The Real Eisman Playbook

Eisman argues NVIDIA's circular financing model is dissolving, ending the 'one trade' AI narrative

22 min Listen to the original Episode page
NVIDIAAI TradeCircular FinancingMarket CorrectionRetail EarningsAnthropic IPOJensen HuangEarnings Guidance

Summary

Steve Eisman claims NVIDIA operated as a virtual holding company through circular financing, lending to AI firms that bought its chips, a structure now dissolving due to Jensen Huang's resignation. He asserts this ends the dominant 'one trade' AI narrative and returns focus to stock picking. Eisman criticizes NVIDIA's Q3 guidance for excluding China revenue and predicts declines. The episode also covers retail earnings, an Anthropic IPO analysis, and distinctions between market corrections and bear markets.

Key takeaways

  1. 1
    The host calculates that the entire market has become an 'AI trade,' estimating that adding Infotech, tech-adjacent companies (Google, Amazon), utilities, banks, and industrial suppliers (Eaton, Quanta, GE Vernova) accounts for 65% of the S&P 500.

    Steve Eisman · 5:04

  2. 2
    the Steve Eisman drama scale. It goes from 1 to 10. And in the case of NVIDIA, it goes to 11. If you own stock in a company with low drama, like say Coca-Cola, it doesn't get more low drama than that. Then, when Coca-Cola reports, all you need to do is go through. the numbers to make sure your thesis is intact. But sometimes the narrative of a sector or stock changes and the drama scale skyrockets. At that point, quarterly earnings reports become far less important because they are backward looking.

    Steve Eisman · 9:22

Episode sections

  1. 0:00 – 3:29

    Intro and Market Context

    Steve Eisman discusses the market correction driven by midterm ramifications and Jensen Huang's resignation. He outlines the episode's agenda, including the 'Jensen Unwind,' earnings analysis, and the Anthropic IPO.

  2. 3:29 – 7:09

    The Jensen Unwind and AI Trade Structure

    Eisman argues that NVIDIA operated as a virtual holding company through circular financing, which is now dissolving due to Huang's resignation. He claims this ends the 'one trade' AI narrative and brings stock picking back.

  3. 7:09 – 11:18

    Quarterly Earnings Analysis Framework

    The speaker provides a framework for analyzing earnings, distinguishing between sectors like banks and tech. He emphasizes that earnings are backward-looking and that narrative changes (drama) can outweigh financial results.

  4. 11:18 – 13:36

    NVIDIA Earnings and Guidance

    Eisman reviews NVIDIA's Q3 results, noting the exclusion of China revenue from guidance. He criticizes the guidance as premature and predicts revenue declines due to the end of circular financing and regulatory risks.

  5. 13:36 – 16:05

    Retail and Consumer Earnings

    The host analyzes reports from Home Depot, Lowe's, Target, and Walmart. He highlights the struggling housing market impacting Home Depot and Lowe's, while noting Target's turnaround success and Walmart's defensive positioning.

  6. 16:05 – 18:06

    Anthropic IPO and Risks

    Eisman finds the Anthropic IPO anticlimactic, citing slowing revenue growth, lack of moats, and massive capital commitments. He warns of potential junk ratings and intense competition from cheaper open-weight models.

  7. 18:06 – 19:15

    Mailbag: Corrections vs. Bear Markets

    In response to a listener question, Eisman distinguishes between corrections and bear markets. He advises long-term investors to dollar-cost average into the S&P 500, while traders might shift to defensive sectors and short-term treasuries.

  8. 19:15 – 22:10

    Mailbag: Liar Loans and Upcoming Interviews

    Eisman explains the mechanics of pre-crisis 'liar loans' in the subprime mortgage market. He concludes by promoting upcoming interviews with tech analysts and an industrials analyst.

Assets discussed

AssetKindStanceContext
NVIDIANVDAequitybearishThe new risks created by the midterms and the changes at NVIDIA dwarf everything else.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

Jensen Huang's resignation effectively dissolved the virtual holding company NVIDIA built through circular financing.

