Hedgeye Podcasts Sep 19, 2026 1h 2m 39m saved
With R. Patrick Kent, Portfolio Manager at Hedgeye Asset Management · Sam Rahman, Portfolio Manager at Hedgeye Asset Management
The spread on triple-C rated corporate debt is 920 basis points over Treasuries. On Kent's arithmetic that is a 14% cost of borrowing for the weakest companies in the market, and those are the companies sitting inside private-equity credit portfolios.
The index says none of this is happening. The S&P 500 is a little over 2% below its all-time high, while roughly 30% to 35% of its members trade above their 50-day moving average and high-yield bonds have turned bearish on Hedgeye's own trend signal.
"So for those following at home, that's a 14% cost to borrow on the crappiest companies and those crappy companies are what your PE credit looks like."
R. Patrick Kent, portfolio manager at Hedgeye Asset Management, ran small- and mid-cap money at Newton Investment Management and built the first diversified public equity impact fund at Wellington Management before joining the firm. Sam Rahman, his co-host on Protect the Pile, was recruited by Ned Johnson to run the Johnson family's multi-billion-dollar public equity portfolio at Crosby Advisors from 2008 to 2023, after fifteen years at Baring Asset Management, where he ended up as head of US equities.
The full episode is covered here so you can skip it. 62 minutes of audio, 23 minutes of reading.
Here are the 14 takeaways that matter.
Key Takeaways
Triple-C credit spreads are at 920 basis points, which Kent works out to a 14% cost of borrowing
The double-B spread is about 273 basis points, up roughly 20 in a month
Those weakest borrowers are what private-equity credit funds own
Only 30% to 35% of S&P 500 members are above their 50-day moving average while the index sits 2% off its high
The consumer sector is down to a handful of names in a bullish trend on Rahman's screen
Physical oil is trading near $120 while the futures curve is near $105, and the two have to meet
Kent expects something in between rather than one side winning
Kent thinks high oil prices flip from an inflationary force to a disinflationary one by 2027, through demand destruction
Data-center capex is being committed at peak prices for steel, copper, diesel and memory at the same time
Software leads three-month momentum, semiconductors still lead 12-month-minus-one-month momentum, and the two definitions are pulling apart
Rahman expects a broader basket of software names and a narrower basket of AI hardware names to survive into 2027
Kent calls the AI safety campaign a PR onslaught covering physical limits on how fast data centers can be built
99% of the productivity gains come from lightweight models on proprietary data, on Kent's own made-up number
Rahman puts 90%-plus of token consumption in basic and semi-basic queries
The AI profit pool will split the way phones did: Apple takes about 80% of gross margin on 20% of the units
A $2,000-a-year NBA courtside package would sell millions of Vision Pros, Kent said, and Apple has not built it
Buyers of credit will soon compare Amazon and Google paper against sovereign debt, and Rahman says the tell is when the companies borrow cheaper
Texas is sorting its power-request backlog into who is real and who is not, and the speculative data-center developers are at the back
1. Breadth Under the Tape
Kent opened the show alone with Rahman, on what the two of them call a free-form Friday, and started with the tape. The S&P 500 was a little over 2% below its all-time high and the surface reading was calm.
The index level hides what is happening underneath
The S&P is around 7620, 7625, wherever I last checked it.
Patrick Kent
Participation is the problem. Kent put the share of names above their 50-day moving average at 30% to 35%, and said the same split shows up between a bearish signal on the Russell 2000 and a bullish one, for now, on the S&P 500.
A third of the market is above its 50-day average
So there's roughly 30 to 35% of names above the 50-day moving average or the moving monkey as we lovingly call it around here.
Patrick Kent
The 10-year Treasury yield held near 5% through a hawkish Federal Reserve meeting on Wednesday. The dollar turned bullish on Hedgeye's signal off the back of the rate rise, which Kent reads as bad for global dollar liquidity — fewer dollars circulating outside the United States to fund trade and borrowing. The rate markets had put the odds of that hike near 90%, he said, and the market still traded as though it were a surprise.
2. The Credit Signal
The clearest deterioration is in corporate credit. High-yield bonds have turned bearish on trend, and the spreads tell a two-speed story: the better-rated end has widened a little, the worst-rated end has widened a lot.
