The Credit Market Warning That Could Hit Stocks | Jeff Snider

Jeff Snider · Atlas Financial

Rising long-term yields reflect central banks leaving zero rates and interfering at the front end, not deficit fears.

Snider rejects the common explanation that bond markets are reacting to government deficits or debt limits. Central banks are coming off zero-interest-rate policies, so they interfere more at the front end of the curve. That pushes longer-term rates higher than in the 2010s and early 2020s. Markets are reassessing the rate environment and the premium for central-bank interference. Current rates are not historically high. The 30-year at around 5.50% is historically low. He says the right approach is to look at spreads, not nominal levels.

“What's really happening is central banks are coming off of zero interest rate policies, and therefore the front end of the yield curve is being interfered with more and more by central banks...”at 19:11

Jeff Snider · Atlas Financial

Government bond yields will fall through the rest of the decade, with the 30-year heading toward 3.50% in a credit downturn.

Snider expects government bond yields to go lower through the rest of this decade because markets keep betting on that, and all the market indications and curves keep pointing there. If the credit cycle downturn deepens and macro impact grows, the 30-year could move from about 5.50% to 4.50%, perhaps 3.50%, quickly. Bonds would also act as a hedge in a credit downturn.

“I expect that government bond yields are gonna go lower through the rest of the rest of this decade because that's what the markets keep betting on and keep all the market indications and curves keep indicating.”at 19:54

Jeff Snider · Atlas Financial

Long bonds would rally as a hedge in an AI-linked credit downturn; 30-year could fall to 3.5%.

In a credit cycle downturn involving AI, the 30-year yield would fall from about 5.5% to about 3.5%. Buying duration is therefore a hedge against downside in the stock market. Bonds have underperformed in pockets since 2022 and 2023, but that does not mean rates have a clear path to double digits, since a downturn lowers growth and inflation expectations and pushes yields down.

“if we have a credit cycle downturn that involves AI, the thirty year is gonna go from five and a half to three and a half. So you can buy duration and be happy with duration as a hedge, at the very least, against being, participating in the downside of the stock market.”at 41:43

Jeff Snider · Atlas Financial

Treasury yields reflect growth, inflation expectations and Fed front-end policy, not deficits or debasement.

Bond yields are driven by growth and inflation expectations plus central bank interference at the short end, not by deficits or dollar debasement. Futures, swaps and forwards show no debasement trade and no risk of Treasuries becoming unnegotiable. The 10-year sits near 5% and the thirty-year rise is explained by central bank policy, not inflation. A downturn would bring lower growth expectations, lower inflation expectations and lower rates.

“It's not deficits, it's not dollar debasement, which by the way is not happening, it's never gonna happen. It's really growth and inflation expectations, plus interference from central banks at the short end.”at 45:01

Andrew Sarna · Forthlane Partners

Bonds are a trade, not a long-term hold; 60/40 is not back and bonds shouldn't be 40% of a portfolio

Sarna says bonds are a trade rather than a long-term hold, mainly because of the deficit outlook and expected bond pressure. He does see some value now. Yields around 5% on the 10-year are far better than the roughly 2% of a couple of years ago. Yields are a good predictor of forward returns, especially at the short end of the curve. For individual investors, taxes are a headwind. He says the 60/40 portfolio is not back and bonds are certainly not 40% of a portfolio.

“I think it's a trade, not a long term hold.”at 1:12:41

Andrew Sarna · Forthlane Partners

The 10-year Treasury yield has a ceiling near 6% because the budget doesn't work above it

Sarna argues there is a ceiling on how high the 10-year yield can go because the government's budget does not work if the 10-year is at 6%. He would not be surprised to see yields eventually fall, though he cannot say when. He suggests investors can reach for a bit of yield pickup on the duration side for that reason.

“I think there is a ceiling to how high sort of the ten year can go because, you know, the government is... The budget just doesn't work if yields are at... Or the ten year is at 6%.”at 1:14:06

Jeff Snider · Atlas Financial

Debt-funded AI overcapacity is building, so it's better the bubble pops now than in two years.

