for compute sellers.

  “there's so much pushback that all this, the build out of data centers seems to be it's going to slow by a lot.”at 18:17

• Josh Brown · Ritholtz Wealth Management

  If neoclouds can't raise debt, hyperscalers gain, and they are stronger than a year ago.

  Brown argues that the demand for AI will not fall even if debt for projects is harder to sell. Consumer and enterprise usage is too embedded to reverse. The constraint is who gets to build capacity. If CoreWeave and Nebius can't sell debt and supply as much compute, the business goes back to Amazon, Alphabet and Microsoft, which have cash flows from their core businesses and are in a stronger position than a year ago. Eisman agreed that the hyperscalers' position has strengthened, but he worries about concentration in the labs.

  “if the Core Weaves and the Nebiuses of the world are not able to sell debt and supply as much compute as they want to, the ball just goes back into the court of the hyperscalers.”at 20:16

US stock market outlook

• Steve Eisman · The Real Eisman Playbook

  The market is pure narrative, and AI is the only narrative that matters for stocks.

  Eisman calls the market 'pure narrative' and says the only narrative that matters right now is AI, so the AI narrative's fate determines the stock market. He supports this with index concentration: about 40% of the S&P 500 is tech, roughly 50% once Google, Amazon and similar names are included, and about 65% once investment banks and AI-power industrials like GE Vernova, Quanta, Eaton and Rockwell are added. He concludes it is 'all one trade'.

  “There is only one narrative that matters right now, and that's AI. So you tell me what happens to the AI narrative and I'll tell you what happens to the stock market, period.”at 39:44

• Josh Brown · Ritholtz Wealth Management

  Nearly all stock-market performance now comes from derivatives of the AI trade.

  Brown argues the economy has become the stock market, with the performance of public companies coming from some derivative of the AI trade. He points out that companies not yet public, such as OpenAI and Anthropic, could cause a recession, and that the bond market's reaction to debt sales showed how much it matters. He says this may not last forever but is more true than ever right now.

  “I really think all of the performance in the market is coming from some derivative of the AI trade.”at 40:12

US housing

• Michael Batnick · Ritholtz Wealth Management

  Housing is in its worst state since the GFC, with owners locked in and no sellers.

  Batnick says the housing environment is the worst since the GFC. He explains that the market is frozen: existing owners are stuck with low mortgage rates and won't sell, and buyers can't afford higher-rate mortgages. He says the situation is bad even though it is not the same as 2008, and he finds it striking that a third of the economy is in a depression-like state while the market ignores it.

  “This is the worst housing environment since the GFC... obviously not nearly the same thing, but it is bad.”at 41:31

• Steve Eisman · The Real Eisman Playbook

  US home prices will fall, but gradually over years, as affordability stays very poor.

  Eisman says housing is locked because of rates and is less affordable than at almost any time in living memory. Affordability is worse than mortgage rates alone suggest because insurance, utility, tax and down-payment costs have exploded. He cites an estimate that total monthly ownership costs are about 50% higher than five years ago. Prices have not fallen yet, but he expects them to come down through rising foreclosures and boomer deaths releasing supply, while demand from buyers who can't afford homes persists. The adjustment is gradual and will take years.

  “I think housing prices are going to come down... They can't afford it unless the price goes down. ... It takes years.”at 43:58

AI pricing & unit economics

• Steve Eisman · The Real Eisman Playbook

  Token maxing is over; cheap open-weight models at a tenth the price are starting a price war.

  Enterprises were ordered to use AI around the clock regardless of cost, and AI costs spiraled (Uber exhausted its AI budget in three months). That spending spree is now ending. Open-weight models cost about a tenth as much and are not as good, but Eisman argues that is enough for most uses ('you don't need a Mercedes Benz for every road trip'). He sees evidence that a price war has started, and expects companies to become more cost-conscious and push staff toward cheaper models, pressuring Anthropic and OpenAI to cut prices.

  “So token maxing is done... You've got open weight models that are a tenth the price. They're not as good as the models, but you don't need a Mercedes Benz for every road trip.”at 1:15

Enterprise AI adoption & AI software

• Josh Brown · Ritholtz Wealth Management

  Investment banks will have much bigger AI budgets in 2027 than in 2026 or 2025.

  Brown says that after two days of meetings with investment banks on Wall Street, every one of them plans a substantially larger AI budget, with more AI-related projects and hiring, in 2027 than in 2026 and 2025. He thinks that enterprise demand will not change even if the financing of AI infrastructure gets harder. Consumer use of ChatGPT, Gemini and Google's AI search will also persist.

  “Every one of them is going to have a substantially larger budget for AI and AI related projects and AI related hiring in 'twenty seven than they had in 'twenty six than they had in 'twenty five.”at 19:11

Software & SaaS

• Josh Brown · Ritholtz Wealth Management

  Usage-based AI pricing feels more sustainable than per-seat SaaS, though it costs more.

  Enterprise software used to sell per employee, such as Salesforce at $80 per head per year. AI contracts have moved toward pay-as-you-go pricing. Brown thinks this is more sustainable than the old per-seat model, and Eisman answers that it is also more expensive. Brown argues companies have more control over usage-based costs.

  “Now it's migrated more toward a pay what you use model. That feels more sustainable to me than the way we thought they were going to sell these services.”at 21:36