Califoreboding

MICHAEL CEMBALEST: Good morning, everybody. This is Michael Cembalest with the October 2026 Eye on the Market podcast. This one is called "Califoreboding," which I will explain. I was seeing clients on the West Coast, mostly California, last week. 

And I have to say, California, as everybody knows, is a very, very strange place. The energy costs are 50% higher than the national median. It's got the highest individual income tax rate. It's got the sixth-highest corporate income tax rate. It's got the least affordable homeownership market. 

It's got the largest number of state-level regulatory restrictions, as a lot of our clients are constantly reminding me. It has the highest state unemployment rate at 5%. And it also has the highest rate of unsheltered homeless people at 66%. In other words, 2/3 of all of the homeless people in California are unsheltered. 

And yet over the last year, California has experienced the third-highest state GDP growth in the country, at about 3 and 1/2%, behind only New York and Idaho, which is a result that, of course, is largely attributable to AI. 

I have a chart in here on the energy costs in California. We'll talk more about this in the energy piece. But what we do is we take the actual fuel sector mix in California in terms of natural gas and electricity, renewable fuels, diesel, gasoline, as they're actually consumed by transport, industrial, commercial, and residential sectors. 

And then we take that same-- and then we figure out how much that cost per BTU. And then we do the same thing to the other states. And there is a whopping premium to do business in California. 

I don't have that much on the midterms. Either way, no matter how things turn out, I don't think it's a huge market mover, at least in the near term, no matter what-- how it turns out, although I will say, I saw this one chart that I'm including here because I thought it was hilarious. 

There was a poll by Ipsos that does a lot of polling. And they asked different respondents, do you think the level of corruption under the Trump administration is better or worse than other recent presidential administrations? 

And only about 10% peak of the GOP respondents said it was worse. About 50% of them said it was better. And 30% said it was the same. And so I'm not going to go into any other detail on it other than to say that this is one of the funniest charts I've seen in a really long time. 

So the title of this Eye on the Market is "Califoreboding." I appreciate the fact that clients are nervous at a time when the markets are all-time highs. I think it's a-- it reflects a healthy discipline in terms of a sell high, buy low mentality. 

But there were a lot of questions that were very much on the-- of negative foreboding of things to come. And specifically, we talked about the elevated share of market returns and growth attributable to AI, how sustainable that is, Anthropic's warnings about open-wave models, stress in private credit, European deindustrialization, and the inability to keep Chinese imports out. 

And then maybe the most important topic was third-rail questions about the US federal debt now that 10-year treasuries have blown through the level that existed in August at the time of that Treasury "Twist" operation. 

So let's get into it. If you're concerned about the sensitivity of the US economy and financial markets to AI, I agree with you. And so I just want to show you a few slides that stress the importance and describe them for those of you listening on audio just to describe just how concentrated things are. 

So this first one you've seen from us before. Somewhere between 60% to 80% of all of the returns in the market, earnings growth, and capital spending has come from a basket of 42 AI-related stocks since January 2024. So that's kind of remarkable because essentially, less than 10% of the S&P is accounting for 60% to 80% of earnings growth returns and capital spending. 

Another one is even though we're pretty close to all-time highs, about 60% of the stocks in the S&P 500 are more than 20% below their all-time highs. And that is a very unusual thing, as we show in one of these exhibits. 

Normally, those numbers would range from 20% to 40%. And now it's 60%. In other words, 60% of all the stocks are trading more than 20% below their all-time highs, even though the market itself is at the top. 

The share of S&P 500 companies outperforming the index on a three-year trailing basis-- about a 30%, one of the lowest numbers we've seen since the early 1990s, again, another measure of just how concentrated the returns are. 

And this is a really important one. Now, I'm expecting this to change. And it's early innings yet. But if you break down the S&P 500 into four buckets, the companies that are building and supplying all the AI-- so that's going to be a combination of utilities, some industrials, and then the hyperscalers, and then the semiconductors and optical networking companies-- it's about 55 companies. Their earnings have skyrocketed. 

If we then take the whole rest of the market, which is 450 companies, and we bucket them into high, medium, and low AI adopters, we're really not seeing yet any huge difference in terms of earnings momentum from the high AI adopters, maybe a little bit higher than the low AI adopters. But we're still in the infancy of something that can be described as transformational for the users of AI, even though it has been completely transformational for the companies that are providing it. 