Snider says he hopes the AI bubble pops now rather than two years from now. Two years from now, many more trillions will be tied up in overcapacity in AI, whether in computing, technology or equipment. The fear is that everyone is building the same capacity, so some firms will lose and their debt will need to be wound down. The damage will be much easier to handle if the bust comes sooner, when the amount at stake is smaller.

“two years from now, you're going to have however many more trillions attached to overcapacity in AI, whether it be computer, technology, or equipment. It's much much easier to handle if it's, say, 2,000,000,000,000”at 6:03

Jeff Snider · Atlas Financial

Vendor financing and GPU-collateral deals repeat the dot-com circular financing, with about $500B of debt at risk.

Snider says the circular financing in AI is the same as dot-com vendor financing. Leaders use their credit profiles to lend to startup customers so those customers can buy their products. The structures are bigger and more complicated now. They include GPUs as collateral and residual value guarantees from vendors. The bust phase will show who was exposed. Some AI firms that borrowed massively will lose, leaving roughly $500 billion of debt to be wound down and resolved.

“That's really the circular finance we're talking about. We don't need to make it complicated. Yes, it gets very complicated with some of these structures, but really all it is is vendor financing.”at 26:45

Jeff Snider · Atlas Financial

Government backstops won't fix AI data-center overcapacity or rescue individual AI stocks.

The US calls AI close to a national security issue, but that doesn't mean it will rescue individual stocks. Government won't take over unfinished data centers or buy leftover compute to keep token economics from deflating. Overcapacity is a different problem from the government's usual stimulus response. Markets also treat a larger government response as a negative sign, because it signals a problem serious enough to need one.

“Are they gonna take over all the data centers that don't get finished and promise to rebuild them? I don't necessarily think that's the case, and that's really what we're talking about with overcapacity.”at 37:09

Andrew Sarna · Forthlane Partners

AI compute is a commodity business at risk of overbuild, with demand collapse leaving spenders exposed.

The market is still undersupplied on compute, but hyperscaler capex is expected to exceed $1 trillion by 2027 excluding SpaceX. A SemiAnalysis estimate says that hitting the 50-gigawatt global target in 2027 would equal all compute built up to 2024. Compute is ultimately a commodity business. Past commodity cycles followed a gold rush, then an overbuild, then a demand collapse, leaving companies that poured everything into capex holding the bag. He struggles to see why companies would stop investing before the overbuild occurs.

“the real concern is this is ultimately a commodity business. And the thing that I struggle to wrap my head around is like, is gonna cause these companies to stop investing before the overbuild occurs?”at 1:06:44

Andrew Sarna · Forthlane Partners

Compute shortage persists, but an eventual data center overbuild is plausible as capacity comes online

Hyperscaler debt issuance is up 357% year over year. Hyperscalers used to issue roughly $5-40 billion a year, and next year is expected to reach about $40 billion. Sarna still sees a compute shortage, so he doesn't think it's time to panic. But a lot of capacity is coming online and it is unclear where demand will be, so he would not be surprised by an eventual overbuild. Negative political rhetoric on data centers ahead of the midterms is growing, and even in Texas data center capacity additions appear to be leveling out. He suggests a decision from elsewhere could force hyperscalers to pull back and save the market from overbuilding.

“There's still a shortage of compute. I wouldn't be surprised if we eventually get to a place where there is an overbuild.”at 1:08:48

Jeff Snider · Atlas Financial

AI is following every previous disruptive-technology bubble, and its debt-funded capex makes it a credit story.

Snider argues AI is following every previous disruptive-technology bubble, and there is no reason to think this time is different. He says the difference is the amount of debt attached to AI capital expenditure, which makes it more a debt story than a stock-market story. Credit markets show signs the cycle has turned: funders are reluctant to finance AI in unlimited quantities at the lowest rates, and anxiety is spreading beyond AI. He calls the AI numbers 'absolutely ridiculous' and says the math doesn't work. The stock market is pricing a smooth AI adoption path into the 2030s, which he thinks is increasingly unlikely as credit repricing continues.