Triple-C spreads are at 920 basis points
And the double B spread to Treasury is at like 273 basis points over something like that. It's up about 20 basis points in the last month. But I think the more interesting story is in triple C credit where the spread is at 920.
Patrick Kent
Kent did the arithmetic out loud and then said who owns that paper.
That is a 14% cost of borrowing, and it is what private-equity credit holds
So for those following at home, that's a 14% cost to borrow on the crappiest companies and those crappy companies are what your PE credit looks like.
Patrick Kent
He also said the energy complex is part of what is happening in credit. Oil is above $100 a barrel, the physical market is closer to $120, and diesel is tight enough that some stations are running out.
The oil price is doing the work
Oil's over 100. Physical markets are closer to 120.
Patrick Kent
3. Rotation and Churn
Rahman said the last two months have been the hardest kind of market to run money in, because the index goes nowhere while everything underneath it moves.
The churn is the story, not the level
I think that's just the nature of what we've been seeing in this market is that there's a lot of rotation a lot of churn and also particularly the last few weeks definitely a narrowing of the market
Sam Rahman
He watches how much the firm's signal-strength list changes week to week, and said the turnover in names coming on and dropping off is unusually high. Falling yields and rising oil prices have knocked out sectors that were consensus long positions — financials and industrials both took hits in recent weeks.
The consumer sector is worse. Rahman ran his own batch screen across consumer names that morning, which he described as a rough proxy rather than the real signal, and said the relative strength chart against the market has been weakening since the spring.
Almost nothing in consumer is still in a bullish trend
It is like there are very few names left that are bullish trend.
Sam Rahman
Positioning is the one thing pointing the other way. The hedge-fund and long-only positioning data the two of them get from the banks is light, particularly in the momentum basket, and there is cash waiting for a reason to move. Neither expects a grand bargain at the United Nations next week; the argument was that less bad headlines would take pressure off crude oil, and that would be enough. Boiled down, they treat this as a one-factor market, and the factor is oil.
4. Oil, Diesel and Refining
The oil market is carrying a gap between what a barrel costs today and what the futures curve says it will cost. Kent put the physical market at about $120 and the front Brent contract at about $105, with that contract rolling within a week or so.
One of the two prices has to move
So either the physical market has to come down to this price or that the price is going to grind higher.
Patrick Kent
He expects a result somewhere in between, and treats the gap as a floor under the oil price rather than a ceiling. He also pointed to a JPMorgan commodities report that went around the market that week, in which the bank effectively withdrew its forecast. Kent said that was the right answer.
Nobody should be confident forecasting this
Why would you think you can sort of like have a firm forecast that you're going to like stick with when you have information that even the people who are involved in all this don't even know how this is going to turn out.
Patrick Kent
On Hedgeye's trend signal, he said WTI's line is somewhere in the mid-$90s, which leaves room for the price to fall a long way and still read as bullish. Rahman's point was that a lower futures price would not fix the physical shortage.
Cheaper futures will not put diesel in the tank
It's going to feel like relief, even though the physical markets are tight.
Sam Rahman
He described pictures circulating on X of gas stations sold out of diesel, and buyers in Asia and Europe unable to source it. Kent, who used to cover energy full-time, added two supply risks. It is hurricane season, and a storm that hits the refining complex is not a remote possibility.
A refinery is running flat out with no maintenance window
And in fact, Exxon just had brought a diesel refinery down just for other reasons because we're running these things full out, right?
Patrick Kent
Refineries get taken down for scheduled turnarounds, he said, and skipping maintenance on what amounts to a giant pressure cooker ends badly. The administration has floated banning exports of refined product. Both of them said that does not solve anything: Rahman's version was that it shaves a few cents off diesel for a moment and then the problem is still there.
5. Demand Destruction Next
Kent's framing is that oil has moved through a first phase and into a second. For months the price was an inflationary force; at these levels it starts destroying demand, and households and companies cut spending on fuel and then on everything else.
High prices turn disinflationary from here
now we are at the point where you really are at demand destruction and potentially if you look forward a disinflationary force given how high these prices are
Patrick Kent
He put the timing on that at 2027, and said the market's expectations for more rate increases next year look erasable.
Rahman agreed and gave the mechanism. Rate increases are landing into what he thinks is the peak rate of change for growth, and data-center construction is the marginal driver of industrial activity.