Technology's contribution to the economy-- we didn't even talk about technology contributions to the economy in 2023 because it was so small. And now it's-- the last couple of quarters, it's been almost a third. So that's really unusual. 

And here's another interesting one. What are the share of states with GDP growth greater than the national average? That number should be somewhere around 40%-50%. It's 20%. So only 20% of the states are growing faster than the national average. And that's consistent with the kind of skew that you're seeing in the stock market. 

And then I just want to spend a minute on construction spending because there's a lot of commentary that, oh, wow, this is fantastic. Look at the construction spending on data centers and power generation-- are soaring. 

Yes, they are. But almost every other bucket of construction spending is falling, particularly manufacturing, which was supposed to be one of the things this administration was focused on reversing. They inherited a declining manufacturing construction spending from the Biden administration. And it's kept on declining. 

And so when you actually put all the numbers together, even though this AI bucket of data centers and power generation has doubled, in dollar terms, it's not large enough to offset the declines taking place elsewhere. So since the beginning of 2024, overall US construction spending is still declining. 

So what that tells us is, yes, grandma, there is an enormous amount of sensitivity in markets to this AI theme. I don't think the AI theme is going to run aground in the next 6 to 12 months. But when and if it does start to slow down, you're going to see some pretty strong reverberations in the markets and the economy. 

Now, on the plus side-- this is another topic that a lot of clients are asking about-- what about corporate profits and the fact that the Fed is hiking? Normally, when the Fed raises interest rates, certainly in the last few cycles in the '90s and the early 2000s, and then around the financial crisis, when the Fed hikes policy rates, you see a spike in corporate net interest costs as a percentage of profits, and the idea being that policy rates go up and it pushes up all sorts of floating-rate debt that companies have. And all of a sudden, their interest costs take up a bigger share of their after-tax profits. 

For whatever reason, and I could theorize this-- the reason why, but for whatever reason, the rate hike cycle that began a couple of years ago did not do that. And corporate net interest costs, unlike the government, where interest costs are soaring, corporate interest costs are still falling as a percentage of after-tax profits. 

And so that's a positive. And essentially, what that's telling you is that the corporate sector has probably extended its duration pretty substantially and is less sensitive to policy rates than it used to be. 

And then the big question on valuations is-- I think two things could be true at the same time. I think we came into this AI boom, let's call it, two years ago, Jan '24. Let's put the marker at Jan '24, when this thing really started to accelerate. 

I think it's true that we came into that with valuations already elevated. But as far as we can tell around the world, they haven't gotten worse. And if anything, if you decompose the returns on technology stocks in the US and in Europe and in Asia and in emerging markets more broadly, almost 100% of the gains have been driven by earnings growth. Multiples in a lot of places have actually flight. And so I thought that was interesting. 

I think Japan is the only place in the world where you've actually had multiple expansion accounts for the majority of the market growth. Everywhere else, it's all been earnings-driven. And that provides a little bit of a cushion for the kind of collapse that people are concerned about. Markets driven by earnings growth are-- tend to be less sensitive than markets driven by multiple expansion. 

And then what about SpaceX as a bellwether? I think it's-- when you start thinking about OpenAI and Anthropic and some of big IPOs coming, I was-- a lot of us are following SpaceX very closely. And certainly, the last 10 or so big tech IPOs since, let's say, 2010 got clobbered, a lot of them, as their lock-up periods-- as the end of lock-up periods approached.

So we're only on day 116. But SpaceX is holding up better than most of them, better than Coinbase and Snap and Lyft and Uber and Rivian and Meta. It's not doing as well as Palantir and Airbnb. But it's in the upper half. 

Now, only 20% of the stocks-- only 20% of the market cap is part of the free float. And there's going to be more coming. But I thought this was positive, that SpaceX was holding its own at this point, 116 days since the IPO. 

Another shift to a slightly different topic-- I-- we got a lot of questions, even though we're in California, on nervousness about what's going on in Europe. And I can understand why. I think part of the political stuff that you're seeing is the byproduct of economic stresses. And so when you start to see the political issues in Germany and France and in Italy, look at what they're facing here. 

This is a table on Volkswagen's manufacturing economics by plant. And this is one of the more jolting things I've ever seen. And I have known each one of-- what I'm-- I've known each one of the tidbits I'm about to tell you for a while. I just have never seen them all laid out in the same place before. 