“The problem is, we gotta get there first, and unfortunately, [AI] is following every previous disruptive technology bubble that we've ever seen, and there's no reason to believe that this time is any different.”at 2:18

Jeff Snider · Atlas Financial

The AI bubble is more likely to deflate slowly like the dot-com bust than crash sharply.

The bigger risk is a .com-style reaction to the AI bubble story, where the air slowly leaks out. A sharp crash lets markets shake out and recover sooner, whereas the dot-com bust took many years to recover from. Nobody knows when the bust arrives, how bad it will be, or which path it takes. The credit cycle is the place to watch first, because it sends the clues that flow downstream to equities.

“the biggest risk is, I think the bigger risk, is that you have a .com style reaction to the overall AI bubble story, which leads to the air kind of slowly inflating out of the AI bubble.”at 34:05

Andrew Sarna · Forthlane Partners

The AI trade will sell off sharply several times during the infrastructure buildout

Sarna says the AI trade has become so central that nearly every market now depends on it, including debt markets. He points to recent weeks where technology dragged the index lower while most sectors rose, plus blowups elsewhere such as European credit and biotech. He is fairly confident that through the buildout the AI trade will repeatedly sell off, so investors should check how exposed they are rather than holding 80% of a portfolio in the S&P 500 and the Qs. He admits timing the top is a fool's errand.

“one of the things that I'm fairly confident is eventually through this build out, we're gonna see the AI trade puke a few times.”at 1:10:59

Jeff Snider · Atlas Financial

Public credit stress is spreading: leveraged loan and junk bond spreads have risen since last September.

Snider argues the credit cycle downturn is not contained to private credit. Stress has spilled into public markets, especially leveraged loans and risky junk bonds, where spreads have continuously risen since last September. He sees this consistent deterioration over more than a year as the most compelling evidence that the credit cycle is turning and developing. He frames credit cycles as multi-year processes, not three-week collapses like 2008, so the trend is a confirmation signal rather than a single event. Energy shock and AI-related fears feed into this repricing of risk.

“Instead, we've seen public markets, especially, you know, leveraged loans, risky junk bonds. Spreads there have continuously risen ever since last September.”at 13:32

Jeff Snider · Atlas Financial

The credit cycle leads the downturn; widespread credit signals should trigger reducing equity risk.

The bubble-bust process begins with the credit cycle, and credit, bond and rates markets provide the big clues about what happens downstream, including for equity investors. Credit signals are one way to decide when to exit the market. When they become substantial, validated and widespread, that may be the point to step back, wherever stocks happen to be.

“it starts with the credit cycle, so that's where you want to focus. Even if you're an equity investor, you want to pay attention to what's going on in credit markets and bond markets and rates markets, because those will be the big clues”at 34:46

Andrew Sarna · Forthlane Partners

Credit spreads are tight, so investors aren't being paid enough to take credit risk

For the most part credit spreads remain pretty tight. As a result, investors are not really being compensated for taking credit risk. That leads him to favor reaching for yield through duration rather than credit.

“for the most part, spreads remain pretty tight, so you're not really being compensated to take credit risk.”at 1:13:57

Jeff Snider · Atlas Financial

Markets are not priced for downturn scenarios, and few have hedges, so a small rise in tail risk would hit hard.

Snider argues that the credit market upside has been priced for continued strength, and nobody is priced for downturn scenarios. Few investors have adequate hedges. So even a small rise in the odds of a bad outcome, such as a major firm failing, could cause a substantial market backlash. Expectations are ratcheting down as information arrives: the economy isn't resilient, debt is larger than thought, and adoption is slower. Investors will reprice risk based on that new information.

“nobody is priced for downturn scenarios. Nobody has enough hedges and everything.”at 30:51

Andrew Sarna · Forthlane Partners

Markets are over-anxious about the AI trade and he expects a rebound.