Steve Eisman ·3:04

How to verify: Review NVIDIA's Q3 2026 earnings report and press releases for any announcements regarding changes in financing practices or leadership structure post-resignation. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA provided circular financing to Anthropic ($20B), OpenAI ($30B), SpaceX ($20B), Core Weave ($2.5B), and Nebius ($2B).

Steve Eisman ·5:44

How to verify: Check SEC filings (Form D or 8-K) for these companies and NVIDIA's investor presentations for disclosed equity stakes or loan agreements. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

Infotech represents 40% of the S&P 500, and adding tech-adjacent companies and industrial suppliers brings the AI-related exposure to 65% of the S&P 500.

Steve Eisman ·5:04

How to verify: Calculate the weight of the Information Technology sector in the S&P 500 using S&P Dow Jones Indices data, and add weights of Google, Amazon, Eaton, Quanta, and GE Vernova. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

NVIDIA's Q3 2026 EPS was $2.68 (up 111% YoY) and Revenue was $116 billion (up 103% YoY).

Steve Eisman ·12:02

How to verify: Verify against NVIDIA's official Q3 2026 earnings press release and SEC Form 10-Q. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

NVIDIA's Q4 2026 revenue guidance is $130 billion, excluding compute revenue from China.

Steve Eisman ·12:17

How to verify: Check NVIDIA's Q3 2026 earnings call transcript and press release for the specific guidance figure and exclusions. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

Home Depot reported EPS of $3.80 (up 3% YoY) and same-store sales comp of 2.2%.

Steve Eisman ·13:56

How to verify: Verify against Home Depot's Q3 2026 earnings report filed with the SEC. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

Anthropic's Q3 2026 sequential revenue growth slowed to 45% from 145% in Q2.

Steve Eisman ·16:54

How to verify: Check Anthropic's S-1 filing or subsequent quarterly reports for revenue growth metrics. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

Anthropic has made $900 billion in various kinds of commitments.

Steve Eisman ·17:15

How to verify: Review Anthropic's S-1 filing or investor presentations for disclosed capital commitments and future funding obligations. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Primary source

Anthropic cut prices by 50% in October 2026 to stave off market share losses to cheaper open-weight models.

Steve Eisman ·17:15

How to verify: Search for press releases or announcements from Anthropic regarding pricing changes in October 2026. For recent events, verify 2026-08-20 through 2026-11-20. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify NVIDIA's Q3 10-Q filing for the exact revenue guidance and the specific exclusion of China compute revenue.
  • Check SEC filings for any loans or financial arrangements between NVIDIA and major AI customers like Anthropic or OpenAI to assess circular financing claims.
  • Review Treasury Secretary Bessent's public statements or official appointments regarding tech sector oversight to verify the 'watchdog' claim.
  • Analyze S&P 500 sector weightings from a financial data provider (e.g., Bloomberg, S&P Global) to confirm the 65% AI-related exposure calculation.
  • Examine Anthropic's IPO prospectus for revenue growth rates, capital commitments, and competitive moats against open-weight models.
  • Compare Home Depot and Lowe's recent earnings reports against housing market data (e.g., existing home sales, building permits) to validate the housing impact thesis.
  • Review historical S&P 500 performance during past corrections vs. bear markets to evaluate the dollar-cost averaging advice.
  • Check NVIDIA's stock price movement and trading volume around the time of Jensen Huang's resignation announcement to assess market reaction.
Full transcriptexpand

0:00Steve Eisman: Let me just discuss the week because it looks like we are in the throes of a correction. The combination of the ramifications of the midterms with the resignation of Jensen Wong caused the market to decline. So, how do you position for a correction? The Jensen Unwind is just starting and it's going to get very ugly. The leaders of the tech world know it and they want Wong back. Not so fast. Home Depot reported they are facing the worst existing home sales. market literally in decades. It looks like Target's attempts to turn the company around are finally succeeding. Anthropic went public this week. A few weeks ago, the Anthropic IPO was important, but no longer. The new risks created by the midterms and the changes at NVIDIA dwarf everything else. And we will be discussing all this at length in today's wrap. rap.

0:59Unidentified speaker (UNKNOWN): Hi,

0:59Steve Eisman: this is Steve Eisman, and this is the weekly rap. This is for the week ending Friday, November 20, but recorded Thursday night, November 19. If you like what we're doing, please subscribe to our YouTube channel. You can also visit steveizman.com. steveisman.com to subscribe on Substack. We are having a Black Friday Cyber Monday special. For Thanksgiving weekend, annual subscriptions are 15% off, which brings the price to 170 a year. The sale runs from 12:01 a.m. Eastern Standard Time, Friday, November 27, through 11:59 p.m. Eastern Standard Time, Monday, November 30. The offer covers annual plans only. If you stay subscribed, you'll keep the $170 price for life. On Wednesday, November 18, we posted an interview on Substack Premium with Gary Gensler. He used to chair both the Securities and Exchange Commission and the Commodity Futures Trading Commission.

2:01Unidentified speaker (speaker_1): We're spending $700. $750 to $800 billion of CapEx in 206, going to a trillion next year in 2030 to justify all those numbers. Right. What does our economy have to spend in aggregate per year $2.4 to $3 trillion a year? It's massive numbers.

2:24Steve Eisman: We talked about the risks in the AI trade. which is top of mind for everybody, how prediction markets should be regulated, and other regulatory topics. Next Wednesday, November 25th, we'll post an interview on premium with Harvard history professor Sven Beckert about his book Capitalism: a Global History. On this week's wrap, we will discuss: one, are we in the throes of a correction? Two, the gentleman. Jensen unwind. AI was one trade because NVIDIA operated as a virtual holding company by financing the whole AI stack. 3. Stock picking maybe back because Jensen Wong's resignation effectively dissolved the virtual holding company he created. 4. Four, How to think about quarterly earnings reports in general when picking stock? stocks. Five, as I'm sure you know, NVIDIA reported. Six, maybe some of you know less about this, but other companies did report too. And yes, Anthropic went public, that's seven. And eight, two mailbags. Let's get started. Before we dig deep, let me just discuss the week because it looks like we are in the throes of a correction. The combination of the ramifications of the midterms with the reservation. resignation of Jensen Wong caused the market to decline on Monday and Tuesday.

3:44There was a small rally on Wednesday, and then Anthropic went public on Thursday. Some investors were hoping that the Anthropic IPO would drive the market higher, and that lasted only a few hours, and the market declined again. From last Friday through Thursday, NASDAQ. was down about 4%. My view is that a few weeks ago the anthropic IPO was important, but no longer. The new risks created by the midterms and the changes at NVIDIA dwarf everything else. And we will be discussing all this at length in today's wrap. First, the Jensen Unwind is just starting, and it's going to get very ugly. The leaders of the tech world know it, and they want Wong back. Not so fast. Treasury Secretary Besson has appointed himself watchdog. Looks to me that if Jensen re-enters the scene, it won't be through NVIDIA's front door. In the meantime, banks, private credit, private

4:44equity, companies, and shareholders have to wrestle with what is about to happen to their balance sheets. NVIDIA reports. reported Tuesday night, and NVIDIA largely, though not completely, represents the end of earnings season. Later, I'm going to discuss NVIDIA's results. But first, I will explain how and why the role NVIDIA played under Wong's leadership is forever changed and how and why stock picking is back. Infotech represents 40% of the SP. If you then add companies like Google and Amazon that are techie, but technically not in the infotech category, you quickly get to 50%. Then add the utility sector, all the large banks, alternative asset managers, and asset managers involved in funding AI data center construction, industrial companies involved in AI data center construction like Eaton, and industrial companies involved in providing the power to AI data centers. like Quanta and GE Vernova, and you easily get to 65% of the SP. In a sense, the entire market has become one AI trade. But it's even more convoluted than that. Jensen Wong did not want to sell all his chips to the hyperscalers. So, in a process called circular financing, he decided to take equity stakes or give loans to a multitude of companies, including Anthropic, 20%. billion,

6:03OpenAI, 30 billion, SpaceX, 20 billion, Core Weave, 2.5 billion, Nebius, 2 billion, and several others. Essentially, he gave them money, and they used that money to buy his chips. He created, I think, a virtual vertical holding company. And while the party lasted, everybody was happy. And now the party's over. His resignation is the effective dissolution of the virtual holding company. All the pieces he financed are now, more likely than not, going to be on their own. Although it's still early to know for sure. But they need to be evaluated on their own merits and not as part of a larger whole. Will they get financing? Are they well run? Who is the leader? What are the projections? What's the cash flow like? The balance sheet? The earnings and the potential. Without NVIDIA backstopping the AI stack, AI is no longer one trade. And that's why stock picking is

7:03back and valuation is back. Because valuation is back, I think this is a good time to take a step back and discuss how to think about quarterly earnings. Earnings seem season begins around 15 days after the end of the quarter. The banks report first, and NVIDIA closes the session. It's all about six weeks long, and there are some days where so many companies report that it can be overwhelming. So, how should you think about earnings season and what should you look at? The first thing to realize is that companies report their results differently. Some just put out a fairly simple press release. Others, like the banks, put out an earnings press release, a very large financial supplement, and a conference called DEC. Perhaps even more importantly is to realize that what investors care about differs between sectors and can also change over time. With respect to the banks, for example, investors care about loan growth, changes in the net interest margin, operating efficiency, and changes in credit quality. In tech, investors used to care only about revenue and earnings growth and backlog. However, now that AI dominates, investors also care just as much about CapEx cash flow and whether a company is going to fund its CapEx needs out of cash flow or the

8:22capital markets. When it comes to software companies, those companies all employ a SaaS model, Software as a Service. In analyzing that model, investors cared about growth in the number of seats and pricing trends. Today, however, investors are looking for any signs that AI is causing software growth rates to slow down. Think of quarterly earnings The as a data point check on your thesis. Suppose your thesis for a company is that revenue will grow at least 15% every quarter for the next several years. If the company reports growth of 16%, then all is good. But if revenue growth slows to 14%, you've got some soul searching to do. Every company makes mistakes, every company hits stumbling blocks. You're going to have to listen to the conference call and do some research to ascertain whether this quarter is just a hiccup or has the story changed. for the worst. That is not an easy decision to make. Also, there is the impact of drama. I call this the Steve Eisman drama scale. It goes from 1 to 10. And in the case of NVIDIA, it goes to 11. If you own stock in a company with low drama, like say Coca-Cola, it doesn't get more low drama than that. Then, when Coca-Cola reports, all you need to do is go through. the numbers to make sure your thesis is intact. But sometimes

9:42the narrative of a sector or stock changes and the drama scale skyrockets. At that point, quarterly earnings reports become far less important because they are backward looking. Right now, the AI narrative is changing almost daily. It was a hyper-growth tale, but because of politics and the invidious scandal, AI data center construction and the entire AI story could, I emphasize, could slow. It could slow a bit, or it could slow a lot. It's too early to tell. So when a company like Cisco reported great numbers last week, all that showed you was that AI CapEx was explosive, but that report tells you nothing about future quarters. One more thing to address is why do investors react so. poorly when a company misses key metrics, even if they miss them only barely? Investors aren't stupid. They know that within limits, companies have the ability to massage their numbers. Accounting rules aren't completely rigid. Companies give guidance for either the next quarter or the next fiscal year. And when they give that guidance, they have sufficient visibility in their balance. backlog, et cetera, to render that guidance. And they also know that they are going to beat that guidance. Underpromise and over-deliver is the mantra for

11:02literally every corporation. So when a company doesn't even meet the guidance, but actually misses it, something is really wrong. The only question is whether it is a one-time blip that will last. a quarter or two, or something really fundamental and long term. Now let's turn back to NVIDIA and its results. Before we get to those results, I want to point out that they are completely backward looking. The forward look is very complicated. Apparently, 20% of NVIDIA's revenue comes from shipments to dicey Asian entities that were probably illegal. Those shipments now could go to zero. Also, assuming that the new management team of NVIDIA abandons or curtails all circular financing guarantees, etc., we need to then know how much of the company's revenue comes from that as well. And right now we don't really know. In the end, NVIDIA's revenue is going to be lower. How much lower, we don't yet know. Maybe it can grow from there, but it has to go

12:02down first. I think. Now the earnings results were, of course, fine. Earnings per share was $2.68, up 111% versus last year. Revenue of $116 billion was up 103%. Nvidia gave guidance for the next quarter of revenue of an amazing $130 billion. And management said that that number does not include, quote, computer. compute revenue from China, unquote. Personally, I think that giving this guidance is a mistake. So much has to be unwound here that the better part of Valor would have been to provide no guidance until management has its arms around all its problems. But I'm sure management thought that if they did not provide guidance, then and the stock would collapse and they wanted to avoid that. The stock was up around 3% after hours. Given how much NVIDIA is down recently, that was not all that impressive and it did not hold. Despite opening up on Wednesday, the stock closed down slightly. By the way, the key sentence, I'm going back to it, in the press release, was that NVIDIA gave revenue guidance that does not include compute revenue from China. And that sounds great. But what does it really mean? Does it exclude direct sales to China? Or does

13:17it also exclude sales to entities in Asia who are then funneling those chips illegally to China? It's just not clear. Anyway, regardless, it's going to be several quarters before we know how much NVIDIA's business model will be negatively impacted. Until then, commentators can say pretty much anything they want. Let's move on past NVIDIA. Lest you need more data. to drive home the fact that the housing market is completely locked, Home Depot reported. Now I have sympathy for this company. It is very well run and they are doing everything they can to grow, but they are facing the worst existing home sales market literally in decades. The lesson here is that a bad macro So trumps a good company almost every time. Now, Home Depot report earnings per share of 380. 386, up 3% versus last year. In this awful housing market, that's not bad. The same store sales comp was 2.2%, which was 50 basis points better than expected. But the company guided to a same-store sales comp for 4Q Q of only 1%, which is below consensus. And that's why the stock was down on Tuesday. Amazingly, the stock is roughly flat this year. I suppose Hope Springs eternal, but there is no sign, no sign, that the housing market will improve anytime soon.

14:36I really, really wish it were otherwise. Lowe's also reported, and the numbers were really not any better than Home Depots. Lowe's reported EPS of 310 10 up 1% versus last year, so not exactly impressive. But for the quarter, Lowe's lows reported. the same store sales comp of 1.8%, which is not exactly great, but was 160 basis points better than consensus. However, those comp numbers are still poor. By contrast, it looks like Target's attempts to turn the company around are finally succeeding. Earnings per share was 265, up in an amazing 49% versus last year. and same store sales grew an impressive 5.6%, which was much better than expectations. It's nice to see a company turn itself around. As we can see from Nike, that can often be difficult. Walmart reported Thursday morning: the results were fine, but not overwhelming. Earnings per share was 65 cents, up only 5% versus last year, and a beat. by a whopping penny. Big deal. Net sales grew a decent 4.4%. More importantly, Walmart is a great company, but it often trades like a high-class safety stock. Since September 30th, the AI trade has, as we all know, reversed, and investors have

15:55sought out safety stocks like Walmart. And the stock has climbed from 105 to 125. Reallocations can be very powerful. And now we can talk about anthropic. Anthropic went public this week, finally, and yet it feels completely anticlimactic. A few weeks ago, the anthropic IPO seemed all important. It was going to set the tone for the next stage of the AI narrative. And then, as I've already discussed, two things happened. The Democrats swept both houses by large margins and Jensen Wong resigned. The ramifications of those two events dwarf the Anthropic IPO. The midterms have created a risk of an AI moratorium or an AI data center construction slowdown. The Jensen Wong resignation raises the specter of an unwind of all of Wang's circular financing deals, and this is what investors will be focusing on literally for months. As for the anthropic IPO, itself, I find the stock oddly uncompelling. The perspective shows that third quarter revenue growth has slowed. In 2Q, sequential revenue growth was 145%. The third quarter was only 45%. I attribute this slowdown to the end of token maxing. For me, the two major long long-term term risks for anthropic are one, LLM creators like anthropic lack.

17:15motes, business motes. It's just too easy to switch between models, and that is why in October, Anthropic cut prices by 50%. It is seeking to stave off market share losses to the cheaper open weight models. And second, Anthropic has made $900 billion in various kinds of commitments. This means it will be raising debt at what I am sure will be expensive prices. It will be interesting to see. what credit ratings the ratings agencies provide to Anthropic. A junk rating is not impossible. And three, Anthropic and OpenAI have to continue to evolve their product. They build frontier models, and staying on the frontier is expensive. Anthropic is a business with slowing revenue growth, increasing expenses, and increasing costs of capital. I just think there are too many risks. here. And that's why I'm uninvolved. And on to the mail there. There are two. First question: how do you position investments in a bear market versus a correction? A correction to me would be a decline in the stock market for a year or two. A true bear market lasts years. We have not had a bear market since the 1970s. I don't foresee a long-term long term bear market because the U.S. economy is more. dynamic than perhaps any time in its history. And it is more dynamic, I think, than any other country in the world. Also, the banks

18:35are well capitalized. So even if the banks experience losses, they have the capital to withstand them. So how do you position for a correction? For those of you who are long term investors, I would not do much of anything. I would continue to gradually buy the S P 500 in. index, dollar cost averaging, and just not watch the market. The US will recover. But for those of you who are more trading oriented and can stomach paying capital gains taxes, I would sell some stocks and buy some combination of a healthcare index, a staples index, and some short-term treasuries. The second question is from Jacob, who asks: Steve, I'm curious, how did you identify the liar loans? if the information in them was falsified? Great question. First, some definitions. Pre-crisis, the subprime mortgage industry created a new type of loan. They called it a no-doc or low-doc loan. That means that very little underwriting was done. The buyer was simply asked what his income was, and the underwriter

19:35did not check it. They took the borrower at his or her word. Moody's securitization database provided information about these loans in each securitization. And examining the database, it became clear that starting in 2006, about 50% of all securitized subprime loans were low doc loans. We guessed about the amount that were liar loans. And to this day, no one really knows the percentage of liar loans in the low doc loans. Jacob refers to liar loans, which are a large subset of the low dock loans. We just don't know how big. What's a liar loan? Some commentators have argued that the consumer simply lied about his or her income to get a mortgage loan. They blame the consumer. I don't think that that is what really happened. Instead, the underwriter would say to the borrower that to get this loan, they needed income of, let's say, $50,000 a year. And the underwriter would then ask the borrower the level of his income. And at first the borrower would be honest and say $40,000. The underwriter would then cajole the borrower to say $50,000, and finally, most did. That is what liar loans were all about. We guessed about the amount that were liar loans, but to this day, again, no one really knows the percentage of liar loans in the low-doc low doc loans. This week on Monday, November 16th, we released an incredibly timely interview with tech

20:54analysts Dan Ives and Gil Luria.

20:57Unidentified speaker (speaker_2): The AI revolution's accelerating. We see that in Korea a few weeks ago in terms of chips, but then the spending, I think investors have said we have to slow this down, not realizing that it was going to take a blue tsunami to actually slow it down. The

21:10Steve Eisman: interview was conducted just after the midterms, and we delved into the midterm implications. as well as all aspects of the AI trade. They gave a deep dive into the bull case I just outlined, and this coming week on Monday, November 23, we will release an interview with Steven Tusa, the industrials analyst at Wells Fargo. Entire industrial landscape. Hope you tune in. The best way to support the Real Eisman playbook is to subscribe to Substack through SteveEisman.com and to subscribe to YouTube. And we greatly appreciate your support. And that's the wrap. This podcast is for informational purposes only and does not constitute investment advice. The hosts and guests may hold positions and stocks discussed, opinions expressed on their own and not recommendations. Please do your own. due diligence and consult a licensed financial advisor before making any investment decisions.

This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.
All insights