Capex is being committed at peak input prices
you're not only building the capex but you're building it at peak prices for everything
Sam Rahman
Steel, copper, diesel and memory prices are all up at once, which means an operator underwriting a data center now is paying top dollar for every input. A pause, or even a slower pace of building, is a second-derivative change from where the numbers have been — and that, he said, is what hits the stocks that moved the most. He noted that Hedgeye's Keith McCullough has had industrial momentum on as a short in ETF Pro for that reason.
6. Two Kinds of Momentum
Kent raised momentum as a basket rather than a sector: it is a list of companies, and the list turns over. He asked Rahman how the momentum baskets actually weight their lookback periods.
The classic momentum screen is 12 months minus the most recent one
And they usually take the first month, the one month out.
Sam Rahman
Rahman explained the reason for cutting the most recent month: a stock that just pulled back should not be penalized for it, because that pullback may be the buying opportunity. Quant models vary, and some weight three-month momentum more heavily depending on what else sits in the model. The naive version everyone knows is the best performers over 12 months, minus the last one.
That is where the two definitions have come apart.
Healthcare and software have migrated into short-term momentum
I just find that composition is shifting like healthcare software some of this stuff has migrated into the shorter term momentum 3 month six month where the 12 month is still dominated a lot by the stuff that peaked back in July.
Sam Rahman
Kent's version of the same observation: software is dominating on three-month momentum while semiconductors still sit at the top of the 12-month-minus-one basket, falling down the three-month rankings as they go.
7. Software and Semis
The AI trade began as a story about software being put out of business. Kent said that narrative has shifted — some software companies lose, some have defensible positions, and the ones that survive have been recovering. Layer positioning on top of the momentum split and he sees a setup where both groups can work.
Software positioning has improved from near-extinction levels
So on a longer term positioning sort of map, they still could go quite a bit higher
Patrick Kent
Semiconductor positioning has been cleared out too, which is what makes the group a candidate for anyone chasing beta into year end, absent bad headlines on AI. Rahman framed the question directly.
The next few weeks decide whether both groups can rally together
Can these groups work together or is it going to be either or?
Sam Rahman
Both said yes, probably. Kent's refinement was that the split stops being semis against software and becomes haves and have-nots inside each group. Memory is his example of a losing side: it is a commodity business, Chinese producers are adding capacity, and it does not matter whether US buyers purchase Chinese output because the price is set globally — the same way US mills do not buy Chinese steel and are still priced off it. On that reading the tightest memory pricing is probably behind us.
Intel is his example of the other side, as the company best placed to build out a US semiconductor manufacturing footprint, with a fab already built that needs filling. In software, he sorted names the same way.
Some software names are in better competitive shape than the narrative allows
Similarly in software, Adobe still sucks, but like others like we're learning like Snowflake, other things that are in that software basket actually not only are not hurt, they're actually probably in really good competitive positions
Patrick Kent
He added Twilio to that list, where the story has changed and the valuation is no longer stretched.
Rahman's 2027 call is where the two baskets end up. The comparison problem facing semiconductor earnings next year is so large, particularly in memory, that he does not think those stocks can grow through it whatever multiple they carry — a broader set of software names and a narrower set of AI hardware names than anyone expected in January.
The precedent he reached for was software itself in 2021, when he looked at what investors were paying on an enterprise-value-to-sales basis and ran the reopening math: everyone had been on Zoom, the comparisons were about to become impossible, and the multiples were at a peak.
Peak multiple on peak numbers is the worst setup there is
I know we don't care about valuation necessarily with these names but if the second derivative shifts on you like this is going to be bad.
Sam Rahman
Those setups feel worst when you are not involved at the end, he said, and then unwind badly because the multiple contracts at the same time estimates are cut.
8. The AI Safety Push
Kent's read on the coordinated AI safety messaging of the previous week was not about safety.
It is a public-relations campaign
It's kind of a PR onslaught.
Patrick Kent
He said it is solving for three things at once. The first is a salvation story: the technology could kill us all, so regulate a trusted few and lock everyone else out. The second is physical. He had checked with a distributor selling into data centers and utility capex, who said demand is enormous but the pace cannot keep accelerating, because permitting will not move that fast.