Markets are extremely anxious and everyone is hyperventilating about the AI trade. The concern is warranted in part and unwarranted in part. Compute is still undersupplied and the long-run overbuild risk is real, but investors are overreacting in the near term. He thinks the market is probably a little over-anxious and expects a rebound.

“[I] think the market is probably a little over anxious right now. I think we're going to see a rebound”at 1:07:25

Andrew Sarna · Forthlane Partners

Sentiment has flipped to extreme fear, so a year-end Santa Claus rally is plausible

Months ago people were posting portfolio return screenshots on X, which is a classic sign to lighten equity exposure. Now the opposite is happening, with everyone paranoid that the AI trade is popping. Sarna contrasts this with the earlier euphoria. He wouldn't be surprised if three months from now a Santa Claus rally has happened and the market was simply over-anxious.

“I wouldn't be surprised if we're talking three months from now and that Santa Claus rally did happen, and everyone was just over anxious.”at 1:17:28

Jeff Snider · Atlas Financial

Private credit underwriting during the bubble years was egregiously poor, as repeated frauds have exposed.

Snider points to a string of frauds, including Tricolor and First Brands, which he calls 'cockroaches'. These weren't large individually, but they showed how bad underwriting was in private credit during the roughly 2021 to 2025 bubble period. Lenders focused on volume and bonuses and did little due diligence. The bust began last year, when investors pulled money out of private credit funds after negative payroll reports made lenders start rechecking borrowers. The backlash was triggered by the exposed underwriting, not by AI.

“[They] demonstrated repeatedly one after another after another, is just how bad the underwriting had been during the bubble period in private credit.”at 7:30

Jeff Snider · Atlas Financial

A shadow bank run is a real possibility, though not the base case, with Apollo-type lenders the risk.

Snider says there is a very real chance of a shadow bank run at some point. He stresses that this is not the base case but a real possibility. A shadow bank could become insolvent, making a firm like Apollo the equivalent of Lehman Brothers. The contagion risk is more limited than with commercial banks, because shadow banks lack depositors. The losses would fall on the equity of wealthy individuals and on the back-leverage providers and lenders who funded those shadow banks. Opaque structures and information asymmetry mean that once one part of the system goes wrong, it is hard to contain. Maggie Lake noted the example was illustrative, not a prediction about any firm.

“there's a very real chance that we do have a shadow bank run at some point. Again, that's not a base case, but it's certainly a possibility where Apollo becomes insolvent.”at 28:43

Jeff Snider · Atlas Financial

The US economy is fragile, not resilient, with years of weak hiring and an energy shock skewing risks to the downside.

Snider rejects the idea that the economy is resilient just because it avoided a 2008 or 2009 collapse. He says the right test is the cycle on its own terms. The economy has seen job losses and weak hiring for several years, and most people feel it. Terms like K-shaped or two-speed economy are admissions that it isn't resilient. When a fragile situation meets an energy shock, the probabilities and scenarios skew more to the downside. That doesn't mean 2009-style layoffs, but performance can keep getting worse.

“We have seen job losses and lack of hiring first and foremost for several years now. And that's the economy that most people are actually living in...”at 22:58

Andrew Sarna · Forthlane Partners

AI capex and fiscal spending are the two growth drivers; an AI capex slowdown would signal weakness

Sarna sees AI investment and fiscal spending as the two main drivers of US economic growth right now. If AI capex falters, that is the likely sign that economic weakness is coming and the economy begins to roll over. At that point he expects yields to fall and investors to get out of the AI trade.

“I view you have two strong or you have two drivers of economic growth right now is really AI and fiscal spend. So I think if you see the AI CapEx falter, that's maybe the sign that you're going to see some economic weakness”at 1:14:29

Jeff Snider · Atlas Financial

Energy shock with $100 oil and expensive fuel is squeezing corporate margins, especially mid-market firms.

Snider cites roughly $100 per barrel oil, gasoline at $4.50 to $5 a gallon, and diesel possibly at $6 to $8. The energy shock, now about eight months old, is hitting corporate margins, particularly in the middle market. This comes on top of an already fragile economy. He cites trucking companies and a large Wendy's franchisee going bankrupt as evidence of margins being squeezed. The energy shock feeds back into credit stress and raises the risk of a downturn.