So what this is looking at is a VW plant in Germany, in Portugal, and in China. And so let's tick through this. Manufacturing labor costs are 15% to 30% of what they are in Europe-- no surprise there. But here's where it gets more interesting. 

Sick leave rate is 10% to 20% in China of what it is in Europe. The number of working hours in a year because of that-- 1 and 1/2 times higher. That's a huge premium in terms of the number of working hours per employee. 

The annual vehicle output per employee-- anywhere from 1 and 1/2 to 2 times higher. And all of that translates into factory costs for vehicles that are 20% to 40% in China of what they are elsewhere. And then you combine that, because a lot of these are EVs, with electricity costs that are cheaper in China. 

So because of all that, I'm not surprised to see this next chart, which shows Chinese auto brands, whether it's BYD or other ones, gaining substantial market share over the last 12 months in Europe at the same time that all the European mass brands, premium brands, US brands, and other Asian brands are actually losing market share. 

And Europe says that they're now prepared to act against China with their own version of the Section 301 tariffs like that the US has if China doesn't meet their list of demands. I have to say, their list of demands-- I'm not sure how China is going to respond to this without chuckling. 

They want China to adopt voluntary export restrictions. And they want to make the provision of rare earth export licenses more predictable. And they want China to unilaterally reduce its industrial overcapacity. And they want better market access for European companies. 

Good luck. The US failed getting any of that out of China since China joined the World Trade Organization in 2001. And I doubt that Europe's going to be able to get it out of them. 

China appears to have the upper hand here. It exports more rare earths and permanent magnets to Germany alone than it does to the entire United States. And so I'm not really anticipating much progress by Europe in this kind of thing. 

All right. Slightly different topic-- I was in California. So of course, there was a lot talk about OpenAI and Anthropic. I thought it was interesting. Anthropic released what is best described as a hit piece on open-wave models. It was a sharply worded hit piece on a model called GLM, which comes from one of the Chinese companies. It's an open-wave model from Z.ai. And it's the kind of thing that generates similar performance to the frontier models at 25% of the cost per task. 

So what did Anthropic do? I don't know. I think it took a few thousand hours. And it cost them $5,000, just $5,000, which was the cost of my plane ticket, I think, to San Francisco. Anthropic engineers years were able to abliterate the GLM model. 

What does that mean? It's an important word that you need to know. It's not "obliterate" with an O. It's "abliterate" with an A. And abliteration refers to how you can go in to these open-weight models and tinker with the weights with the-- for the express purpose of disabling all the safety restrictions. 

And so what was interesting was the Anthropic said, before we tinkered with it, this model, GLM-5.3, had similar safety controls in terms of refusing to do bad things as Mythos Preview, Opus 5, and Opus 4.8 from Anthropic. Once they abliterated the model, its refusal rate on bad requests went from 97% to 5%. In other words, few thousand hours and $5,000, and they were able to completely remove any semblance of safety controls in this model. 

And so, yes, it's a hit piece. But I have to begrudgingly concede that Anthropic is making a pretty good point here about what happens when bad actors get a hold of these open-weight models. 

And by the way, when they did this, there was no degradation in performance of GLM. The performance was just as good. It's just that the safety controls were gone. 

No sign yet of how the administration is going to handle this. I do think something's going to have to give. It's partially a signal flare from Anthropic that they're trying to close the moat before the IPO. But at some point, I think the administration is going to have to do something here, maybe some kind of restrictions on enterprises adopting open-weight models that are this easily tinkered with. 

All right. Let's talk about the the 10-year Treasury yield. So the 10-year Treasury yield was drifting up a few basis points at a time for most of the year. And then the Treasury secretary said, I dare you to cross this line. And everybody did. 

And so the Treasury has been selling off more aggressively after the Treasury August "Twist" operation than before. And so I'm getting tons and tons of questions from clients on the federal debt. 

Now, look, I don't think this is the final denouement of US overindebtedness at the federal level. I don't think that's what's happening. I think this current rise in yield is a function of inflation that's still a little bit sticky over the Fed target, tons of hyperscaler debt, and so into the long end of the curve, which we've written about a lot over the last couple of months, and GDP growth that's actually above expectations. If the fuse were lit under this administration, I guess it wouldn't surprise me. But I don't think that the fuse is being lit right now. 

That said-- and my hope is to be retired before this happens. But in 2030 or 2031, we will reach this crossover point where 100% of government revenues in the United States will be needed just to pay entitlements and interest on the debt. 

And so that historical gap that existed between that-- in other words, from 1965 until this year, with the exception of the COVID recession-- there was more government revenue than just what you needed to pay on entitlements and interest. That won't be the case anymore starting in 2030-2031. 

And you could anticipate that there would be some kind of bond market riot at some point then. And you could anticipate that you might see rating agency downgrade. You can anticipate that there might be some kind of sovereign wealth fund reluctance to participate in Treasury and agency auctions at the same pace. 

To me, when you look at the TARP, for instance, how the TARP vote failed in the Congress, and then the markets fell again, and then the TARP was adopted, there are some very, very, very difficult and painful choices that will eventually have to be made. My sense-- and I think this is a positive, what I'm about to say. 

I think at some point, there's a bond market riot. And rather than having the Fed come in and monetize the debt, my expectation is you will get, after that bond market riot, enough politicians using that as cover to make some very difficult decisions. 

And we'll talk about this more in the future one day. And I'll probably write about this in the outlook next year because we are getting a little bit closer to 2027 outlook. But when you look at the deficit reduction options that are on the table, they've all been well known for a really long period of time. The Bowles-Simpson report many years ago walked through them. We have a chart in here on what the big ones are that move the needle. 

You're going to need-- and it's almost all tax increases and cuts to mandatory spending. There's almost no discretionary spending left to cut. It's already been cut. And so elimination of itemized deductions, a new uncapped payroll tax on earnings above a certain level-- so taxes on different kinds of income-- reduced tax subsidies for certain employment-based health benefits, and then lots of changes to means testing of Medicare, changes to Social Security indexation, taxation of VA disability payments-- all of these things are enormously unpopular in both parties. But these are the options that are on the table. 

And let me just mention also, probably at some point, more strict caps on contributions to qualified retirement plans, particularly for wealthy people-- So this is what's going to have to happen one day. And we should probably all get well versed on what these different things are before it happens. 

Last topic for this month-- there's a lot of discussion about what's going on in private credit. And we've written before extensively and in the last two alternatives papers that I write in December of every other year about how the underwriting standards a few years ago, maybe five years ago, seven years ago, in private credit were substantially tighter and better and stricter than underwriting standards in the broadly syndicated leveraged loan market. 

But the more capital flowed into private credit, the weaker those underwriting standards got-- happens all the time. So now people are tracking the stress in private credit because measuring stress tends to be a precursor to actually-- actual realized defaults, which is what people track in the first place. 

But it's as much art as science in terms of, what is the level of stress in private credit? Some of the sources that you'll see out there are saying that it's only 1%-- stress is only one to 3% of the book when they're looking at things like BDCs. And examples include Lincoln International, S&P, Houlihan Lokey, Proskauer Rose, KBRA. These are the usual suspects. And their stress levels are just 1% to 3%. So what's the problem? 

The issue is that the definitions of stress from all these providers are different. Some of them include covenant default. Some don't. And so I went to our risk management and private credit teams in Asset Management. And I said, let's do this the right way. If we start out with a blank sheet of paper, how would we track stress in private credit? 

The good news was, hey, Mike, they said, we're already doing this. And so we made-- I made some adjustments to a couple of definitions. But we agreed on the following five categories, which is any loan that's been deemed non-accrual by the lenders, a distressed PIK, which refers to a loan where the borrower decides to capitalize interest instead of paying it, a maturity extension of more than two years for loans that are priced below 95% of par, any loan that's been foreclosed upon and converted some debt to equity, and loans that are valued below $0.80 on the dollar for any other reason other than the ones just listed. And this stress level has gone from-- has doubled since 2022. It's gone from 8% to 16%. 

And so what is that telling us? It tells us that a reasonable chunk of the private credit is defined by the BDC universe of positions are experiencing some kind of stress. 

But just like private equity and venture, which have a historically low rate of monetizing positions from vintages like 10 years ago. The private credit industry is also kicking the can down the road and deferring the price discovery and loss realization on some of these positions to another day. 

And I can't say that I blame them. That's probably what a lot of us would do if faced with these kind of decisions. And it's certainly, by the way, what banks do and when the residential mortgage book-- when people can't pay, they-- first option is always a restructuring rather than a foreclosure. But it's important to understand that someone that says that the level of stress in private credit is only 1% to 3% is probably not counting the kind of things that you and I would count as relevant. 

That's it for this month. Thank you for listening. And I will see you all on this podcast after the midterm elections. Thank you very much. Bye.