Construction does not run at 30% growth rates
And anyone who's ever tried to build anything, it's like you can try to push these things at 20 30% growth rates, but like you still got to get the inspector to come out and sign off on the thing and it needs to go through the town meeting in this area.
Patrick Kent
Better, he said, to tell a story about slowing down for safety reasons than to admit you have run into the limits of what can be built. The third is liability: he cited that morning's story about Anthropic saying its models were used to break into OpenAI email accounts as the beginning of an open-ended exposure.
The bigger AI risk is what people do with it
we probably cause ourselves more problems with AI long before AI gives us problems
Patrick Kent
Rahman agreed and named the strategy: regulatory capture, getting the government to fence off the incumbents as open-source and other models arrive. What struck him was the speed of the political response, with Dario Amodei, Sam Altman, the New York Times, the evaluation body METR and effective-altruism-linked figures all landing inside about three days.
The political class arrived fast
how quickly the political classes came running to support this narrative of hey we need to regulate this
Sam Rahman
Kent's criticism was of the pacing rather than the substance: the window was tight enough to look coordinated. Rahman's answer to the regulation argument was that the machinery already exists — the FTC and a century of bodies overseeing product liability and product oversight.
There is no need to invent a new regulator
So it's like not like we need to really reinvent the wheel to do this.
Sam Rahman
9. An AI Claim Skeptic
Kent set out a position he wrote up about three weeks ago.
He does not doubt the technology, he doubts the claims
I'm not an AI skeptic, I'm an AI claim skeptic, right?
Patrick Kent
His argument is that the claims are sized to the sellers' ambitions rather than to what buyers need, and that a regulated closed-weight artificial general intelligence is a solution looking for a problem.
Nearly all of the productivity gain comes from small models on your own data
I'm picking a number. I'm totally making this up, but if I just think real world, like 99% of the productivity gains you will actually get are from like a lightweight model running on proprietary data that you can like customize solutions for your own business or your own enterprise or your own industry like law or whatever
Patrick Kent
The example he gave was a law firm cutting three paralegals with a tool trained on that firm's own history of work, which does not need to be replaced every two months. Rahman agreed and gave the usage split.
Basic queries are most of the tokens
I don't know what the number is but it is a large number like 90 plus% of token consumption goes for really basic semi-basic queries
Sam Rahman
The remaining 10%, running frontier models on heavy research work, is where the margin is. Kent's analogy for how that profit pool divides was smartphones.
Apple takes most of the profit on a minority of the units
It's like Apple's like 80% of the gross margin but 20% of the units.
Patrick Kent
Rahman put the profit split closer to 60% to 70% Apple against a similar share of units for Android, and said AI applications will end up skewed the same way, if less extremely.
10. The Missing App Layer
What has been visible so far is spending, cost, price inflation and what Rahman called the dodgy financing deals. What has not arrived is a consumer application anyone can point to.
Enterprise coding is what is actually driving usage
So that's kind of what's been driving token consumption has been coding using AI.
Sam Rahman
He named Codex and Claude Code as the moments the enterprise case became obvious, and said he has been waiting for the consumer equivalent. Kent has started using Meta AI.
The first consumer application that looks genuinely usable
This is like the first real look at a consumer application that has the potential to be super useful, high utility, really, really easy to use, like no necessary coding or extra steps required.
Patrick Kent
That led to a disagreement about how much of Meta's consumer reach counts. Threads is a distant second to X, Rahman said, after launching into a moment when everyone was sour on X and ramping quickly; Instagram still feeds it traffic, and he and his wife both end up in conversations there by tapping through. His point was about distribution rather than the product.
Nobody puts a consumer product in front of more people faster
They're putting it in front of consumers very quickly.
Sam Rahman
Kent put Apple in the same category and Google close behind, and said the model layer will stop mattering to most users — the fight that decides returns is at the application layer. The dot-com comparison he drew is that the spending can still end badly while the behavior it creates persists: Amazon, eBay and Priceline were real utilities built on a bubble. Rahman's version is that the crash seeded consumer and enterprise habits for decades afterward.
Kent's suspicion is that the closed-weight, one-platform-to-rule-them-all model is the thing that gets disproved. Everyone in 1999 expected the internet to decentralize power, and 25 years later it concentrated into the Mag 7; he wonders whether the OpenAI-versus-Anthropic race to be the single platform ends up being the fragmentation people expected the first time.