“You throw in $100 per barrel oil... suddenly you have, especially in the corporate middle market space... then you got energy shock that's really hitting margins.”at 9:11

Jeff Snider · Atlas Financial

Banks are exposed to the credit downturn but less than in 2008, so a Lehman-style panic is unlikely.

Snider argues the current credit cycle differs from the last one, which centered on bank balance-sheet expansion. Banks have acted more responsibly. They are still exposed, but not the way the whole system was in 2008. So a downturn is unlikely to become a banking-sector panic with a Lehman-style failure. The downturn can still cause a credit crunch, job losses and financial volatility, with shadow banks as the more likely failures.

“Banks have acted responsibly. That doesn't mean that they're not exposed, because they certainly are, but they're not exposed in the same way that everything was in 2008.”at 16:02

Jeff Snider · Atlas Financial

French bond yields carry a political-uncertainty premium, not a sign that French fiscal limits have been hit.

Snider says French government bond pockets of concern are not due to the market judging that France has hit its fiscal or debt limit. The deficit and debt are part of it. The main driver is political uncertainty, and therefore social uncertainty, which the bond market finds hard to price because the outcome could go in one of two directions. The French case is unique compared with other European government bonds and the US.

“it's not the market saying that the French fiscal situation has hit its limit, it's really the market is [pricing] an uncertainty premium to French debt over the political uncertainty.”at 19:11

Jeff Snider · Atlas Financial

The Fed will overreact to the energy shock and raise rates again, as it already did once.

Snider argues the long end of the Treasury curve is repricing partly because the Federal Reserve interferes at the short end. He asks whether the Fed will overreact to oil prices and the energy shock and raise rates. His answer is yes. The Fed already did it once and is likely to do it again. This adds to the higher premium for central-bank interference that markets are pricing in the 2020s.

“Is the Fed gonna overreact oil prices and the energy shock and thereby raise rates? Well, answer is yes. They already did once, and they're likely to do it again.”at 21:03

Jeff Snider · Atlas Financial

Stocks will wobble as AI support fades and valuations look obscene, but a slow deflation beats an October 2008 crash.

Snider argues equity prices are less supported as credit markets reprice AI risk, and many equity values have become 'absolutely obscene'. This makes the stock market grow wobbly. He doesn't expect an October 2008-style crash. He compares it to the dot-com bust, which took three years to bottom with repeated false rebounds. The biggest risk he sees is a dot-com-style reaction to the AI bubble, where the air slowly leaks out.

“There's less support for equity prices and equity values that are gotten, in many places, gotten to be absolutely obscene. So the stock market grows a little bit wobbly, but that doesn't mean that it turns into, you know, October 2008 either.”at 33:33

Jeff Snider · Atlas Financial

Markets price no accelerating inflation; TIPS breakevens have been sideways for years.

If markets expected CPI to get out of control, demand for TIPS would surge. Instead TIPS breakevens have been sideways for years. Inflation is elevated versus the 2010s but not accelerating. The 2021-22 supply shock was transitory, and each oil spike, including the energy shock earlier this year, produced only a short disturbance. The five-year breakeven hit a multi-year high, but from a low base, so it is not a 1970s replay.

“the marketplace was saying, yes, you're gonna get some short run disturbance in the CPI based on oil prices, but it is not going to be some kind of 1970s replay.”at 47:51

Jeff Snider · Atlas Financial

Silver's run to $120 was a supply squeeze; he expects silver to fall toward $30 per ounce.

Silver's move to $120 was an obvious supply squeeze, visible in London and Shanghai lease rates, not a repricing driven by dollar debasement. Like the 1980 Hunt Brothers episode, it was a small-scale bubble. It went too far, too fast and will take a long time to consolidate. Gold, which must also run higher, should outperform silver but will be dragged down with the precious metals group until fundamentals reassert themselves. Gold and silver are poor inflation hedges and act more as error hedges.

