Transcript of Market Cracks Under AI Debt Burden | The Weekly Wrap
Steve Eisman
0:00Oil and energy prices continued to0:01influence the market on a daily basis.0:04The inflation data was worse than0:06expected. There was negative news about0:08core wheat and that news caused tech0:10stocks and the market to decline. The0:12news from France [music] is particularly0:14bad these days. France's debt to GDP is0:17at 120%. [music] A dangerous level. The0:19banks kicked off earnings season. Let me0:21just explain the concern some investors0:24have with respect for the impact of0:25Agentic AI on banks and other [music]0:28financials. We've had 17 years of0:31amazingly benign credit quality and0:33investors are wondering if the good0:34times are over. And this brings me to my0:36conclusion about the state of the US0:37economy. Let's get started.0:40[music]0:46[music]0:48Hi, this is Steve Eisman and this is0:50another edition of the weekly rap. This0:53is for the week ending Friday, October0:5516, but recorded Thursday night, October0:5815. I just want to repeat an1:00announcement about a change we made1:02recently. Monday and Friday episodes are1:04available to premium Substack1:07subscribers earlier. The Monday1:10interviews are released 12:00 p.m.1:12Eastern Standard Time every Sunday, and1:16the Friday rap is released at 12:00 p.m.1:19Eastern Standard Time every Friday. The1:22Monday and Friday episodes will still be1:24available on YouTube and all audio1:27channels at their usual times, Monday at1:30noon and Friday at 4:15 p.m. For1:34Substack Premium, this Wednesday, we had1:37returning guests Gretchen Morgansson and1:39Tom Gober to discuss life insurance1:41susceptibility to being defrauded. And1:44next Wednesday, we have another1:46returning guest, Dwight Collins, head of1:49sales at Autonomous Research, for his1:51thoughts and insights on the impact of1:53AI on financial companies. On this1:56week's rap, we will discuss, as usual,1:59number one, the war in the Middle East.2:01Number two, the telep2:07correction. Number four, bad news in2:10France. Five, the impact of agentic AI2:13on financial institutions. Six, and2:16last, earning season began with, of2:19course, the banks. Let's get started.2:22The war is still in a stalemate. Oil and2:24energy prices continue to influence the2:26market on a daily basis. On Wednesday,2:29the inflation data was worse than2:31expected. Interestingly though, interest2:33rates declined on Wednesday. Why? There2:37was negative news about coreweave, which2:39I'll discuss in a bit. And that news2:41caused tech stocks and the market to2:44decline and some investors purchased2:46treasuries as a safety play despite the2:49inflation news. Now you've heard of the2:51SAS apocalypse. We've spoken about it2:53many times and now there is a new2:55apocalypse and I think I have coined2:57this term and I call it the tel2:59apocalypse. This is not an AI story but3:02an Elon Musk tale. On Friday, last week,3:06October 9, AT&T, Verizon, and T-Mobile3:09were down 9 to 13%. Why? SpaceX's most3:13profitable division is called Starlink.3:16It is a company that uses satellites to3:18provide cellular service all over the3:21globe. Sounds great, but the service is3:23not nearly as good as the service3:25provided by telecom providers. Now,3:28Starlink announced that it was acquiring3:30Grain Management's 14 megahertz of 8003:33megahertz spectrum. Now, I don't want to3:35get too technical, but this will3:37definitely improve Starlink service in3:39the US. However, as best as I can tell,3:42this is still not nearly good enough.3:44Starlink needs to build a terrestrial3:47network, and that can be unbelievably3:50expensive. The most economic way for3:53Starlink to do this is to reach an3:55agreement with one of the big three in3:56the US and use their rails to provide3:59service. And none of the big three,4:01believe me, have any interest in helping4:03Starlink. So, the telep4:09narratives are powerful and this one is4:11not going away. By the way, between the4:14big three, T-Mobile is down the most4:16this year, down 25%, whereas Verizon is4:20up five and AT&T is down 9% this year. I4:25think the reason for T-Mobile's poor4:27performance is that it used to have a4:29high multiple. It's been a much better4:31company than Verizon and AT&T, and so4:34used to have a multiple in the 20s,4:36which was double double the multiple of4:38its two competitors. But now that the4:40entire landscape has changed, T-Mobile's4:432026 PE multiple is only 14 times.4:46Moving on, and this is probably the key4:48news of the week. There was a very4:50interesting piece of news about AI that4:52is indicating stress in the AI4:54narrative. Coreweave is a public company4:58that operates data centers devoted5:01solely to AI. It's called a neocloud5:04company. It is competing with the5:06hyperscalers all of whom have much much5:09larger balance sheets to build these5:11data centers. Coreweave has had to take5:13on a lot of debt. Its market cap is only5:1643 billion but it has taken on 355:18billion plus in debt and it has5:21additional capital lease obligations of5:2310 billion. This is not a company with a5:25great balance sheet to say the least.5:28When it raised its debt, the coupon it5:31had to pay was 9%. Which is very high.5:35However, it looks like the fixed income5:37markets are getting really nervous about5:39core. Recently, the debt has declined in5:42price below 90 cents on the dollar. So5:45that the yield is around 15% and that5:48means the fixed income markets are5:50quickly losing confidence in5:51cororeweave. On this news on Wednesday,5:54Cororeweave was down 15%. Nebius,5:57another Neocloud company, was down 10%5:59and the entire AI complex went down on6:02Wednesday as suddenly investors are6:03becoming afraid of AI debt levels. Last6:06week, the Financial Times reported that6:08OpenAI claims to have 50 billion in6:11annualized revenue as of the end of6:14September. Now, I have to say, I have no6:17idea what that means. Is that 1 day* 3656:21or 1 week x 52 or 1 month * 12? Who6:25knows? To me, this annualized revenue6:28data point that both Anthropic and6:29OpenAI employ is merely designed to help6:32them raise money. Regardless, the most6:34important part of the story was that6:36previously OpenAI had claimed that it6:38had 70 billion in annualized revenue,6:42and I have no idea how 20 billion6:44evaporated, but that is not a good sign.6:47Moving on, the news from France is6:49particularly bad these days, and I'm not6:51just talking about the riots. Investors6:53are losing confidence in France's6:55ability to manage its deficit. The6:57spread between the French 10-year and7:00German 10-year boons is at 140 basis7:03points, a multi-year high. The problem7:07is that France's debt to GDP is at 120%.7:10A dangerous level. There is no consensus7:13or desire by French politicians to cut7:16expenditures. In fact, Marie Le Pen, who7:18may become the next president, is7:20advocating cutting the retirement age7:22from 62 to 60. Of course, someone could7:26immediately point out that US debt to7:28GDP is also 120%. So, isn't that a7:31problem, too? Well, of course, it's a7:33problem, but it's a lesser problem. Why?7:36The dollar is the reserve currency of7:38the world. More importantly, the entire7:40global financial system functions on7:43treasuries. The repo market, which is a7:45multi-trillion dollar market for banks,7:48only uses treasuries. When large7:50institutions want to park money, they7:52buy treasuries until there is a large,7:54safe, and liquid alternative to7:57treasuries. The US has the luxury of8:00time, but not endless time. In other8:02words, no one needs to own French8:05sovereign bonds, but many large8:07institutions need to own US treasuries.8:10And now for the banks. The banks kicked8:12off earnings season. But before we get8:14to the banks, let me just explain the8:16concern some investors have with respect8:19to the impact of Agentic AI on banks and8:22other financials. When you put your8:24money in a bank, generally you put it in8:26a checking account that pays nothing or8:29a savings account that pays little. In8:32actuality, for most people, not all8:34people, but for most people, you only8:36need a limited amount of cash in the8:39bank. the extra you could put it in a8:41money market fund that pays more. Now8:43hold that thought for a second. Now my8:45wife and I have our investment accounts8:47at Schwab and we are very happy with8:49Schwab. I have no complaints. Now I have8:51some amount of cash in my account. Now8:53if I do nothing, Schwab sweeps my cash,8:58pays me very little and earns the9:00short-term interest rate for itself.9:03However, if I execute a trade and buy a9:06money market fund through Schwab with my9:08cash, I earn the interest. Now, Schwab9:11does not make this hard, but I have to9:13execute the trade. The important point9:15here is that because of consumer9:16laziness, banks and brokers like Schwab9:19have access to cheap sources of funding.9:21For a typical bank, for example,9:24checking and savings accounts are the9:26most important parts of the bank's9:28funding and therefore the most important9:30sources of its net interest margin. Now,9:32let's think about Agentic AI. What9:35investors are worried about is the idea9:37that consumers will use Agentic AI to9:39sweep their excess cash out of banks and9:42brokers and into nice paying money9:44market funds. If this were to happen, it9:47would seriously impair the net interest9:49margins of banks. Of course, banks9:50aren't stupid. They are going to try and9:53protect their franchises. This week, JP9:55Morgan announced that it would not allow9:58outside agents access to its customer10:01accounts. the fight begins. Now, that is10:04the long-term risk for banks, but that10:06risk extends to other parts of10:08financials such as personal, auto, and10:11homeowners insurance. Companies like10:13Progressive and Geico wield powerful10:15brand names. They are very price10:17competitive, and consumers flock to them10:20because of their brands. Agentic AI does10:22not care about brands. It cares about10:25credit ratings and price. Imagine asking10:28an Agentic AI to find for you auto10:31insurance on your new car. Leaving aside10:34credit ratings, Agentic AI only cares10:36about price. And this could seriously10:38impair the franchise value of personal10:41lines insurers. Though, Progressive10:42reported some numbers on Wednesday.10:44Progressive does something no other10:46company does, at least not that I know10:48of. Instead of reporting quarterly, it10:50reports monthly. and the September10:52report was disappointing in that new10:55premium volume declined for the first10:57time since November 2023. Now, the most11:00likely culprit is that pricing is soft.11:02So, Progressive is pulling back. I doubt11:05Aentic AI has gotten big enough yet to11:08make a material difference, but that11:10will not stop investors from11:11speculating. And Progressive was down 4%11:14on Wednesday. Now, I'm going to share11:16some thoughts on a bunch of the large11:17banks that reported this week, and then11:19show you how to think about valuation11:20given the results. For the large banks,11:23the major variables in terms of revenue11:25and earnings are fixed income and equity11:27trading, investment banking, which is11:29composed of M&A fees and equity and11:31fixed income underwriting, net interest11:34income, which is a function of loan11:35growth and the net interest margin, and11:37of course, credit quality statistics.11:40All of these variables are important in11:42understanding earnings results. However,11:45credit quality statistics are the most11:47important variable in terms of drawing11:49implications about the overall economy.11:52To summarize, this quarter's results11:54briefly, credit quality remains benign.11:57As for revenue, the overall results were11:59good, but not as good as prior12:01recorders. The problem was that for most12:04companies, fixed income trading was12:06weaker because rising rates put a crimp12:09on client volume. Before we discuss12:11valuation, I want to focus on how the12:13large bank results are of particular12:16importance because they provide great12:18insight into the state of the current12:19credit cycle. The four large banks, JP12:22Morgan, City, Wells, and Bank of America12:24have large and broadbased lending12:26businesses. I'd note that credit quality12:29is a less important metric for Goldman12:31Sachs and Morgan Stanley. On the12:33consumer side, the big four provide12:35credit cards, auto loans, and several12:36other types of consumer loans. On the12:38commercial side, they make commercial12:40real estate and corporate loans and12:41other types of commercial loans as well.12:43They don't do everything, but they're12:45big enough and broad enough to provide a12:47window, though not a complete window,12:49into the health of the US economy. Now,12:51there is growing fear that credit losses12:53will start to mount for the first time12:55since the great financial crisis. We've12:56had 17 years of amazingly benign credit13:00quality, and investors are wondering if13:02the good times are over. Yes, private13:03credit is exposed to software and13:06software is getting hit by AI, but are13:08there signs of broad credit problems13:10emerging in the actual data? When the13:13major banks report, they provide reams13:16of data. JP Morgan's quarterly earnings13:19release supplement is actually 29 pages13:22long and is filled with data on every13:24page. That is typical of how all the13:26banks report. to see trends in credit13:29quality. The best place to look is to13:31examine consumer and commercial13:33nonacrrewing loans. Now, a non-acrrewing13:36loan is a loan that is seriously13:37delinquent, usually 90 plus days. A bank13:40no longer reports interest income from13:42these loans and reserves for future13:45losses. I am putting on the screen13:48consumer and commercial non-acrrewing13:50loan data for JP Morgan, Wells Fargo,13:53City and Bank of America for the third13:55quarter of last year, the second quarter13:58of this year and this third quarter for14:00this year. These four banks combined14:03provide a large window into credit14:05quality trends. If if there was a14:07serious trend of deteriorating credit,14:09we would see increases in non-acrrewing14:11consumer or commercial loans or both on14:14a year-over-year basis and on a14:17quarterly sequential basis. We certainly14:19saw large increases in consumer14:21non-acrruals leading up to the GFC. We14:24are not seeing that now at all. The14:28credit data is benign in both consumer14:30and commercial. Now, for those of you14:31who are audio listeners, I'll just focus14:33on JP Morgan and Bank of America. For JP14:36Morgan, total non-accrewing loans were14:389.414:40billion in the third quarter of this14:42year. They were down 7% versus last year14:45and flat versus the second quarter. The14:48numbers of Bank of America were also14:49very benign. Total non-accrals were 5.714:53billion, up 3% versus last year and down14:574% versus the second quarter.14:59Conclusion, the credit trends reported15:01by the large banks are very benign. Now15:04maybe a credit cycle will emerge but we15:06are not seeing it in the banks. That's15:08an important data point. The banks have15:10the broadest credit exposure. So their15:12results are a concurrent indicator. And15:14this brings me to my conclusion about15:16the state of the US economy. Since the15:18GSC, many commentators have tried to15:20predict the next crisis. This is it,15:22they say, and then it isn't. Perhaps the15:25most important story of the past 1715:27years is how powerful and resilient the15:29US economy has become. Yes, there are15:32definite problems in private credit with15:34respect to overexposure to software. And15:37perhaps these software issues are15:38harbingers of an approaching credit15:40cycle. Yet the banks have shown once15:43again that credit quality in the US is15:45still okay. Whatever bad happens in the15:48future, I don't think it will be in the15:50banking system. It could emerge in15:53private credit or it could emerge in AI15:56related debt like cororeweave. That's15:58why I'm now much more focused on the16:00fixed income markets continued reaction16:02to coreweave and whether there is16:05contagion to other parts of AI. Again,16:09in my view, the banks are not leading16:11indicators this time. In the past, I16:14have said as the banks go, so goes the16:16economy. Not this time. The banks are16:19extremely well capitalized and their16:20credit quality is still benign. The16:23potential issues in AI and in private16:25credit will determine whether and when16:27there will be a recession. It would be16:30much easier to predict the future if the16:32banks were the leading indicators16:34because the data that banks produce is16:36voluminous and transparent while private16:38credit is much more opaque making future16:41economic weakness much trickier to16:43foresee. Frankly, I'm starting to think16:46that the entire future of the US economy16:48hinges on the success or failure of AI.16:51That's why I talk about it so much. As16:54for the financial results of the banks16:55this quarter, they once again posted16:57strong results largely because of16:59extremely strong investment banking and17:01trading. The banks are partial17:03beneficiaries of the AI boom as17:05companies with AI financing needs are17:07going to Wall Street for help. These17:09results produce record levels of returns17:12or almost record levels of returns.17:14Here's a brief synopsis. JP Morgan17:17earnings per share was 557 versus 50717:2010% growth. Both revenue and earnings17:23per share get this missed which is a17:25very very rare event for this company.17:28There was a lack of operating leverage17:30as revenue was up 10% but expenses were17:32up 12%. Also, the buyback was a poultry17:35700 million and fixed income trading was17:38down 4% versus last year. The return on17:41tangible common equity was a strong 20%.17:44Bank of America reported a poor quarter.17:46It missed on revenue because fixed17:48income trading declined by 17% versus17:51last year. It also missed on earnings17:54because of a government fine. The17:56company paid a $1.2 billion fine because17:59of failures with respect to anti-money18:02laundering. a big no no. AML is18:05something regulators are very strict18:06about and it's kind of shocking that a18:08bank of equality of Bank of America is18:11still having problems in this area. The18:13market was quite shocked by this fine as18:16the stock was down over 4% on Wednesday.18:18Still, Bank of America's return on18:20tangible common equity is a very18:22respectable 15%. So long-term, the18:25company's fine. Wells Fargo has a18:27history of occasionally disappointing in18:30the quarter and it did so once again18:33this quarter. It beat on earnings but18:35missed on revenue. Now I really don't18:37understand why Wells Fargo seems to miss18:40every other quarter or so. I just don't18:42get it. But they just do it. Earnings18:44per share was $187 versus $166. Up 13%18:48versus last year. That was a beat.18:50Revenue missed. Total revenue was up18:53only 2% versus last year. In fact, Wells18:56has not grown revenue in multiple18:58quarters. The company just seems to have19:00trouble growing its revenue. If you are19:03interested investing in a large bank,19:05this is not the one I would look at.19:07Let's move on. Under the leadership of19:09Jane Frasier, City continues to report19:11good results. Company reported earnings19:13per share of 285 versus $186, up 53%19:17versus last year. And while higher rates19:19hurt the fixed income results of19:20Goldman, City did fine. The return on19:23tangible common equity was at 11 and a19:25half percent. That was lower than the19:27recent 13% achieved in the last two19:30quarters, but still pretty good for19:31city. Right now, we are in a pretty19:33golden age for investment banking.19:35Trading volumes are high. AI is creating19:37massive financing needs and M&A is19:39strong. And that's why despite the19:40weakness in fixed income trading,19:42investment banking results remain strong19:43at Goldman and Morgan Stanley. Goldman19:46reported a good quarter of 1701 earnings19:49per share versus 1225. So that's 39%19:53growth. The overall results were good,19:55but if you want to get nitpicky, FIC,19:57which stands for fixed income trading,19:59was weaker than expected. On the other20:01hand, equities did very well, up 72%20:03versus last year. Goldman's results20:05indicate that higher rates can hurt the20:07fixed income trading business as clients20:09take a pause because higher rates scare20:11them. And like Goldman, Morgan Stanley20:13produced powerful numbers. Revenue,20:15while not at record levels, was up 12%20:18versus last year. EPS of 303 was up 8%20:21versus last year and the return on20:24tangible common equity was a very strong20:2623%. As for how to value banks, the20:29easiest and most common and consistent20:31way is to look at the return on tangible20:32common equity. The higher that20:34percentage, the higher the price to20:36tangible common equity. That's why20:38Goldman, JP Morgan, and Morgan Stanley20:41are valued at 2.6 to 3.4 times tangible20:44book. Their return on tangible common20:46equities exceed 20%. At the same time,20:49Wells Fargo, Bank of America, and City20:51are at 1.2 to 1.8 times because their20:55return on tangible common equity is 1220:58to 16%. What should you do about21:00investing in the banking sector right21:03now? Now, some of the froth has come out21:06since the July peak, but frankly, not21:08that much. To invest in banks at these21:11valuation levels requires you assuming21:13that the investment banking cycle21:15remains very strong. I said exactly that21:18when the banks reported second quarter21:20results and I reiterated there is just21:22not much margin for error at these21:25valuation levels. To me, it's just not21:28that great an entry point. This week on21:30October 12th on our free service, we21:32posted an interview with recurring guest21:34Gary Marcus. We discussed all the21:36controversies surrounding AI, what's21:39real and what's not. Next week on21:41October 17th on our free service, we21:43will release an interview with Danny21:45Moses, my former partner for My Big21:47Short Days. Danny now runs his own21:49podcast called On the Tape, and we had a21:51broad ranging discussion about the21:53deficit, the Fed, and controversies21:56surrounding AI. The best way to support21:58the Real Eisman Playbook is to subscribe22:00to Substack through steveisman.com.22:03And we greatly appreciate your support.22:06[music] And that's the wrap.22:09This podcast is forformational purposes22:11only and does not constitute investment22:14advice. [music] The hosts and guests may22:15hold positions in stocks discussed.22:17Opinions expressed on their own and not22:19recommendations. Please do your own due22:21diligence and consult a licensed22:23financial adviser before making any22:24investment decisions. [music]
3,689 words · 560 lines