11. Vision Pro's Use Case
Kent demoed an Apple Vision Pro a couple of months ago and has not stopped thinking about what Apple has not done with it.
A courtside subscription would move millions of units
If you launched the Vision Pro with a NBA courtside package for pick a number, $2,000 a year subscription, it would sell millions.
Patrick Kent
He listed the rest: concerts with Live Nation, IMAX films at home, the best seats at a baseball game, on the glass at an NHL game — sold anywhere in the world. Rahman, who had been at the US Open a few weeks earlier and does not otherwise like virtual reality, made the same point about a final he could not get seats to. He said the ecosystem has to be built alongside the device, and that Steve Jobs understood the order of operations.
A device without a use case is pointless
there's no point introducing a device that doesn't have a use case
Sam Rahman, relaying Steve Jobs
Kent's precedent was the iPod, which was not the first MP3 player and did not need to be.
The store was the product
The bigger deal was the iTunes store.
Patrick Kent
That is the gap in the Vision Pro
So like the Vision Pro hasn't sold that many because there's just not that much to do with it, man.
Patrick Kent
He said the hardware also needs work. It is bulky, and storage, memory and power all need improving. If Apple misses this, he said, Jobs would not have.
12. Deficits and Gold
Kent flagged the topic he wanted for another week: whether deficits matter. He raised the 6% federal deficit from his own Early Look note that morning, and the point that it exists without a recession or any external shock to explain it.
The interest bill is the part that is hard to look at
this really is kind of eyebrow raising when you sort of think about that growing pie of interest payments as part of the budget
Patrick Kent
That arithmetic, he said, is what gold bulls have been pointing at for decades: without a policy change there is no way out of the math. On the efficiency drive, he noted that hiring a staffer in his twenties known by the nickname big balls to cut the deficit did not get the job done.
Rahman's read on the next two years is that spending stalls. If the Democrats take the House, the purse closes and little gets through apart from emergencies such as defense.
13. Who Borrows Cheaper
The second half of that thought is where the credit to fund AI comes from. Rahman said the capital requirements are large enough, and increasingly global, that buyers of credit will find themselves comparing a developed-market government against a hyperscaler.
The comparison will be Amazon or Google against a sovereign
Well, that'll be the indicator when they can borrow at better rates.
Sam Rahman
Kent's reference point was a line from the trailer for Aaron Sorkin's The Social Reckoning, which has a character saying they are post-government.
The dystopian version is corporations as nation states
if you want to believe in like one of those dystopian futures of like the technology companies, these sort of corporate technology companies are the new sort of nation states.
Patrick Kent
He said he would like to see where Johnson & Johnson's credit trades against the US government right now, and that spreads for the highest-rated corporate borrowers have been tight.
Triple-A corporate spreads are already close
those spreads have been relatively tight
Patrick Kent
Rahman expects a deadline effect. The 2028 presidential election could change the equation for every large spender, so operators will pull spending forward while they can, subject to the regulatory, local, power and labor constraints that cap how fast anything gets built.
Everything gets pulled in ahead of 2028
they're going to try to frontload as much as they can before the end of 28 because whatever happens after that, nobody knows.
Sam Rahman
14. The Power Queue
The last stretch was about who ends up owning the data centers. Kent's expectation is that the large buyers end up with capacity built speculatively by weaker credits — the neoclouds financing construction with borrowed money.
Telecom did this once already
It's like you saw these independent ones that came out and then they went through bankruptcy or close to it or ended up on the pink sheets and then got bought out.
Sam Rahman
AT&T and Verizon bought those assets out of bankruptcy, he said. Texas is now sorting the same problem at the front end. Faced with power requests running to multiple gigawatts against what the state can actually generate, it is grouping the backlog by who is legitimate, meaning real financing, local approvals mostly done and backup power arranged, and who is not.
The back of the line is full of developers who will never build
And I think what you're seeing is that there is a very long tale that's in the back of the line that are mom and pop cowboy developers that are never going to see the light of day
Sam Rahman
He cited a Bernstein chart he had sent Kent showing how much of the incremental announcement flow came from speculative builders. His expectation is that a buyer's market develops for permits and half-finished projects, because a large balance sheet can afford to sit on a project it did not borrow to start. The same sorting is already happening with suppliers and contractors.