“There's still more downside in silver because it went way too far, way too fast, and it's gonna take a long time for silver to consolidate. I really think we're gonna see $30.20 dollars per ounce silver simply because it was not a fundamental repricing of silver, it wasn't dollar debasement.”at 50:26

Jeff Snider · Atlas Financial

Copper's record high reflects US tariff-driven inventory hoarding, not AI or a supercycle; it will likely fall.

The short-run factor pushing copper to record highs is tariffs. About 70% of the world's above-ground supply sits in the US doing nothing, which squeezes supply for the rest of the world. Supply is also not growing, since Chilean output keeps disappointing, but the price move is driven by non-economic factors, not dollar debasement, AI data centers, solar or EV demand. Copper therefore has a near- and intermediate-term outlook similar to silver, and a tariff removal could send prices sharply lower.

“the short run factor that has got copper prices up to record highs in this year have to do with tariffs. You've got somewhere around 70% of the world's above ground supply, therefore, accessible inventory sitting in The United States doing nothing”at 51:28

Jeff Snider · Atlas Financial

Dollar's reserve role isn't being replaced; yuan gains come at the euro's and yen's expense.

BIS FX derivatives data show the US dollar on one side of roughly 88 to 90% of FX transactions. The yuan's rising share of FX swaps comes at the expense of the euro and yen, not the dollar. TIC data show the private system buying Treasuries steadily, while foreign governments sell only during dollar shortages. The Eurodollar system is the only reserve currency system that can perform the required functions, so there is no meaningful move away from dollar or Treasury assets. Over the long run, digital ledgers may slowly take over some functions, but the dollar will not collapse.

“On the other side of almost every single FX transaction, it's roughly 88 to 90%, is the US dollar, which means that the Chinese are not replacing the dollar, they're replacing the Europeans to be on the other side of that dollar swap”at 59:51

Jeff Snider · Atlas Financial

Stablecoins are only a midpoint bridge, a half step off the Eurodollar system rather than its replacement.

Stablecoins are a significant step toward decentralized ledger money because they move reserve functions off bank balance sheets. But they are limited in capabilities and capacity. For Snider they are not enough of a step toward a new system. They reinforce the Eurodollar system and may serve as a midpoint bridge to a truly decentralized digital ledger money in the future, which would help adoption down the road.

“they're still limited in their capabilities and capacities, and for me, they're not quite enough of a step toward a new system. It's sort of like a half step off the Eurodollar system as we have it.”at 1:02:46

Andrew Sarna · Forthlane Partners

Commodities are a halo trade: find assets that AI won't disrupt

Sarna says commodities are really the halo trade, pointing to things that aren't going to be disrupted, as opposed to the AI trade. He suggests diversifying into these assets rather than piling into the AI names that drove past returns. He acknowledges these may not be as high-flying as recent AI returns, but says this is not a time to be greedy.

“you bring up the point on commodities where that's really the halo trade. Find things that aren't going to be disrupted.”at 1:11:23

Andrew Sarna · Forthlane Partners

A 7% US deficit is unsustainable, and voters and politicians will eventually be forced to address it

Looking out ten years, the US cannot keep running a 7% deficit. He sees no candidate in the future willing to cut spending, and spending has become politically acceptable. A Wall Street Journal chart showed voter concern about deficits has plummeted over the past decade, which he sees as a problem because the fiscal situation will eventually force people to care. He says the new government reaction function is to spend more, which should make bond investors nervous.

“If I look out ten years, you know what's gonna happen. Like, the US government cannot spend or run a 7% deficit indefinitely.”at 1:12:41

Andrew Sarna · Forthlane Partners

Renewed US escalation with Iran after the midterms is the risk that most worries him

Sarna says the thing that keeps him up at night is that once the midterm overhang passes, the administration could re-escalate with Iran. He says this should keep investors up at night and that it worries him, although he balances it against a bull case of extreme sentiment.

“The one thing that keeps me up at night is without the overhang of midterms, well, the administration reescalate things with Iran, which should keep investors up at night.”at 1:16:40