Scarce contractor capacity goes to the names with real money
they're going to prioritize the guys that are coming in with real capital, real name brand, and real fin real everything are going to get preferential treatment in terms of contracts and supply
Sam Rahman
Amazon, Blackstone, Brookfield and KKR are the buyers he named for that supply.
Bonus Insights
The next episode is a politics one. Kent said they are trying to book JT Taylor, Hedgeye's Washington policy analyst, before the midterm elections, and possibly more than once.
Rahman has not committed to Meta AI. He has used it for a few days and is weighing the access it wants.
I'm still kind of like deciding whether do I really want meta to have access to everything
Sam Rahman
Kent's own relationship with the technology has limits. He said he is not using Meta AI to that extent yet, and described his worse days.
a little bit close to just unplugging everything and going into the woods sometimes
Patrick Kent
The format is deliberate. Free-form Friday means no guest and no agenda. Kent said he always wonders whether there will be enough to talk about, and they ran the hour with another half hour left over.
The bottom line from Rahman: next year sorts both halves of the AI trade rather than picking one, with a broader set of software names surviving than anyone expected in January and a narrower set of semiconductor and AI hardware names getting through the comparisons.
Products, Companies & Tools Mentioned
Hedgeye Asset Management (Where both speakers manage money, and the publisher of Protect the Pile; its trend signals frame every call in the episode)
S&P 500 and the Russell 2000 (Bullish for now and bearish respectively on Hedgeye's signal, which is the divergence Kent used to open)
JPMorgan (Its commodities team withdrew its oil forecast in a report that circulated that week, which Kent called the right answer in a system this complex)
Exxon Mobil (Took a diesel refinery offline, which Kent used to make the point that refineries are being run without maintenance windows)
Morgan Stanley (One of the banks whose hedge-fund and long-only positioning data they watch; it reads light, especially in momentum)
Intel (Kent's pick as the company best placed to build a US semiconductor manufacturing footprint, with a fab already built that needs filling)
Adobe, Snowflake and Twilio (Kent's sort of the software basket: Adobe on the losing side, Snowflake and Twilio in better competitive shape than the narrative allows)
OpenAI and Anthropic (The two closed-weight platforms racing to be the one platform, and the subjects of the week's safety campaign and the email-hacking story)
Codex and Claude Code (The coding tools Rahman says made the enterprise case obvious and now drive most token consumption)
Meta AI and Threads (The first consumer AI application Kent rates, attached to the distribution machine that made Threads ramp and keeps it fed from Instagram)
Apple and Apple Vision Pro (The device Kent says is missing its iTunes moment; he wants an NBA courtside subscription attached to it)
NBA, NHL, Live Nation, IMAX and the US Open (The content partners they think would make a headset worth buying, priced as a subscription)
Amazon, eBay and Priceline (Kent's examples of real utility built during the dot-com period that outlived the crash)
Amazon, Blackstone, Brookfield and KKR (The balance sheets Rahman expects to take over permits, projects and contractor capacity from speculative data-center developers)
Johnson & Johnson (Kent's test case for how close the best corporate credit now trades to the US government)
AT&T and Verizon (Bought the independent telecom builders out of bankruptcy after the last infrastructure cycle, which is Rahman's template for the neoclouds)
DeepSeek (Kent's reference point for how an AI headline shock lands, describing the week as a DeepSeek moment part three)
Books & Resources Mentioned
ETF Pro (Hedgeye's ETF product, where Rahman noted Keith McCullough has industrial momentum on as a short)
JPMorgan's commodities note (The report withdrawing the bank's oil forecast, which Kent said went around the world that week)
Bernstein data-center research (The chart Rahman sent Kent showing how much of the incremental announcement flow came from speculative builders)
Kent's Early Look (His morning note for Hedgeye, where he raised the 6% deficit and the growing interest bill)
METR (The AI evaluation body Rahman named among the groups that surfaced during the safety push)
The Social Reckoning (Aaron Sorkin's companion film to The Social Network; Kent quoted a line from the trailer about being post-government)
Listen to the full episode
🔴 YouTube | 🔗 Episode page
Watch the full episode:
If this was worth your time, send it to someone closer to the industry than you are.
Get the latest market chatter as it happens: