Subscribe now

West Texas Intermediate ran from $57 in January to almost $115 by April. Then it rolled back over, and it sits near $85 today. The market exhaled. Crisis averted, energy normalizing, nothing to see here.

Except the pump never normalized. Gasoline is still $4.05 a gallon, a few cents off its May high, while crude has given back a quarter of its peak. The barrel and the pump have come unstuck, and that gap is the whole story. We think the market is grading the wrong number.

Read more

For fifteen years every selloff was bought on the same faith, that a chair stood behind the tape ready to rescue it. That chair takes his seat Wednesday, and he spent his career arguing against the rescue. The wrinkle is that disinflation is finally showing up, the one thing that usually buys the cut, and we do not think it pays this time.

Subscribe now

Markets have a reflex, and like all reflexes it was trained. For fifteen years, through every wobble and every crash, the lesson was the same. When things got ugly enough the Federal Reserve would cut, or flood, or at minimum signal that help was on the way, and the long end would come back down to rescue whatever had broken. Buyers learned to step in ahead of it. The reflex paid often enough that the market stopped asking whether the rescue was actually coming and started treating it as a law of nature.

You can date the lessons. The pivot in early 2019, when the tape broke on too much tightening and the Fed reversed within weeks. The flood in 2020, when the balance sheet nearly doubled in a matter of months. The quiet backstop in the spring of 2023, when a handful of regional banks wobbled and a lending facility appeared almost overnight. Each time the pattern held, and each time it taught the same lesson, that the Fed’s tolerance for market pain was finite and the put was real. A whole generation of investors has never once been positioned against it.

This week the reflex is firing again, quietly, in the way investors lean on a dovish pivot the moment a print softens. And it runs straight into a problem the tape has not priced. The man who breaks the tie at the June 16 and 17 meeting is Kevin Warsh, and the rescue is the exact thing he has spent his career arguing the Fed should stop providing.

So here is the claim, stated plainly so you can hold us to it. The market is pricing a Fed put that is being retired in real time, and the disinflation now arriving will not buy it back. We will show our work on real data, the kind anyone can pull and check. If credit starts pricing stress, if the long end rolls over, and if the new chair blinks dovish on Wednesday, we are wrong and we will say so. None of those is happening yet.

The Reflex Is Firing Into a Calm Tape

Start by being honest about what this is and what it is not. This is not a crash call. It is not even a selloff call anymore. The scare that rattled the headlines a couple of weeks ago has already been bought. The S&P closed today around 7,554, back near its highs. Small caps led the recovery, the Russell up better than four percent on the week. More than sixty percent of the index sits above its 200-day average, the broad tape is healthy, and volatility has bled back down to a 16 handle after touching 21 when the worry was fresh. The dip got bought, exactly the way fifteen years of training said it would.

That texture is the tell, because it tells you what the reflex just did and what it now assumes. It worked. Weakness showed up, buyers stepped in ahead of the rescue, and the tape is back at the highs wearing a 16 vol. The bet underneath a thousand screens is no longer to buy this dip. It is the quieter, more dangerous one. That the next dip gets bought too, because the backstop is still there.

We think that bet is leaning on a backstop that is being dismantled. The tape is calm enough that the interesting question has nothing to do with whether something is breaking. The question is whether the thing everyone is counting on to catch the next break still exists.

The Referee Takes His Seat

Kevin Warsh was sworn in on May 22. There has been no policy meeting since Jerome Powell’s last on April 29, so this week is unambiguously his first in the chair. The decision itself is a formality. A hold at 3.50 to 3.75 percent is priced at something like ninety-nine percent, and we can move past it. The meeting matters for everything around the decision. The projections, the dot plot or whether he keeps one at all, the language on whether an easing bias survives, the tone on the balance sheet, the cadence he sets at the podium. Those are the tells, and every one of them points the same way.

He inherits a divided committee. The April meeting produced an 8-to-4 vote, the most dissents at a single FOMC since 1992, with some members wanting a cut and others objecting to an easing bias sitting in the statement at all. Warsh inherits that split, and on the public record he lands with the hawks. He has argued for years for a smaller balance sheet, for a Fed that holds less duration, for sound money, and against the reflexive accommodation that defined the prior regime. We want to be careful and call that what it is, his stated views and the market’s reasonable reading of them, which is a different thing from announced policy or a meeting outcome.

Watch the projections in particular. The prior dot plot still carried a cut or two penciled in for this year. If the median loses them, or if the committee sketches a hike, that is the market’s signal to flip from pricing easing to pricing its opposite, and it would land on a tape still leaning the old way. Watch whether the easing-bias language survives at all. And watch whether Warsh treats the dot plot as an institution worth keeping or signals he would rather scrap it, because a chair who distrusts forward guidance is a chair telling you not to count on a telegraphed rescue.

Look at what he is taking over. The Fed’s balance sheet ballooned in every crisis of the last two decades, each expansion a rescue, and even after two years of slow shrinking it still dwarfs anything seen before 2008. A chair who has argued for years to shrink it further wants to retract the most visible arm of the put. The reflex assumes the balance sheet rides in again the moment things break. He has spent his career arguing it should not, and on Wednesday he sets the agenda.

The Engine He Wants More Of

The reason the long end refuses to come down has nothing to do with growth or next month’s inflation print. It is supply, plain and simple. The government is funding enormous deficits, the market is being asked to hold a mountain of long-dated paper, and it is demanding to be paid for the privilege. That premium, the extra yield investors require to hold a long bond rather than roll short ones, has gone from deeply negative to clearly positive over the last two years and keeps grinding higher.

That is the engine pulling the 30-year toward five percent. Read it against what Warsh wants and the implication is uncomfortable. The force already lifting the long end is the force the new chair would rather lean into than fight. A Fed that wants less duration on its balance sheet and a steeper curve is a Fed comfortable with a higher long end. The rescue the reflex is pricing runs directly against the chair’s own stated preference.

And this is not a passing technical that fades when sentiment improves. The deficits are structural, the issuance calendar is relentless, and there is no obvious marginal buyer stepping in to fund it cheaply now that the Fed is shrinking its own holdings rather than adding to them. The premium is simply the market’s price for that imbalance, and it does not melt away on one soft inflation print.

The scale is the point. Federal debt held by the public sits near a post-war record as a share of the economy, and the deficits that feed it are structural rather than cyclical. Someone has to absorb every new auction, and the price of absorbing it is the premium you read off the long end.

For years that someone was foreign. Overseas official buyers once funded close to a third of the debt, and that share has fallen for a decade even as issuance exploded. The marginal buyer now is domestic and price-sensitive, and a price-sensitive buyer demands to be paid.

And the cost of carrying the pile compounds. Federal interest expense now runs alongside the entire national defense budget, and every basis point the long end rises makes next year’s deficit worse, which means more issuance, which means more premium. That loop is the engine, and it does not care what core inflation printed last month.

Now decompose the move, because the inflation read is the one most likely to be gotten backwards here. Split the 10-year into its two pieces, the real yield and the inflation compensation buried inside it, and the story is clear. The real yield sits near the high end of its fifteen-year range while the market’s inflation expectation has drifted lower, down toward 2.3 percent. Real is doing the work. The bond market is demanding more yield because the real cost of money is rising and because someone has to be paid to absorb the supply. The long end is high because of who has to fund the government and at what price, and a soft inflation print does nothing to change that. So the usual rescue logic, that cooler inflation lets the Fed ease the long end back down, misfires at the source.

Disinflation Arrives, and the Rescue Still Doesn’t

Here is the part that should, on paper, hand the dip-buyers their catalyst. The disinflation they have been waiting for is actually showing up.

May CPI was a hot headline on a soft core. The headline ran 4.2 percent over the year, dragged up by energy at 6.7 percent on the month and gasoline at 8.6 percent, the last echo of an oil scare that has run since February. But the core, the part the Fed actually steers by, came in soft, up two tenths on the month and 2.8 percent over the year, with shelter cooling. Strip the energy and the disinflation is doing exactly what a patient committee would want to see.

The gap between that hot headline and that soft core is energy, and energy just broke.

Crude round-tripped the entire war premium. Over the past week front-month WTI collapsed about fourteen percent to roughly 80 dollars, an eight-week low, as hopes built around a US-Iran interim deal that would lift oil sanctions and reopen the Strait of Hormuz. The dollar is soft alongside it. Both work to pull the headline down toward the core, which is the direction that takes pressure off prices.

So put it together. A soft core, a collapsing energy headline, a soft dollar, inflation expectations drifting lower. A normal Fed eases into this. That is the trap in the bull case, and it is a good trap, because the disinflation is genuine. But walk it through. The thing keeping the long end elevated was never the front-end inflation story. It is the supply premium, the cost of duration in a world of relentless issuance. So the very disinflation that should green-light a rescue does almost nothing for the 30-year, and a chair who wants more term premium has no reason to manufacture one. The excuse arrives. The rescue does not follow it. That is the whole piece in three sentences.

Credit Isn’t Pricing a Rescue Either

If something were actually breaking, the part of the market whose entire job is to price default risk would be the first to say so. It is silent, and the silence cuts against the reflex from the other side.

High-yield spreads sit near 2.8 percent, investment grade near 0.75 percent, both close to the tightest levels of the cycle and well under the 300 basis point line where credit even begins to register concern. Equities took a real scare a couple of weeks ago and then ripped all the way back, and through both moves these spreads barely twitched. Pull the chart back far enough and the discomfort is obvious. This is the same flavor of calm credit wore in 2007, camped out at the tights deep into the year, looking fine right up until it was not, and then gapping rather than drifting when it finally moved.

There is a reason the tightest part of the credit market makes the least reliable warning. Spreads compress when capital is plentiful and reaching for yield, which is precisely the late-cycle condition that tends to precede the turn. The blow-outs in 2008 and 2020 did not build slowly off wide levels. They detonated off tight ones. The equity reflex treats spreads this tight as an all-clear. We read them as the absence of a cushion, the market pricing perfection into a regime that is busy removing the backstop perfection was leaning on. We are not calling a credit blow-up here, and we want to be precise about that. We are making the narrower point that spreads this compressed leave no room for error if the rescue everyone is pricing turns out not to come.

What It Means If You Borrow

If you do not trade a single share, this section is for you, and it carries the same weight as everything above it. To an operator, the rescue the market is pricing is concrete. It is the cut you were counting on, the refinance you were waiting to get cheaper, the rate you penciled into next year’s plan.

Watch how the cost of long money actually reaches you. The 30-year fixed mortgage sits at 6.52 percent because it keys off the 10-year Treasury yield, and the 10-year is the very thing a steeper-curve chair wants higher rather than lower. There is no separate dial the Fed turns to bring your borrowing cost down while it holds the long end up. They are the same dial. A chair comfortable with a high long end is a chair comfortable with your mortgage, your maturity wall, and your line of credit staying expensive.

This is the split we keep coming back to. There is the economy that owns the assets, that watched the index hold near its highs and felt fine, and there is the economy that borrows to operate, that lives on the price of credit and feels every basis point of it. The first economy is debating whether to buy the dip. The second is deciding whether to break ground, refinance a building, or roll a line of credit into a long end that refuses to come down. A retired Fed put is a footnote to the first economy and a regime change to the second.

The places it bites first are predictable. Commercial real estate has a wall of maturities coming due that was underwritten when money was cheap and now has to refinance into rates that are not. Small businesses run on floating-rate credit that prices off a short rate the chair intends to hold firm. Private portfolios carry floating debt that a rescue cut was supposed to relieve. None of those borrowers got the memo that the backstop is being pulled, and Wednesday is where the memo gets sent.

And the cushion most operators assume is there, a Fed that cuts the moment the economy softens, is the very thing in question. Policy is still restrictive against an inflation rate that is cooling.

A normal Fed, looking at a core running under three percent and falling, would already be easing. This one is signaling it would rather wait, and that tells you something about the regime. The bar to a rescue is higher than it has been in a generation.

The labor market is the tell to watch, and it is cooling under a calm headline. Quits are the truth serum here. Workers stop volunteering to leave well before the unemployment rate moves, and quits have rolled over while unemployment stays low. By the time the jobs number cracks, the decision is already behind you. So read Wednesday’s projections the way a borrower should, as the clearest twelve-month signal you will get on the price of money. If the dots drift toward no cuts, or toward a hike, plan as if the short end stays high and the long end does not come to you. Finance against the rate you can actually get today, not the one you were hoping a rescue would deliver.

What Would Change Our Mind

We hold a view, so we will tell you plainly what breaks it, because a thesis without an off-switch is just a position you have married.

We are wrong, and the rescue is closer than we think, if three things turn. Credit starts pricing risk, with high-yield spreads widening off these tights instead of sleeping through the tape. The long end rolls over, with the real yield coming down and the supply premium releasing rather than building. And the chair himself blinks, with a dot plot that pencils in more cuts than the last one, or an easing bias kept and reinforced, or a press conference that leans dovish. Any genuine move on that list and we re-risk without apology. Right now not one of them is in motion.

Short of that, the setup is what it is. The reflex to buy weakness is leaning on a backstop that the man taking the chair on Wednesday has spent his career arguing against, and the disinflation that would normally force his hand lands on the front of the curve while the long end stays a fiscal story he is content to leave alone. For investors, that makes the dovish pivot the low-probability bet rather than the base case, and it makes the dot plot and the bias language the highest-leverage things to watch this week, well above the rate decision itself. For everyone whose plans run on the cost of money rather than the level of the index, it means the same thing in plainer terms. The cut you were waiting on is no longer the base case.

The reflex was trained over fifteen years. The trainer just left the building. We find out on Wednesday whether the market has noticed.

This is the board as we read it heading into the June inflation print and Kevin Warsh’s first meeting in the chair. The growth data is firmer than the mood, with orders and borrowing both up. Yet breadth healed on the surface while its momentum rolled over, defense bounced, and credit is still asleep at the tights while the labor data cools one print at a time. The deficit does not close, so the long end carries the weight. And for the first time in a long while, cash pays you to wait.

Forty-eight charts, twelve pillars, and the live book at the end. None of it is a forecast. It is a map of where the weight sits right now, and what would have to change for the call to change.

A word on cadence before the board. The May book never shipped, the calendar got away from us, and we would rather name that than pretend the gap was by design. This one puts us back on the rhythm.

Subscribe now

Read more

The yield back-up is real and it is fiscal, not a price scare, and that one fact sorts the whole tape. The leaders flashed for weeks. The laggards are catching down on schedule. Two June catalysts, a CPI print and Kevin Warsh’s first meeting in the chair, both cut the same way. The book is built for exactly that.

Subscribe now

The Setup: One Fact Sorts Everything

A month ago the question was whether the leaders were right. They were. Breadth thinned while the index made new highs, the term premium ground higher, and the real 10-year backed up toward the high end of its decade-and-a-half range. We told you the tape was turning in order, leaders first, and that the followers had not gotten the message yet.

This is the follow-up, and the message is starting to land. Friday’s payroll cracked the calm. The index wobbled off its highs, the VIX popped to a 20-handle, and megacap tech rolled a second leg lower after a one-day bounce. Then the very next session it steadied, which is exactly how a tape that is rolling rather than breaking behaves. Step back and the picture is clear enough. The S&P is still up high-single-digits on the year and sitting a few percent off its all-time high, carried by a shrinking handful of names, while the leading indicators underneath it have been deteriorating for weeks. The leaders are still leading. The laggards are still lagging. The gap between them is the trade, and over the next month we think it resolves toward the leaders.

The whole outlook rests on one fact, and here it is plainly. The back-up in yields is real and it is fiscal. Not an inflation print. The real 10-year sits near 2.2%, the high end of its fifteen-year range and still shy of the 2023 peak around 2.5%, while breakevens have behaved and parked near 2.35%. This is a market repricing the price of money for a fiscal reason, term premium and fiscal dominance, the long end demanding to be paid more to hold duration in a world of relentless supply. The 30-year is pushing 5%, and it printed just over it last week.

That distinction is the entire piece. A real-rate regime is a sorting machine, and it sorts without mercy, paying zero-duration cash a real return while it punishes anything long-duration, anything that trades like a bond, anything whose multiple was built on a discount rate that no longer exists. It crushed gold. It carries the long end. And it quietly compresses the multiple on the growth complex that led for two years. Get the cause of the yield move right and everything downstream falls into place. Get it wrong, call it an inflation scare, and you fight the wrong battle into both prints.

The Engine: Term Premium, and a Chair Who Wants More of It

Start with the thing actually moving the long end, because it is also the thing the new Fed chair wants to amplify.

Our Fiscal Pressure read, the 10-year term premium, is back above the 95th percentile of its trailing five-year range. That is the engine pulling the 30-year toward 5%. It is not a forecast of inflation and it is not a growth scare. The bond market is charging rent on duration in a regime of structural deficits and heavy issuance, and that read tells us, as well as anything we track, why the long end refuses to rally even as the front end prices a more restrictive Fed.

Here is what changes the calculus for the next twelve months. The chair is now Kevin Warsh, and June 16-17 is his debut. His agenda is not subtle and it is not new, he has been writing it for years. Sound money. Active balance-sheet shrinkage. The deliberate removal of the easing bias that the prior regime baked into every statement. And, most consequential for the long end, a steeper curve engineered through balance-sheet composition, shifting the Fed’s holdings to push term premium higher rather than suppress it.

Read that against Figure 2 and the implication is stark. The force already driving the 30-year toward 5% is the force the new chair wants to lean into, not against. A Fed that wants more term premium is a Fed that is comfortable with a higher long end, comfortable with a steeper curve, comfortable letting duration reprice. That is a regime that rewards short duration and rate-insensitive defensives and punishes the long-duration growth trade. It runs against the reflex the last fifteen years trained into everyone, the reflex that says the Fed will eventually come to the rescue of the long end. This one might not want to.

The Gap: Laggards Catching Down, Credit Still Asleep

Markets turn in order. They always have. The leading indicators move, the lagging confirmations follow, and the gap between them is where the information lives. Right now the gap is wide and closing.

What the leaders flashed weeks ago has been the same trio the whole way down: breadth thinned, the term premium climbed, and real rates backed up while the index kept printing highs on fewer and fewer names. This week the followers started to catch down. The VIX is on a 20-handle. The spread between QQQ and SPX one-month implied vol hit a four-year high, the options market telling you exactly where it expects the next leg of pain, in the megacap-tech complex that carried the index. Net new 52-week highs sit near 3% on a ten-day average, breadth still narrow even on the bounces. Megacap tech rolled a second leg lower this week after a one-day relief pop. The tape is doing the thing the leaders said it would do.

And then there is the one confirmation that has not arrived, the tell at the heart of the whole setup. Public credit spreads are still at the tights. High-yield option-adjusted spread sits near 2.75%, investment grade near 0.75%, both in the record-tight decile. The bond market that is supposed to sniff out stress first is fast asleep.

We do not read that as a refutation. We read it as the lag itself. Cash credit is the last domino, the slowest of the lagging confirmations, and the fact that it has not moved is exactly consistent with a tape that is turning in order with credit at the back of the line. The complacency is loudest in the place that should be most alert.

We want to be precise about what Figure 3 says and does not say. Labor is not stressed. The labor read is neutral, sitting right around its average. The point is that credit is priced tighter than even a neutral labor backdrop warrants. Spreads are pricing perfection, into a regime that is removing the easing bias that perfection was leaning on.

Now look one layer down, because the camouflage matters. Public credit is asleep, but the public number is dominated by a handful of insulated megacap issuers who can fund themselves in any weather. The cracking is happening on the private side, where the marks are slower and the exits are gated. Direct-lending stress, by the data we track, is running above its multi-year norm. Business-development-company redemption gates have reportedly come down as quarterly redemption requests pressed against the caps, which is the polite institutional way of saying the line for the exit got too long. And a rising share of leveraged-loan defaults appears to be running through liability-management exercises, the amend-and-extend machinery that keeps a default from printing as a default and keeps the headline rate looking benign. We flag these as directional reads off the private-credit tape, not pinned, audited figures, because that corner of the market does not mark cleanly or report on time.

That last point is the whole game. The public spread is camouflage. The stress is real, it is just being held in the slowest-marking, least-liquid corner of the credit complex, and the public side mean-reverts to the private, not the other way around. When the gap closes it closes downward, toward where private credit is already telling you the risk sits. Up in quality, short in duration, and no reach for credit here.

The index near its highs on 57% participation is a narrow tape. A few names are doing the work, and the names doing the work are precisely the ones a higher discount rate punishes. That brings us to the internals.

Read more

The leading reads were flashing for weeks while credit and volatility slept. Friday was the tape catching up to an inside that had been coming apart since February.

Subscribe now

Markets do not turn all at once. They turn in order.

The leading parts go first, quietly, in the places that never make the front page, and the lagging parts go last, loudly, on the day everybody is watching. A framework earns its keep in the gap between those two events. The print broke the tape on Friday, but the story had already broken weeks earlier, and the point is the sequence: the leading reads were flashing for weeks while the lagging reads slept, and what looked like a single ugly session was the moment the surface finally caught up to an inside that had been coming apart for months.

The Sequence Is the Point

Here is the claim, stated plainly so you can hold us to it. The reads that move early were already moving. Breadth internals had been hollowing out since February. The term premium was sitting at the top of its own range, dragging the long end higher for a reason that has nothing to do with growth. Real rates had pushed to fifteen-year highs. All of that was true before Friday. The reads that move late, credit and volatility, were still asleep. So when the May payroll print landed hot and the tape finally cracked, it was the surface catching up to a structure that had been failing for weeks.

That is the whole thesis. What leads is the internals, the term premium, the real cost of money. What lags is price, and price is what most people call “the market.”

The index made new highs into early June. Underneath it, fewer and fewer names were participating. By the time the headline made its high, only a slim majority of the five hundred were even above their own long-term trend. The surface said all-time high. The inside said fewer soldiers carrying the flag every week. That is distribution, and it had been running for four months in plain sight.

What Actually Happened on Friday

The proximate cause was the jobs report. May payrolls came in at +172,000 against a consensus near 80,000. On the headline, that is hot, more than double what the street penciled in. The reflex read was simple: strong labor, no cuts, sell duration. And the tape obliged.

The S&P 500 had its worst day since October, closing around 7,384. The Nasdaq fell more than four percent. A nine-week winning streak ended. It was the first down week in ten. Those are real numbers and they hurt, and we are not going to dress them up.

But pay attention to what fell together, because this is where the easy story breaks. On June 5 the Dow dropped 1.35 percent and the Russell 2000 dropped 3.5 percent. Both. At the same time. There was no clean rotation out of the megacaps and into the forgotten old-economy names, no healthy passing of the baton from growth to value. Small caps and cyclicals went down with the giants. When everything sells at once, that is money leaving risk, not a baton pass.

The semis took the worst of it over several sessions. Broadcom fell about 3.8 percent on June 5 after failing to raise its AI outlook, and the larger double-digit damage you may have seen quoted was the cumulative multi-session slide across Micron, AMD, and Intel, not one name on one day. The difference matters. A cluster of AI-adjacent names repricing over a week is a story about crowded positioning unwinding, not a single earnings miss.

And the labor headline itself deserves a second look. The headline was hot, yes. Underneath it, the quits rate is rolling toward the 2.0 line, the level that historically marks workers losing the confidence to walk, and initial claims are grinding higher. That is late-cycle flow softening hiding under a strong stock number. The headline is the lagging read of the labor market. The flows are the leading read. Same pattern, one layer down.

The Real-Rate Back-Up

This is the part the consensus is most likely to get backwards, so we are going to spend real time on it.

When yields back up, the lazy interpretation is inflation. Prices are running, the Fed will have to stay tight, term yields rise to compensate. Sometimes that is right. This time it is not, and the bond market itself tells you why.

The front end led. Over the last twenty trading days the 2-year climbed twenty-two basis points while the 30-year barely moved, up three. When the short end leads a sell-off, the market is repricing the policy path, not a long-run inflation scare. It is saying the Fed will be higher for longer, or higher again, before it is saying inflation has gotten loose.

Now decompose the long end into its two pieces, the real yield and the inflation compensation buried inside it.

The real yield is at +1.1 standard deviations on our scale, with the actual real rate near 2.1 percent, the high end of its fifteen-year range, though still shy of the late-2023 peak near 2.5 percent. The breakeven, the market’s inflation expectation, is at +0.6, with the number near 2.36 percent. Real is running hotter than inflation compensation. That is the signature of a real-rate-driven move. The bond market is demanding more yield because the real cost of money is rising, and because someone has to be paid to hold a lot of long-dated government paper. It is not demanding more yield because it fears prices.

One honest caveat, because the inflation read is not a clean grind lower. That 2.36 percent breakeven is a pullback from a mid-May spike near 2.50 percent, when a Middle East oil scare briefly pushed inflation expectations up. So the breakeven coming down is partly the unwind of an oil fright, not a serene fade. The conclusion holds, the move is real-rate-led, but we want you to see the texture, not a smoothed cartoon.

What Is Actually Carrying the Long End

If the back-up is real and the front end is leading the policy path, what is doing the work at the very long end of the curve? This is where one of our proprietary lenses earns its place in the argument.

We track the 10-year term premium through what we call the Fiscal Pressure Index. Term premium is the extra yield investors demand to hold a long bond instead of rolling short ones, and it captures the part of the long rate that is not about expected policy at all. It is about supply, about deficits, about who is willing to fund the government at the long end and at what price.

The Fiscal Pressure read is above its 95th percentile against its own recent history, around 0.75 percent, above where it peaked in 2023. That is the engine under the 30-year. The long bond is near five percent because the market is demanding a fatter cushion to absorb the supply, not because anyone expects five percent inflation for thirty years. This is fiscal dominance showing up in a price. The deficit has a yield now, and you can read it off the curve.

That is the leading-read story in three figures. Breadth thinning since February. The front end leading the repricing. The term premium carrying the long end. None of it required Friday’s print to be visible. All of it was visible before.

The Laggards Were Still Asleep

Now turn to the reads that move late, because their silence is the tell.

Volatility spent months pinned near its lows. It did not start pricing risk while breadth was hollowing out, while the term premium was climbing, while real yields were grinding to multi-year highs. It waited until the tape broke, then jumped about forty percent in a session to a 20-handle. That is the textbook behavior of a lagging read. It reacts to price, it does not lead it.

Credit is the other laggard, and credit is still sleeping at the tights.

High-yield spreads sit near 2.74 percent, investment grade near 0.74 percent, both close to the lowest levels since the financial crisis. Spreads this tight say there is no default risk worth pricing. And the historical rhyme is uncomfortable. In 2007, investment-grade spreads camped out near 80 basis points well into the year, looking calm, looking fine, right up until they were not. The blow-out came after, with high-yield gapping toward eighteen points the following year. Tight spreads here are a sign that the part of the market that prices safety has not looked up yet.

Here is what makes the credit calm more than just complacency, and this is where the word camouflage earns its keep. Underneath the public index spreads, private credit is already cracking. The private credit default rate is running at 6.0 percent on a trailing-twelve-month basis, a record by Fitch’s count. Business development companies have been gating redemptions, slowing the door so investors cannot all leave at once. And once you count liability-management exercises, the polite restructurings that keep a loan from being formally tagged as a default, the leveraged-loan default rate runs close to double the clean number. The public tape shows you 2.74 percent and a quiet VIX. The private books show you something else. The stress is real. It is just wearing camouflage, off-exchange, marked quarterly, out of the index.

Credit Is Ignoring the Labor Read

Our second proprietary lens makes the credit complacency precise. We call it the Credit-Labor Gap, and it does one thing well: it measures whether credit markets and the labor market are telling the same story.

Read it carefully, because the honest version is more interesting than the dramatic one. The labor side is neutral, a z-score of roughly -0.03. Labor is not screaming. It is not stressed. It is sitting right at its average. So we are not going to tell you the jobs market is falling apart, because the composite does not say that.

The credit side is the problem. High-yield spread stress is at -0.96, nearly a full standard deviation into complacency. So the gap is this: credit is pricing a full sigma less risk than even a merely neutral labor backdrop would justify. The two are not looking at the same economy. Labor says fine. Credit says better than fine, nothing to see. When credit is this much calmer than the rest of the data warrants, the gap usually closes the hard way, with spreads moving toward the data rather than the data drifting up to meet the spreads.

Breadth Was Rolling Over for Months

We showed you one breadth read at the top. Here is a second, independent one, because a single internal can mislead and two pointing the same way is harder to wave off.

The McClellan Summation Index is a longer-horizon measure of advancing versus declining issues, a way of asking whether the broad market is accumulating or distributing. It peaked in early March near 394. It then fell off a cliff, collapsing toward 36, even as the headline index kept printing new highs into June. Two different breadth tools, measured two different ways, both saying the same thing months apart from the price high. The participation was draining while the index climbed.

This is the same shape as the share-above-the-200-day chart, and it is the same shape as the labor flows under the labor headline, and it is the same shape as private credit under public spreads. Surface up, inside down. The non-confirmation was not subtle and it was not new. It just was not on the front page.

And it sits on top of a concentration problem that makes the thin breadth more dangerous, not less. The top ten names are roughly 40 percent of the index, a higher share than the dot-com peak. The other four hundred ninety carry the rest. When breadth narrows in a market that concentrated, the index can keep rising on a handful of names long after the typical stock has rolled over. That is exactly what the figures show. The flag stayed up because ten people were holding it.

A Hawkish Bear-Flattening

Put the rate move and the policy path together and the curve gives you a clean read.

The 2s10s spread sits near +38 basis points and it flattened on the way up, with the front end leading. There are two ways a curve flattens. One is friendly: the front end falls because the Fed is about to cut, the long end holds, the curve steepens later as relief arrives. That is a bull-steepening setup and it is what risk assets want. This is the other kind. The front end rose and dragged the curve flatter, a bear-flattening, the market pricing tighter policy for longer. The curve is signaling resolve.

And the path has moved with it. The market has flipped from pricing cuts to pricing a hike by year end. We are not going to put a precise probability on it, because the number bounces around and pretending to three-decimal certainty is its own kind of dishonesty. The direction is the point. The conversation has shifted from “when do they cut” to “do they have to go again,” a regime change in expectations that happened underneath a stock index sitting at all-time highs.

Two Economies in One Print

Step out of the bond market for a moment, because the same divergence that runs through credit and breadth runs through the household data too.

Consumer sentiment is at -2.57 standard deviations, a record low mood. People feel terrible. And yet retail spending is at +0.03 year over year on our scale, essentially average. The wallet is still moving even though the mood has cratered. That gap between how people feel and what they spend is the two-economy signature, and for us it is the thing that decides how you read the American consumer right now.

It resolves the way these things always resolve, by cohort. The top of the distribution has assets, the assets went up, and that cohort keeps spending no matter what the survey says. The bottom of the distribution lives on wages, and real average hourly earnings have turned slightly negative year over year. Their paycheck is buying a little less than it did a year ago. That is who answers the sentiment survey with a record low and who keeps the headline spending number from rolling over only because the top cohort is carrying it. One print, two economies. The aggregate hides the split, and the split is the story.

Oil Round-Tripped, the Rate Move Did Not

Oil is the obvious place a skeptic would look to call the rate move inflationary after all. It does not hold up.

Crude spiked when the Strait of Hormuz scare hit, the same scare that briefly pushed breakevens to 2.50 percent in mid-May. Then it round-tripped, back to roughly $90.54 by June 5. And here is the tell: oil gave it all back, but the rate move did not. If the yield back-up had been an oil-and-inflation story, yields would have eased as crude retraced. They did not. The breakeven settled back near 2.36 percent and the real yield kept climbing. The reflation cover is gone. Strip out the oil round-trip and what is left is a clean real-rate move, which is exactly what the term premium told us in Figure 4.

The Dollar Says Stress, Not Yet Global

One more cross-check. The external channel.

Normally, rising real yields pull the dollar up as capital chases the higher real return. This time the broad trade-weighted dollar is off its highs even as real yields rise. The external channel, the one that transmits US tightness to the rest of the world through a strong dollar, is actually loosening a touch. The internal channel, the real cost of money at home, is tightening hard. That combination tells you the stress is domestic for now and has not yet spilled into a global dollar squeeze. It is a relief valve, and it is worth watching, because if the dollar turns and joins the real-rate move, the stress stops being a US story.

This Was Not a One-Session Panic

It would be comfortable to file Friday under “bad day, moving on.” We do not read it that way, and the reason is the calendar. Two events sit directly ahead, and both can extend this rather than end it.

The first is the May inflation report on June 10. Consensus has core around 2.8 to 2.9 percent, soft. And here is the counterintuitive part that we want to be very clear about: the dangerous print for risk is the soft core, not a hot headline. A hot headline this month is mostly oil, and the Fed looks through oil. A soft core, though, takes away the market’s excuse to keep dreaming about cuts in a world where the front end is already pricing the opposite. The soft print removes the last reason to disbelieve the curve.

The second, and the bigger one, is the FOMC on June 16 and 17. The decision itself is close to a foregone hold. The event is the projections. Whether they drop the last cut from the dot plot, whether the dots shift up, that is what moves markets, not the rate line. And this is the first meeting under the new chair, who has been clear about wanting a steeper curve and more term premium, pursued through balance-sheet composition: shorter holdings, less mortgage exposure, more bills. If that intent shows up in the projections, the term premium story in Figure 4 stops being a passing reading and becomes policy, and the thing carrying the long end gets an official sponsor that is unlikely to step back for a long time.

So this is not one ugly session to shake off. It is a setup with two near-term catalysts that can confirm it.

What It Means If You Do Not Trade It

Most of the people who read us do not trade a book, and this section is for you. You run a business, you sign the checks, you plan the year. Here is what the framework says you should take from all of this, in plain terms.

The cost of capital just went up, and it went up for the durable reason. This is not a temporary inflation spike that the Fed will cut its way out of in a quarter. The real cost of money is rising and the term premium is rising with it, which means borrowing long is more expensive and likely to stay that way. If you are financing anything multi-year, a buildout, an acquisition, a real estate decision, the rate you get quoted now is not a fluke to wait out. Plan around it.

Your demand base is splitting. The top-cohort consumer is fine and will keep spending. The bottom-cohort consumer is feeling a record-low mood and a paycheck that buys slightly less than last year. If you sell to the broad middle and lower end, the softness in sentiment is going to show up in your receipts before it shows up in any headline, because the headline is held up by the cohort that is not your customer. Watch your own numbers, not the aggregate.

And plan for a widening gap between the economy you read about and the economy you bill. The headline will keep saying the consumer is resilient because the aggregate is resilient, and your receipts may say something else entirely, and that gap will keep widening for as long as assets and wages keep diverging, because it is not a data wobble that mean-reverts next quarter, it is the two economies showing up in your own ledger.

What Would Change Our Mind

Strong views, weakly held. Here is what breaks this read, stated up front so you can hold us to it.

The thesis is wrong if three things happen together, not one of them, all three. Credit spreads have to stay tight, meaning the laggard never wakes up and the camouflaged stress really was nothing. Breadth has to re-broaden, meaning the share above the 200-day turns back up and the McClellan Summation reverses higher, so the inside re-confirms the surface. And real yields have to roll back over, taking the pressure off the term premium and the cost of capital. Tight credit and broad breadth and falling real rates, all at once, and the distribution story dies. We would take the other side gladly, because that is a healthier market than the one we are describing.

Any one of those alone is not enough. Spreads can stay tight while breadth rots, that is just the laggard sleeping longer. Breadth can bounce while real yields keep biting, that is a relief rally inside a downtrend. The three have to come together, because the whole argument is about the reads moving back into alignment from the leading edge in, not the lagging edge out. Until they do, the sequence stands.

We watch the parts that lead, and we wait for the part that lags to catch up. This week it caught up. The inside has been coming apart since February.

The surface finally noticed on Friday.

Get 20% off for 1 year

1. Two Numbers, One Release

The Bureau of Economic Analysis released the April Personal Consumption Expenditures price index at 8:30 this morning. Core PCE printed 3.3% year over year, in line with consensus and the highest reading since October 2023. Headline PCE printed 3.8% year over year, the highest reading since May 2023.

The release contained both numbers. The Federal Reserve writes policy from one of them. Most household budgets feel the other.

The market read the Core print as the actionable number. The S&P 500 closed at a fresh record at 7,520. The Nasdaq joined it. The curve disagreed. 10-year Treasury yields closed at 4.67%, the 30-year at 5.18%, both fresh 2026 highs.

Three readings of the same release. The equity tape says fine. The Fed’s preferred gauge says manageable. The long end of the curve says no.

Read more

20% off Yearly Subscriptions ends on Friday! Subscribe now to lock in the price.

Get 20% off for 1 year

1. The Widest Reading We’ve Ever Seen

The May print of the University of Michigan Survey of Consumers settled at 44.8.

That is the lowest reading the survey has ever produced. Below the 1980 stagflation trough at 51.7. Below the 2008 financial-crisis trough at 55.3. Below the 2022 inflation trough at 50.0.

In the same week, the S&P 500 closed at 7,473, a fresh all-time high and the eighth consecutive weekly gain. The Dow set a record at 50,580.

One country. Two readings. The widest page in the book.

Read more

One market is repricing risk in real time. The other two are asleep. This Beam is the whole read: where the stress actually sits, the mechanism underneath it, what the consensus is missing, and what we are doing about it in Crosscurrents, our live model book, where paid subscribers see every position and every change as it happens.

Lock the 20% before it closes.

The 30-year Treasury closed at 5.11% on Wednesday. It touched 5.18% the day before. The long end of the curve is repricing the price of money in front of everyone, and while it does, investment-grade credit spreads are two basis points off their tightest level of the year and the S&P is a rounding error from its record.

One of these markets is wrong. Our framework just told us which one.

Get 20% off for 1 year

Read more

The portfolio is Crosscurrents, funded and live on PiTrade under @LighthouseMacro. About a week in, three positions on: SHY, GLD, and XLP. Every position, every mark, every rebalance, timestamped and auditable as it happens, the misses included. Early marks went out against SPY. A macro book that holds short-duration Treasuries, gold, and defensives should not be measured against a long-only equity index, so we benchmark Crosscurrents against QAI, the liquid multi-strategy comp. The benchmark should say what the strategy is. The framework is not a worldview we narrate at you. It is infrastructure you can watch work, with a real book attached to it. Watch it move in real time: https://app.pitrade.com/user/lighthousemacro

The work is built for two readers. Market participants who trade the macro, and operators whose decisions ride on it without ever placing a trade. The same framework serves both.

What follows is a Beam in three parts. Part 1 is free. Parts 2 and 3, the plumbing under the long end and the institution that just changed hands, are for subscribers, with ten more charts and the full positioning read.

The price is $500 a year. For a limited time, the first year is $400. After that, standard pricing.

PART 1: The Tape Doesn’t Agree With Itself

Free preview · Parts 2 and 3 are for subscribers

Three things happened over the last week that we should sit with before we say anything else about them. On Tuesday, with the indices grinding toward records, the NYSE common-stock tape still printed more new 52-week lows than new 52-week highs, 51 against 42. On Thursday the S&P 500 closed above 7,500 for the first time, at 7,501. The Nasdaq Composite printed an all-time high at 26,635. The Dow crossed 50,000 for the first time since February. Then Friday took most of it back. The S&P closed Friday at 7,408, down 1.2% on the week’s high and roughly back where the week started. Jerome Powell’s last day as chair of the Federal Reserve was Friday. Kevin Warsh, confirmed by the full Senate on May 13, took the chair this morning and chairs his first FOMC on June 16 and 17.

Two of those are facts about the tape. One is a fact about the institution. All of them are unfolding inside the same five-day window, and that window also holds an inflation print at 3.8% year over year, the largest one-month jump since May 2023, a Treasury refunding running 70% above its quarterly projection, and oil that will not stay below $100 because the Strait of Hormuz is still on the table.

It is a lot. The market is treating it like a normal week.

We are not. Our composite read of sentiment against structure, what we call the Sentiment-Structure Divergence, sits in FEAR plus WEAK alignment. Sentiment is not euphoric and structure is deteriorating at the same time. That is the patient regime, the one where the data keeps printing and the eventual move comes from a single thread breaking rather than from the headline. The reading is the spine of this piece. Part 1 is the tape. The plumbing underneath it, and the institution sitting on top of it, are Parts 2 and 3.

The Setup

Start where the disagreement is loudest. The cap-weighted indices are at fresh records. The common-stock tape underneath them printed more new lows than new highs the same week. That is not a rounding error in the data. It is the data telling two stories at once, and the gap between them is the entire subject of this piece.

The records are real. The Mag 7 carried the index there, and inside the Mag 7, Nvidia carried the Mag 7. The “everything is great” narrative, read closely, is an “AI infrastructure is great, and also Nvidia” narrative. The index is doing what a cap-weighted index does when leadership narrows. It obscures what the median name is doing.

The Data

The percentage of S&P 500 stocks above the 50-day moving average sits at 47% as of May 14. That number is fine for the middle of a normal cycle. What matters at this index level is whether the breadth thrust has fired, the move from below 30 to above 70 within ten sessions, which is the marker that confirms participation broadening into a durable regime. We have not seen one of those since the spring of 2023. The path from the March 27 low at 20% up to 47% now is healing. Healing is a recovery off a washout. Thrust is a regime change. They are not the same event, and only one of them confirms a record.

Healing with the index at record highs is the configuration we have run back through 2007, 2014, 2018, 2020, and 2022. It does not always resolve into a top. It always shows up before one.

Our composite read of where the index sits relative to its 200-day against where the median stock sits relative to its 50-day, what we call the Structure-Breadth Divergence, flared toward its distribution band earlier this cycle and has since faded back below zero. On the 3-month smoothed series the read is negative today. We are not going to argue a top off a composite that is sitting below its own mean, and we are going to say that plainly rather than quote the one-day spike that would tell a louder story. The honest read is in the simpler internals, and they line up.

The cumulative Advance-Decline line has kept pace with the index. The median name has not. The S&P recovered its first-quarter drawdown faster than its own breadth did, which leaves a window where the index made back its losses on an advance the broad tape never confirmed. That is the circled region on the chart. An index can do that. It cannot do it indefinitely without either the breadth catching up or the index coming back to meet it.

The Mechanism

This is the engine. Q1 earnings growth ran +22.8% for the Mag 7 with Nvidia, +6.4% for the Mag 7 without Nvidia, and +10.1% for the S&P 493. Pull one company out of the cohort and the cohort’s growth more than halves and falls below the other 493. The top ten names are 38% of the S&P. The Mag 7 alone are 33%, the most concentrated the index has been in fifty years.

When breadth narrows into the cap-weighted top while the bottom of the index is still printing new 52-week lows, the index can rise while underlying participation thins. Capital is rotating out of the median name and into the largest names. The index does not show you the rotation. The high-low split and the median-name lag are how you read how far the rotation has already run.

What Consensus Is Missing

Consensus is reading Thursday’s record as confirmation of where the cycle sits. We read internal composition as the check on that, and the two have not agreed for two months. The gap is wider now than it was a week ago.

Narrowing participation while the index makes new highs resolves one of two ways. Participation broadens, which is the bullish path. Or the leaders roll over to meet the laggards, which is the bearish path. The leaders are still leading. The breadth read does not tell us the record is wrong. It tells us the question is when the leaders crack, not whether the laggards were a false alarm. The headline tape says they will not crack. The internals say that is the wrong thing to be sure of.

What Would Change Our Mind

The breadth thrust fires: the percentage above the 50-day crosses 70% within ten sessions of being below 30%. New 52-week highs on the common-stock tape outnumber new lows convincingly and stay there for a week. The cumulative Advance-Decline window closes because the median name catches up to the index rather than the index falling to the median name. Any one of those is a yellow light flipping toward green. Two together and the tape thread inverts, and we say so.

None of them are happening yet.

The So-What

The framework reads this regime as one of capital preservation, and the breadth picture is a large part of why. An index making highs on thinning participation is not a reason to be short. It is a reason not to chase the record. The patient regime is patient on purpose. We wait for the thread to break rather than guess which one and when.

That is the tape. It is the part anyone can see if they look past the index print. The two parts that follow are the parts the index is built to hide: the plumbing underneath the long end, and the institution that just changed hands on top of all of it.

Before you go: the book is live, and what is behind the paywall

The book is live. Crosscurrents runs on PiTrade under @LighthouseMacro, funded and auditable as it moves, the honest misses included. Watch it before you pay for the thinking behind it: https://app.pitrade.com/user/lighthousemacro

Parts 2 and 3 of this Beam are for subscribers. Part 2 is the plumbing underneath the long end, the balance sheet, term premium, funding stress, reserves, the spreads that move first when reserves get scarce. Part 3 is the institution: the dissent count Warsh inherits, what the curve and the committee are actually saying versus what consensus thinks they are saying, and where the framework puts the asymmetry. Ten more charts. The full positioning read.

Pricing. The price is $500 a year. We ran one launch window in the spring. It opened and closed on schedule, exactly as we said it would, and we are not repeating it. This is not that. We are holding the first year at $400 for a limited time, after which it returns to standard pricing.

If your decisions touch the macro environment, and almost everyone’s do, this is the cheapest insurance you will buy this year against being the last to know which thread broke.

Get 20% off for 1 year

Read more

The Horizon · May 2026

Subscribe now

Executive Summary

Q1 2026 real GDP grew 2.0% annualized. Roughly 1.1 percentage points of that, more than half of the entire print, came from computer and software investment. Inside the consumer line, healthcare alone accounted for 47% of the spending growth. Strip those two layers and the underlying private-sector economy ran closer to flat than to two percent. The headline reads resilience. The composition reads fragmentation.

This Horizon documents that fragmentation across six pillars and connects it to the rates regime that is repricing alongside it. Term premium on the 10-year sits at roughly 70bps, the floor of the pre-QE distribution. The framework’s honest level is 150. The Federal Open Market Committee printed four dissents at its April 29 meeting, the most at any single FOMC meeting since October 1992. Powell exits the chair on May 15. Kevin Warsh’s confirmation vote is this week. The institutional anchor is moving in five days, into a committee that has already begun fracturing in both directions.

We call this regime The Splitting Cycle. Two parallel cycles overlapping at the seam, held together by a tape that is still pricing one of them and ignoring the other. The framework’s job in 2026 is to read the composition.

We are not calling for an imminent recession. We are documenting a structural divergence between the headline economy and the composition underneath. The investment side is being underwritten by four companies pointing at $625 billion in 2026 capex. The consumer side is being held up by healthcare and government. The credit market is pricing the surface. Labor flows are giving mixed signals after the May 5 JOLTS print. The Fed is fracturing. The plumbing is running closer to capacity into the May 11 to May 13 auction window. The 30-year cracked 5% on May 4 and partially retraced on Iran headlines.

All of those are happening at once. None of them are noise.

Part I: The Splitter

The Q1 GDP advance landed at 2.0% annualized on April 30. Below the 2.3% consensus, above the Atlanta Fed’s 1.2% running estimate, and almost exactly where the St. Louis Fed’s tracker put it. The headline was a soft beat. The composition is the news.

Inside the print, non-residential investment grew at a 10.4% annualized rate, contributing 1.39 percentage points to the 2.0% headline. Consumer spending grew 1.6%, contributing 1.08 points. Government spending grew 4.4%, adding 0.73. Imports rose, subtracting from the total. Exports added back. The contribution shares tell the story the headline buries.

Look inside the investment line. Computer investment grew at a 67.4% annualized rate. Software grew 22.6%. Together, those two line items contributed roughly 1.09 points to GDP, more than half of the entire 2.0% headline. Equipment, intellectual property products, and private inventory all rose. Residential and nonresidential structures both fell. The investment cycle is concentrated by category, sitting in the hardware and software stack that runs the AI buildout.

Now look inside the consumer line. Real PCE grew 1.6%, of which 47.2% came from spending on health care alone. Durable goods consumption was flat. Non-durables fell at a 0.2% rate. The consumer is not weak in aggregate. The consumer is being held up by health care, which is the least discretionary category in the basket and the one most sensitive to demographics rather than business cycle.

The cycle is not lifting. The cycle is splitting.

The capital cycle running through that investment line has a name we have used before. It is the AI capex super-cycle, and the Q1 2026 reporting season just extended it without flinching. The four largest hyperscalers, Microsoft, Meta, Alphabet, and Amazon, collectively guided to roughly $625 to $640 billion in 2026 capital expenditures. Including Oracle and the broader top-five hyperscaler universe, the analyst consensus has now moved to roughly $700 to $750 billion. That is up from the January estimate of $620 billion. Big Tech raised the spending guide aggregating 2026 by $80 to $130 billion in a single quarter.

The dollar magnitudes are easy to gloss past. The macro magnitude is not. Four companies pointing at $625 billion in 2026 capex represent roughly 2% of US GDP being directed at AI infrastructure. There is no peacetime precedent at this concentration outside the late-1990s telecom and dotcom buildout, and the AI cycle is now running hotter on capex/revenue ratios than that one did at its peak. Per FactSet, blended S&P 500 earnings growth this quarter sits at 27.1%. The Magnificent Seven blended at 61.0%. The S&P 493, even with the Mag 7 cap weight stripped out, prints at 11.1%. One side of the capital cycle is producing the entire earnings tape on its own.

The pricing evidence inside the buildout is the part that surprises us most. Semiconductor industrial production minus broader manufacturing IP runs at +7.1 percentage points wide, persistent through every quarter of the last three years. Production of the things AI demands has been growing materially faster than production of everything else. Yet semiconductor PPI is flat. Core goods CPI is barely above 1%. The buildout is showing up as quantity expansion, not as price pressure.

Quantity without price is the signature of a cycle the Fed has no obvious reason to lean against. The 2021 goods cycle showed up as both more units and higher prices, which forced a hawkish response. The 2025-26 AI cycle shows up as more units at flat or falling prices, with the relative price story pointing the wrong way for inflation hawks. That changes the macro implication. Real yields, credit duration, and the equity multiple paying for back-loaded monetization are where the bill comes due, not headline CPI.

The capital intensity ratios make the dependency concrete. Capex as a share of revenue runs 86% at Oracle, 54% at Meta, 47% at Microsoft, 46% at Alphabet, and 25% at Amazon. Bank of America’s credit team estimates AI-related capex consumes roughly 94 cents of every dollar these companies generate after dividends and buybacks across 2025-26, up from 76 cents in 2024. Barclays models Meta’s free cash flow falling roughly 90% in 2026. The cycle is being financed by the residual after capital returns to shareholders, and the residual is shrinking.

That dependency is the thing the headline tape is not pricing. The S&P 500 makes new highs because the AI sleeve carries the index and the multiple compounds. The 493 is growing earnings at 11%, which is mid-cycle, not late-cycle. The 7 is growing at 61%, of which Nvidia alone does most of the lifting. Take Nvidia out of the Mag 7 group entirely and the remaining six grow earnings at roughly 6%. That is below the 493. The breadth of the earnings beat is narrower than the index print suggests. The dispersion of capex commitments across the same companies is widening.

The macro thesis we wrote in late April still holds. AI is fragmenting the cycle. The Q1 GDP print made that quantitative. The hyperscaler capex guides extended it. The earnings concentration confirmed it. The investment side accelerates. The consumer side eases. The engine creating the divergence is sustained and unhedged.

What would change our mind on the splitter thesis. Computer-electronic products inventory-to-shipments rises sharply, telling us supply has caught demand. Semiconductor IP rolls over relative to manufacturing IP. Real yields rise meaningfully and high-yield spreads widen while capex expectations stay aggressive, forcing the buildout to slow on financing constraint. Or one of the four hyperscalers cuts its 2026 guide on the next print. None of those have happened.

Part II: The Anchor Moves

The Adrian-Crump-Moench decomposition puts the 10-year Treasury term premium at roughly 70bps as of late April 2026. A year ago it sat at 40. Five years ago it ran negative. The framework’s honest level, the median reading from 1970 to 2007, sits at 150bps. We are halfway home.

The conditions facing the bond market today look meaningfully more like the pre-2008 regime than the 2014-21 stretch when term premium ran below zero. Structural deficits at full employment. Treasury funding short and deferring duration. The Fed out of the buyer seat. Inflation re-anchored at target with materially wider variance than the 2010s carried. None of those features were present during the QE-era anomaly. All of them characterized the regime where term premium averaged 150 to 250 for forty years. We anchor at 150 because that is the median across that distribution, and 70 is the floor of it.

The May 6 Quarterly Refunding Announcement was the structural test on the long end. We wrote the May 4 Beacon (”The Honest Level”) arguing that May 6 sets the pace. The market priced the worry into May 4. The 30-year yield briefly cracked 5%, the highest level since November 2023. The 10-year hit 4.45%, a nine-month high. Treasury supply concerns drove the move. Then the announcement landed. Treasury’s Q2 2026 net marketable borrowing announced at $189 billion, $79 billion above the $110 billion estimate Treasury published in February, a 72% overshoot inside a single quarter. Coupon sizes unchanged for the ninth straight quarter. The “at least the next several quarters” forward guidance language stayed. Net new cash from refunding $41.7 billion. Auction sizes set for $58 billion three-year, $42 billion ten-year, $25 billion thirty-year over the May 11 to May 13 auction window.

The market reaction was a partial retrace. The 30-year closed May 7 at 4.97%. The 10-year at 4.41%. The QRA was the non-event the announcement language signaled. The supply path was the catalyst the market priced two days early.

The destination did not move on May 6. The pace did. Treasury continues to fund through the front of the curve, leaning on bills, deferring duration. Each quarter that passes with stable coupon sizes is another quarter that adds to the eventual coupon issuance trajectory. Analyst expectations have shifted to 2027 as the year coupon increases begin. The framework reads that as a deferral, not a solution. The term premium reset is the price of that deferral, paid in basis points by buyers who require compensation that QE-era buyers did not.

The Federal Reserve is the second pillar of the rates regime, and the institutional anchor is moving in a way the dot-plot does not yet reflect.

The April 29 FOMC kept the target range at 3.50% to 3.75% for the third straight meeting. The vote was 8-4. Four dissents. The most a single FOMC meeting has produced since October 1992. Cleveland’s Beth Hammack, Minneapolis’s Neel Kashkari, and Dallas’s Lorie Logan dissented over the inclusion of easing-bias language in the statement. Governor Stephen Miran dissented in the opposite direction, calling for a quarter-point cut. Three hawks pushed back against the language. One dove pushed for action. The committee is fracturing in both directions.

A 33-year span without four dissents at a single meeting is not a number we cite for color. It is the institutional read on a committee that has lost the procedural consensus that has typically governed Fed decisions through every cycle of the last three decades. That fracturing matters now because the chair changes in five days. Powell’s term as Chair ends May 15. Kevin Warsh has been nominated, and the Senate confirmation vote is expected this week.

Warsh’s confirmation hearing on April 21 went fine on the substance. He committed publicly to Fed independence, declined to fix any future rate decision in advance of Senate questioning, and signaled what one analyst called a “regime change” plan for the institution. The procedural risk is concentrated in committee. Sen. Tillis is blocking the nomination over an investigation into Powell, and the Senate Banking Committee runs a 12-10 Republican advantage. One Republican defection in committee blocks the floor vote. As of this writing, the path to confirmation by mid-May is intact but contingent.

The forward question is not whether Warsh hikes. It is whether the dot-plot trajectory bends meaningfully more hawkish at the margin under a chair whose historical voting record sits to the right of the recent FOMC center. We expect it does. The committee that just produced four dissents on easing-bias language will not absorb a Warsh chair without the dot plot moving. The path of short rates that anchors the front end stays anchored. The supply path that anchors the long end keeps grinding. The curve steepens through the back, not through the front.

That is the Fed-side mechanism for term premium repricing. The fiscal-side mechanism still sits underneath, and the two operate in the same direction. The destination is 150bps. The path is grind, occasionally gap, partially retrace, repeat. May 4 was a single session of that pattern. The next one comes when the next catalyst arrives.

What would change our mind on the term premium thread. Term premium compresses sustainably below 50bps for three consecutive weeks. Or 10-year trades back below 4.10% with no growth shock to explain it. Or Warsh signals continuity at his confirmation hearing in a way that the market reads as Powell-trajectory preservation. Or Treasury executes a meaningful weighted-average maturity extension over the next two quarters that pulls the duration absorption forward and compresses the back-loaded supply story.

Part III: K-shape With a Caveat

The labor data took a turn on May 5 that the framework has to handle honestly.

The March JOLTS print, released that day, showed the hire rate jumping to 3.5% from 3.2% in the prior month. Hires rose by 655,000 to 5.6 million, the biggest single-month rise in over a year. The quits rate ticked up to 2.0% from 1.9%, just at the framework’s pre-recession threshold rather than below it. Job openings sat at 6.9 million, unchanged. Layoffs and discharges held at 1.2%. The print partially reversed the labor-flow weakening narrative we and others have been writing.

We have to read this two ways at once. The first read is straightforward. The flow data softened the late-cycle thesis at the margin. Hires accelerating off a crisis-era low is not consistent with a labor market about to break. Quits stopping their slide and re-anchoring at the threshold instead of below it removes one trigger we had named in earlier pieces. The Credit-Labor Gap, which we computed at -1.68 from the March data when it was the latest reading, will compress mechanically as the May print updates. Some of the framework’s spread-widening triggers fade.

The second read is the one we sit with longer. The March print is one data point. It is the first JOLTS release in six months that has pointed in the direction of recovery rather than continuation. The composition story underneath the headline did not change. Tech-sector layoffs continued through April. Microsoft, Meta, and Oracle cumulative announced reductions ran in the tens of thousands. Healthcare and transportation continued to carry net job creation. Bottom-half employment expanded. Top-half employment in tech, professional services, and finance contracted. The headline payroll has been firm because the bottom-half is still hiring. The K-shape composition has not closed.

Real wages tell the same story. The Q1 Employment Cost Index showed wages and salaries rising 0.8% quarter-over-quarter and 3.3% year-over-year, decelerating modestly from the prior trend. Inflation-adjusted wages and salaries grew 0.1% over the year. Effectively zero. Workers’ real spending power is being held flat by a CPI print that has run sticky on the goods side and re-accelerated on energy through the back half of Q1. The aggregate labor income that funds consumption is barely growing in real terms. Healthcare drove half of consumer spending growth in Q1. The other half came from trade-down, government transfers, and savings drawdown.

This is the part the headline labor and consumer data hide together. The hire rate jumped back, which is real. The quit rate stopped falling, which is real. Composition is still K-shaped, which is also real. Real wages are at zero, which is the hardest fact to argue with. The labor market is holding without enough force to fund a consumer currently spending out of an aggregate saving rate near 4%, with the bottom 60% running well below that.

The next labor surprise can come from either direction and both paths are consistent with the framework’s read. Top-half cuts continue and eventually compress the headline payroll. Or bottom-half hiring continues and the headline holds while real income growth stays flat. Either way, the K-shape composition holds. Either way, the consumer remains the lagging variable in the cycle.

What would change our mind on the labor thread. Real disposable personal income re-accelerates above 2% for two consecutive prints. The quit rate moves back above 2.2% on the next two JOLTS releases. Tech-sector layoffs reverse with hiring announcements at the same firms. ECI wages re-accelerate to 4% year-over-year while inflation cools, restoring real wage growth.

Part IV: Credit Looks the Other Way

The high-yield option-adjusted spread closed May 7 at 279bps. Below the 300bps complacency line we use as a regime marker. Roughly 200bps below the long-run average. The spread has tightened straight through the late-March SPX correction, the Iran supply shock, the FOMC dissents, and the GDP composition surprise. Credit has not flinched.

Investment-grade spreads tell the same story from the other side of the credit curve. IG sits near 80bps, the tightest in roughly three decades, in the bottom-decile of its post-2000 distribution. Both segments are pricing perfection. The relevant question is what they are pricing it against.

The Credit-Labor Gap, which compares high-yield spread positioning against labor fragility, was at -1.68 on the March composite refresh. Deeply through the -1.0 complacency threshold. The math on that gap will compress with the May 5 JOLTS print, because the labor side moved in the direction of recovery. The gap may not compress all the way. Quits at the threshold is not quits above it. Real wages at 0.1% is the consumer-income condition that prevents the labor signal from closing the credit-spread divergence on its own.

The market structure side is the second cross-check. The Structure-Breadth Divergence (SBD), which measures the gap between price-vs-trend and breadth-vs-trend on the S&P 500, breached its +1.0 distribution-warning threshold on April 14, traded above it for a week, dropped back into the +0.6 range, and re-breached on May 6 with a single-session jump from +0.72 to +1.17. The breach was real and volatile. By May 7 it had retraced to +0.90, just below the threshold.

The Market Structure Index (MSI) sits at +1.05 as of May 8, after a 3.1 z-score swing in the prior month. Structure is back. Breadth is participating. But the participation is shallow. The percent of S&P 500 members above their 50-day moving average sits at 52% with the index near all-time highs. Comparable record-tape episodes typically run 70%+. The 50-day rule is the cleanest read on rank-and-file participation, and the rank-and-file is barely participating at the index level.

The cross-asset composition tells a fourth story. Equities at all-time highs. The 10-year at 4.41% with the 30-year nearly at 5%. The dollar testing support. Gold near $4,800. Bitcoin consolidating. None of those are the cross-asset signature of a clean risk-on tape. The equity index is doing something the rest of the asset complex is not buying.

What would change our mind on the credit-and-structure thread. HY OAS compresses below 240bps and stays there for two months while quits stabilize back above 2.1% for two consecutive JOLTS releases. Or breadth thrust higher: percent above 50d crosses 60% with confirming percent above 20d above 70% on the same five-day window. Or the May 6 SBD breach stays retraced for two weeks of consecutive closes below +0.5. Any of those would mean the warning is wrong and the rally has earned the next leg on actual broadening.

Part V: The System Runs Closer to Capacity

The plumbing signal that matters is in the funding spreads. The SOFR-IORB spread, the cleanest read on whether reserves are doing the work they need to do, sits roughly 5bps negative on most days. By the level alone, reserves are still ample. The pattern around the level tells a different story.

Episodes through late 2024 and 2025 saw SOFR-IORB widen on quarter-ends and tax dates, the largest single-day spike being 32bps on October 31, 2025, the highest reading in five years. Those episodes are no longer isolated. They cluster around predictable cash-management dates, growing in magnitude. The April 2026 tax season passed without an acute incident, in part because the Fed had already halted QT on December 1, 2025, and begun reserve management purchases of approximately $40 billion per month in Treasury bills. Those purchases are technical management of seasonal flows. They do not rebuild the buffer. The Fed exiting the seller role does not put the Fed back in the buyer role.

A single quarter-end blip is noise. Repeated blips at predictable dates, growing in amplitude, are the system telling us that the buffer between abundant and scarce reserves is thinner than the level alone implies. We treat SOFR-IORB as a regime indicator rather than a level indicator for that reason. The level says ample. The frequency and amplitude of dislocations say closer to the threshold than ample suggests.

Reserves themselves sit at roughly $3.0 trillion as of early May, modestly above the $2.5 to $2.7 trillion Lowest Comfortable Level of Reserves (LCLOR) band Fed officials describe as the threshold below which funding stress historically emerges. The trajectory matters more than the level. RRP exhaustion, which carried the system through 2022 to 2024, was the first stage of buffer compression and finished long before the auction window we are about to walk into. The relevant pressure point now is the absorption capacity of the system that has to take down the supply.

The May 6 QRA confirmed the pace. Coupon sizes unchanged for the ninth straight quarter. Forward guidance language preserved. Net new cash from refunding $41.7 billion. The May 11 to May 13 auction window prices that confirmation. $58 billion three-year on Tuesday, $42 billion ten-year on Wednesday, $25 billion thirty-year on Thursday. The questions that matter run through the auction stat sheet. Tail or stop-through. Bid-to-cover. Indirect bidder share, which proxies foreign demand. Direct bidder share, which proxies real-money domestic. Primary dealer takedown, which is what is left over when the indirects and directs are out.

That last category is where the constraint binds. Primary dealers absorb whatever the rest of the auction does not, and dealer balance sheet capacity is regulated by the Supplementary Leverage Ratio. SLR treats Treasuries the same as risk assets for capital purposes. Each dealer’s ability to warehouse a heavy auction without flinching is a function of capital headroom against that ratio. The 2025 SLR exemption rollback that was floated and shelved would have given dealers more room to absorb supply. It did not happen. Dealers run closer to their internal warehouse limits today than they did a year ago, and the back-loaded coupon trajectory the QRA preserved compounds against that constraint quarter after quarter.

The Fed’s response to the next bind is the question Treasury is implicitly running through every QRA. Buyer-of-last-resort interventions during episodes like September 2019 happened on a 24- to 48-hour timeline. The system today is not at that point. It is closer to that point than the headline reserve level implies, and the auction calendar between now and the August refunding compounds the test.

What would change our mind on the plumbing thread. The May 11 to May 13 auctions all clear without tails, with indirect bidder share above 70% and primary dealer takedown below 18% on the thirty-year. Reserves expand back above $3.2 trillion through a sustained Fed bill-purchase program scaled larger than current pace. SOFR-IORB averages below zero through two consecutive quarter-ends without any single-day spike above 10bps. Any of those would mean the system has either absorbed the supply cleanly or rebuilt enough cushion that the next acute event recedes from the calendar.

Part VI: What Comes Next

Three calendar items shape the next thirty days, and each one is a fork.

The Iran path opened on May 6. The White House announced a 14-point memorandum of understanding intended to end the war. The framework calls for a 30-day window of nuclear talks, an Iranian moratorium on enrichment, US sanctions relief, and the release of frozen Iranian funds. Trump halted Project Freedom, the US escort effort in the Strait of Hormuz, citing progress in negotiations. Brent crude fell 8% to close at $101 on the announcement. WTI fell 7% to $95. By May 7 Brent was at $100, WTI at $94.81. The IEA had estimated the conflict was disrupting roughly 14 million barrels per day of crude transit through the Strait of Hormuz at peak.

The Iran fork has three branches. Branch one: the deal completes within 30 days. Oil moves toward $75. Energy equities rotate, the inflation tailwind compounds, the Fed’s “look through the supply shock” framing was right. Branch two: talks break down inside the window. Strait re-closes. Oil moves toward $115. Headline CPI re-accelerates, services inflation lags but pulls higher, the Fed has no cover to ease. Branch three: the talks drag through the 30 days without resolution. Oil stays in the $90 to $100 range. Premium stays priced. The macro outlook absorbs uncertainty as a variance, not a level.

The Warsh confirmation this week is the second fork. The committee vote is the binding constraint, not the floor vote. A clean confirmation puts Warsh in the chair on May 15 and shifts the dot-plot trajectory hawkish at the margin. A blocked nomination forces a renomination process that could leave Powell as a holdover or push to an interim chair, an institutional event the Treasury market has not had to price in 30 years.

The CPI print on May 13 is the third fork. Headline already running 3.3% with energy on a downward path post-Iran announcement. Core has been holding around 2.7%. The supercore reading, which strips housing, is what the Fed actually targets, and it has been the variable that determined the path of cuts in every prior cycle. A supercore print at or below 0.2% monthly buys the Fed time. A print at or above 0.4% removes the cover Powell has been using to look through the energy spike, and the FOMC dissent dynamic gets louder before Warsh even arrives.

Each fork resolves, in some form, before June 5. The framework’s job between now and then is to be patient with which branch fires and which threshold breaches. The book is mostly cash for a reason.

Bottom Line

The headline economy reads 2% growth, 3.3% inflation, 4-something unemployment, all-time-high equity prices, and tight credit spreads. By any single line, the picture is fine.

The composition reads differently. AI capex is doing more than half the GDP heavy lifting. Healthcare is doing half the consumer’s. The Fed is fracturing into its largest dissent count in 33 years. Term premium sits at the floor of a regime it doesn’t belong in. Real wages are at zero. The 30-year cracked 5% on a Monday and partially retraced by Wednesday on geopolitical headlines. The plumbing is running closer to capacity than the headline reserve level implies. The structural-breadth divergence breached and re-breached inside two weeks. The credit-labor gap was at the deepest reading since 2018 going into a labor print that partially relieved it.

None of that is new this week. Most of it has been visible for a quarter or longer. What changed in the last 30 days is the aggregate compression. The catalyst window we framed in the April 27 Beacon was supposed to test the framework. The framework reads strained, not broken. Strained is information.

The cycle is splitting. The anchor is moving. The book waits for the gates to align. The framework waits for the composition to either hold the surface up, or pull it down. We don’t know which way that resolves. We do know which way the composition is leaning.

The framework reads the tape. We trade the reads. The launch goes anyway.

Behind the Watch: the stack expands

The framework is what we have built over the last decade and a half across institutions. Twelve pillars, three engines, one master composite, a public model, and an active book. The Substack has been the primary surface for the framework since the publication launched. As of this Horizon, the surface starts to widen.

Three new surfaces alongside the Substack, one anchor that ties them together, and a rollout sequence that runs as the build queue clears.

The portfolio, public

The Lighthouse Macro portfolio is going public on PiTrade. Advisory through Pioneer Advisory LLC. Brokerage and clearing through Interactive Brokers LLC. The platform’s creator program lets us publish positions in real time, with subscribers able to follow and copy at $10 one-time per portfolio per the platform’s terms.

The portfolio is the framework, expressed within the platform’s available order types. Eight target positions plus an SGOV cash floor. Active versus staged separated by trigger. SPY as the public benchmark, because that is the lingua franca of the platform and because the position-level RS gates for equities all bench against it.

Subscribers will see the full position book by ticker and weight, updated in real time. Performance versus SPY. Trades as they happen, with timestamps. Strategy description and methodology. What subscribers do not see is the dollar amount of capital committed; that is private by platform design.

A note on the structure. The portfolio runs on PiTrade as a publicly tracked tactical macro book, with brokerage and clearing through Interactive Brokers and advisory through Pioneer Advisory LLC. Different vehicle from a hedge fund or a separately managed account, different regulatory architecture, different fee structure. The discipline is documented in writing every Sunday in The Horizon. The trades land on the platform with timestamps. The framework and the book live in the same room.

The portfolio launches with what the framework actually allows on day one. The first thirty days will look like patience, because the framework requires patience right now. The patience is the position.

The research surface, opened

We have been building the Lighthouse Macro research dashboard against the OpenBB Workspace. The first pillar-level apps ship to paid subscribers as they clear the build queue.

The mechanic. Lighthouse_Master.db, the SQLite warehouse that backs every chart and every composite we publish, is exposed to OpenBB Workspace through a FastAPI bridge running on Lighthouse Macro infrastructure. Subscribers point OpenBB at the bridge and read the full database directly. Roughly 2,100 series. Roughly 4 million observations. Refresh cadence matching the underlying source pipelines, daily on the composites.

What that means in practice. Every chart we publish in a Beam or a Beacon or a Horizon will be reproducible on the dashboard inside two clicks. Every pillar composite queryable as a live time series. Every threshold, every regime classification, every divergence flag we cite auditable against the source data.

The roadmap is dozens of pillar-level apps in the build queue. Labor flows. Credit-Labor Gap. MRI regime. Plumbing dashboard. Sentiment composite. Each pillar gets its own app surface as the architecture rounds out. We will publish each one as it ships, with a one-paragraph explainer in the Substack Note that week.

This is the layer of the stack that most directly answers the question we hear most often from institutional readers. “Where does the data come from, and can I see it?” Soon, you can.

The channel, and the bot

The Lighthouse Macro Telegram channel is shipping, paired with a framework-aware bot that paid subscribers can query directly.

The channel handles the cadence the Substack cannot. Live macro briefings during the New York session. Threshold-breach alerts the moment a pillar reading crosses. Auction post-mortems within an hour of the print. Real-time read-throughs on economic releases. Anything that benefits from being delivered in minutes rather than days lives on Telegram.

The bot is the more interesting piece. Trained on the full Lighthouse Macro framework documentation, the pillar specifications, the published Beacons, Beams, and Horizons, and the live data layer through the OpenBB bridge. Subscribers ask it framework-aware questions and get framework-aware answers in real time. “What is the current MRI regime, and which pillars are driving it?” gets a current-data answer. “Which composites have moved most over the last week?” gets a live ranked list. “Walk me through the Credit-Labor Gap calculation as of today’s print” gets the formula, the inputs, and the current reading.

The Substack remains where the framework gets articulated and where positions get explained. The bot is the layer that lets subscribers interact with the framework in between publications, on their own questions, on their own data.

Access to the channel and the bot is included in the standard paid subscription. Pairing instructions and onboarding ship to subscribers in the Substack Note when each goes live.

The grandfather, and the floor

This is the most important paragraph in the Horizon, for subscribers and for prospects.

Every paying Substack subscriber on the books before the launches above is grandfathered into the full Lighthouse Macro stack as it expands. That includes PiTrade portfolio access at the subscriber rate, the OpenBB dashboard, the Telegram channel, and the LHM bot. The Substack subscription is the anchor. Existing subscribers carry forward at their current rate, with permanent rate preservation against the full-stack pricing that goes live when the surface area is complete.

Founding members lock for life at their original rate. Subscribers on the books at the current $500/yr ($50/mo) rate carry forward at that rate as long as the subscription stays continuous through the public-pricing transition. The price you pay today does not move on you because we shipped four more products around it.

What that means for prospects considering today’s rate. The next move is up. We have not set a date, and we are not setting one in this Horizon, but the direction is determined. The full Lighthouse Macro stack at the current rate is what you get for subscribing before the next pricing move. After that, the rate moves and the grandfather rule above is what makes the prior rate stick for the cohort that got in.

The path forward

We are honest about what this means.

What stays the same. The framework we publish on Substack today, the Beam cadence, the weekly Beacon, the monthly Horizon, the Note rhythm, all of it continues at the current cadence under the current pricing. The free tier remains free. The paid tier remains paid at the current rate for current subscribers.

What changes. The surface area expanding. Hundreds of charts a month across Beams, Beacons, Chartbooks, and Horizons. Dozens of OpenBB apps surfacing the same data infrastructure that backs the publication. A live Telegram channel during the trading day. A framework-aware bot answering subscriber questions in real time. A public portfolio with timestamped trades. All of it sitting behind one credential.

The line between free and paid moves as the paid surface grows. The early-cycle macro framework explainers stay free, because they belong to the educational mission and the brand. The real-time readings, the live data access, the bot, the portfolio, and the deep-dive composite work move further into the paid tier. Some pieces that currently live above the fold will move below it as the stack rounds out.

The next public pricing move will reflect the stack as it exists at that moment, not as it exists today. The grandfather rule above is how we make that move without penalizing the cohort that got in early. Subscribe today and you are in that cohort.

If the framework, the book, and the stack we’re building are the kind of work you want in your inbox at the current rate, the math is straightforward enough.

Subscribe now

The remainder of this Horizon is for paid subscribers.

The Book at Launch Readiness

Eight framework positions plus a strategic cash floor. The framework requires patience right now, and the book reads patient. The active sleeve at 10am Monday sits at 13% on the tape; everything else stages behind named triggers, with the cash floor paid roughly 4% to wait. The cash floor is the framework operating as designed.

Framework readings as of 2026-04-29 close. Bench-corrected RS regimes; rate-of-change and distance-from-MA z-scores computed against 252-day rolling distribution.

The book by exposure

Long the term-premium and fiscal-dominance thread, expressed through gold at a structural pullback inside a multi-year breakout that is still intact on weekly timeframes. Long the short-end leg of the steepener trade, staged behind the price gate clearing the stop zone. Long defensive equity through staples, utilities, and healthcare, each staged behind relative-strength rotation against SPY. Long energy operators, staged on the next Iran-print catalyst. A small AI-infrastructure expression on the systematic screen, staged behind a momentum reset. Tail hedge via volatility through the catalyst window. Strategic cash floor parking the residual at roughly 4% carry while triggers wait.

Every line carries named triggers in writing. Relative-strength against the corrected benchmark; rate-of-change at two windows; distance-from-MA at two windows. The full position book, the per-line sizing, and the live state of each gate publish in real time on the PiTrade feed when it launches, and inside the trading day on the Telegram channel as triggers fire.

The active sleeve at 10am Monday

Two positions clear the framework’s day-one gates and fire at 10am Monday. The gold half-tranche, under a named exception against the multi-year breakout that is still intact on weekly timeframes. The volatility hedge, RS-exempt by design, full size against the four-fork catalyst window. Both are documented in the standing trade log with the named exception on the gold price gate cited.

Every other line waits. The patience is the position.

Today’s read on the systematic screen

The LHM systematic screen reads risk-on at the technical level. The cleanest cluster sits in broad equity and growth, concentrated in tech and AI-infrastructure exposure. The defensive sleeve the macro thesis prefers reads Red or Mixed on relative strength, technically short of clearing the gate today.

The framework’s response is the staged-versus-active split above. The macro view says lean defensive. The technical screen says the defensives are still pending. The book holds the targets and lets the gates fire.

Risk Matrix

Eight scenarios shaping the next thirty days, with probability, impact, trigger levels, and the pillars each touches. Portfolio reasoning updated weekly. Triggers fire intra-week in the Telegram channel.

Watchlist

What we are watching for staged deployment, and what would invalidate the current book. Live readings update in the Telegram channel as triggers fire.

Bottom Line

The catalyst window is open. The Fed anchor moves in five days. The plumbing runs closer to capacity each quarter. The composition of the cycle is splitting between an investment thread that has its own gravity and a consumer thread that does not.

The book runs the framework through a second gate. The portfolio’s first day on PiTrade starts patient. Most of the target weights sit behind triggers. The discipline is the published edge.

The stack runs the framework through a third gate. PiTrade puts the book on a public timestamped feed. OpenBB puts the data and the composites on a live workspace. Telegram puts the cadence inside the trading day. The bot puts the framework inside the subscriber’s own question. Four surfaces, one credential, rolling out as the build queue clears.

What we built over the last decade and a half was the framework. What we are launching now is the surface for it. The Substack stays the anchor. The current rate stays available through the rollout window. The grandfather rule stays permanent for the cohort that locks before the next pricing move.

The framework reads the tape. We trade the reads. The launch goes anyway.

That’s our view from the Watch. We’ll keep the light on...

Don't Navigate in the dark. Subscribe.

Bob Sheehan, CFA, CMT
Founder & Chief Investment Officer
Lighthouse Macro | Research | @LHMacro

Term premium on the 10-year Treasury sits at roughly 70 basis points. A year ago it sat at 40. Five years ago it sat below zero. The drift is in one direction.

We anchor the framework’s honest level at 150bps. It is the median reading from 1970 to 2007, and we use it because the conditions facing the bond market today look more like that pre-QE regime than the 2014 to 2021 stretch when term premium ran negative. Structural deficits at full employment. Treasury funding short and deferring duration. The Fed out of the buyer seat. Inflation re-anchored at target with materially wider variance. None of those features were present during the QE-era anomaly. All of them characterized the pre-2008 regime, when term premium averaged roughly where we say it should sit today.

Subscribe now

70bps is halfway home. The base case is straightforward: the front end stays anchored by a patient Fed, the long end keeps repricing higher as supply grinds against capacity, and the curve steepens through the back. That is what the data has been doing for two and a half years.

The next catalyst is Wednesday, May 6. The Quarterly Refunding Announcement gives Treasury’s read on issuance composition over the next three months and gives the market the chance to push the long end before the issuance hits. Refunding does not move the destination. It moves the pace.

The Reading: 70bps Today

We use the Adrian-Crump-Moench decomposition because it has the longest continuous monthly history, the New York Fed publishes it without revision games, and the cross-sectional behavior over four decades behaves the way theory says term premium should behave.

History matters here because the eye gets tricked by the recent past. From 1970 through 2007, term premium averaged a little above 150bps. The 1980s and early 1990s ran higher, regularly above 300bps. The mid-2000s ran lower, occasionally below 100. The full pre-crisis distribution clustered roughly between +75 and +250.

Then the world changed.

From 2014 through 2021, term premium ran negative for extended stretches. That was not a market judgment about future risks. It was an artifact of three forces compounding. The Fed owned a quarter of the Treasury market and absorbed duration aggressively through QE. Foreign official buyers, primarily China and Japan, ran reserve accumulation strategies that treated Treasuries as inelastic demand. And global growth and inflation expectations were anchored low enough that the path of short rates over a decade looked extraordinarily compressed. Term premium measured what was left over after those three forces. There was not much left over.

Those three forces have all reversed. The Fed has been a passive duration seller for two-plus years. Foreign official demand has flatlined and ceded share to private buyers who require compensation rather than absorbing reserve flows. Inflation expectations have re-anchored at target but at a variance materially wider than the 2010s carried. The conditions today look meaningfully more like the pre-2008 regime than the QE-era anomaly. The framework’s anchor falls out of that observation. Term premium lived in the 100-to-200 range for most of forty years under conditions that broadly resemble where we are today. We anchor at 150 because that is where the median sat across that distribution.

The 70bps current reading puts us at the floor of the pre-2008 range. The repricing has begun, and it has gone halfway. The framework’s argument is that the conditions warrant continuation.

The Spine: Fiscal Supply and Duration Mismatch

Every term premium framework that survives contact with the data starts here. The federal government issues debt. Someone has to hold it. The compensation that someone demands depends on how much debt is coming, what maturity it carries, and what the alternatives are.

The first leg is volume. The federal deficit is running at roughly 6.5 to 7 percent of GDP. The level is remarkable not for its magnitude but for its composition. We are running peacetime, full-employment deficits at a scale historically associated with recessions or wars. The cyclical adjustment is meaningful. If the economy entered a normal recession tomorrow, automatic stabilizers and revenue rolloff would push the headline deficit toward 10 percent of GDP without any new policy. We are starting from a structural baseline that leaves no room for the business cycle.

That baseline is not closing. CBO projections show deficits widening through the next decade under current law. The interest line alone, which was a manageable rounding error during the ZIRP years, now compounds against a debt stock that has roughly doubled since 2008 and a marginal funding cost that has materially repriced. We spend more on interest today than on defense.

The second leg is composition. Treasury has consistently chosen to fund a disproportionate share of issuance through bills rather than coupons. Bills are short-dated, easier to absorb, and roll cheaply when the front end is anchored. Coupons require buyers to take duration, which is where capacity constraints bite.

The bills share has ridden the upper half of the 15-to-20 percent range that the Treasury Borrowing Advisory Committee has historically described as appropriate, and has done so for two years now. The accounting of that choice is straightforward. By issuing bills today, Treasury defers the duration problem. By deferring the duration problem, Treasury accumulates an obligation to issue more coupons later. The math runs through eventually.

The third leg is maturity.

Weighted-average maturity sits near 92 months, the high end of the post-1980 range. This sounds like Treasury has done the duration work. It has not. WAM is a stock measure. The deficit run-rate is a flow problem. The deficit is wide enough that simply maintaining the current WAM requires accelerating coupon issuance from here. Letting WAM drift back toward Treasury’s historical six-year reference would require even more. Either path puts duration on the market faster. WAM tells us where we are. The forward question is what it takes to stay there.

The supply trajectory follows from those three legs.

This is the spine of the framework. Volume that does not stabilize, composition that defers duration, and a maturity profile that sets up an accelerated coupon issuance trajectory. Term premium is the price of all of that, paid in basis points, by buyers who require compensation that QE-era buyers did not.

The Cushion That’s Gone

For most of the post-2008 era, the Fed was the marginal absorber of duration. QE1 through QE4 added roughly five trillion in Treasuries and MBS to the System Open Market Account. That demand came in regardless of price, and it was the dominant reason term premium ran where it did during the QE years.

Quantitative tightening reversed that. Starting in 2022, the Fed began letting maturing Treasuries roll off the balance sheet at a defined cap.

The pace of QT moderated in 2024 and again in 2025 as Fed officials grew increasingly attentive to reserve scarcity. There is now active discussion about whether QT should end entirely. We expect it will, sometime in the next two to three meetings, depending on how reserves and money-market functioning behave. Ending QT is not the same as restarting QE. The Fed exiting the seller role does not put the Fed back in the buyer role.

What absorbed the duration that QT released? Primary dealers, on their balance sheets. Foreign private buyers when the currency-hedged yield pencils. Domestic real-money allocators when liability-driven mandates demand it. And money-market funds, indirectly, by lending into repo and term funding rather than buying Treasuries outright.

That last buyer matters. From 2021 through 2024, the Reverse Repo Facility absorbed peak balances above two trillion dollars. Money-market funds parked cash with the Fed at the RRP rate when private alternatives offered nothing. As the front end repriced and bills became more attractive, money-market funds rotated out of RRP and into Treasury bills directly.

That rotation was a one-time event. The buffer is gone. There is no longer two trillion of money-market liquidity sitting at the Fed waiting to flow into Treasuries when supply hits. That liquidity has already flowed. It is in the bills market now, and it will roll, but the cushion that absorbed every funding shock from 2022 to 2024 has been spent.

Reserves are the other side of the same coin. They have come down materially from peak. They have not yet hit the level Fed officials describe as scarce, but they have moved from abundant toward ample, and the trajectory matters. Without RRP as a release valve, reserve scarcity becomes the next pressure point.

The Plumbing Layer

When QE ran the duration market, vol was suppressed because the buyer of last resort had effectively unlimited capacity. With the Fed out of the seat, vol shows up first in the plumbing, before it shows up in the long bond.

Three signals carry this layer.

The first is funding spreads. SOFR-IORB and EFFR-IORB are the cleanest reads on whether reserves are doing the work they need to do. When reserves are abundant, SOFR runs at or below IORB. When reserves tighten, SOFR drifts above IORB, and the spread becomes a real-time gauge of dealer balance sheet stress.

Episodes through late 2024 and into 2025 saw SOFR-IORB widen on quarter-ends and tax dates. Those episodes are no longer isolated. They are the early warning that the system is operating closer to its capacity than the headline reserve number suggests. The pattern matters. A single quarter-end blip is noise. Repeated blips at predictable cash-management dates, growing in magnitude, are the system telling us that the buffer between abundant and scarce reserves is thinner than the level alone implies. That distinction is why we treat the SOFR-IORB spread as a regime indicator rather than a level indicator. The level says ample. The frequency and amplitude of the dislocations say closer to the threshold than ample suggests.

The Fed knows this. It is the reason QT pace has been throttled twice and why the ending-QT discussion is no longer hypothetical. The plumbing is doing the speaking, and the policy response is following.

The second is auction performance. The bond market gives us a quarterly stress test in real time. Tails, the gap between the auction stop yield and the pre-auction when-issued level, are the cleanest measure of whether dealers are stepping up.

Tails on the 10-year sat near zero through most of the 2010s. They have run consistently positive at four-to-eight basis points since 2022, and the trailing twelve-auction average has not retraced. The market is requiring a level of concession to absorb 10-year supply that simply was not asked of it during the QE years. Bid-to-cover has run softer than the trailing twelve-month average more often than not. Indirect share has been volatile, with weakness concentrated in the months immediately following announcements of larger issuance sizes.

These are not crisis signals. They are pace signals. They tell us the market is requiring more concession to absorb supply than it required two years ago, and the trend is not retracing.

The third is the futures-spot basis. When dealers are balance-sheet constrained, the basis between Treasury futures and the cheapest-to-deliver cash bond widens. The basis trade has historically been a relatively small position concentrated in a handful of hedge funds, but it has grown to a multi-hundred-billion-dollar size, and dislocations in the basis are the canary for broader funding stress in the Treasury market. Vol and dealer capacity both transmit through this channel.

Inflation and the Path

Term premium does not include the expected path of short rates. That path is the other component of the long-term yield, and inflation expectations bound where term premium has room to move.

The 5-year, 5-year forward breakeven sits in a range broadly consistent with the Fed’s 2 percent target plus a modest risk premium. Anchored is the right word. Stable is not. The variance of inflation outcomes is materially wider than the 2010s carried, and term premium has to compensate for that variance even when the central tendency lands at target. A 2 percent expected path with a 50bps standard deviation is not the same risk profile as a 2 percent expected path with a 150bps standard deviation. The framework reads the latter as the current regime.

Growth runs the other direction. If growth surprises higher, the Fed stays restrictive longer, the path of short rates carries the load, and term premium can sit lower. If growth rolls, the Fed eases, the path falls, and the curve steepens through the front rather than the back. Our base case is neither extreme, which leaves term premium as the marginal pricing variable. That is the working assumption. It is also the one most exposed to a regime shift in either direction.

Subscribe now

The Honest Gap: What We Don’t Price

Six lenses are load-bearing today. One is not. We will tell you which.

Pension and insurer regulatory demand for long-duration assets is a real long-end force, and it is not in our tactical framework. The mechanism is straightforward. Defined-benefit pensions and life insurers carry liabilities with very long durations. Liability-driven investment frameworks and NAIC capital rules push these institutions to hold duration on the asset side to match. When equity markets rally and pension funded status improves, pensions de-risk by rotating from equities into long-duration fixed income. That flow can act as a structural force in the back end of the curve. The 2022 UK gilt episode exposed the fragility of LDI on the way down. The slow grind of pension de-risking is the same mechanism on the way up.

We do not load on it tactically because the pace is wrong for our timeframe. We work on a three-to-six-month horizon. Pension de-risking is a multi-year flow. The signals are slow, the data lags, and the conviction that any individual quarter shows it cleanly is low. The position sizing implication of getting the regulatory demand piece exactly right over a six-month window is small relative to the position sizing implication of getting fiscal supply or plumbing wrong.

This is on the build list. A multi-factor term premium model that incorporates LDI flows, NAIC capital rules, pension funded status dynamics, and insurer reserve requirements would tighten the framework’s strategic positioning view. That work will ship. It will not ship before May 6.

We tell readers this because the alternative is overclaiming. The 150bps anchor sits on the fiscal supply leg, the duration mismatch leg, the Fed balance sheet leg, the plumbing leg, and the inflation-and-growth path. It does not require the regulatory demand leg to hold up. The pension flow is real, the LDI bid in the back end is real, the insurer reserve dynamics are real. They are simply slow enough that they look like a level shift across the framework’s window rather than a tactical signal inside it. When the model exists, we will fold them in. Until then, the honest answer is that we know the leg matters and we know our framework does not currently price it. That is a statement about our toolkit. The bond market itself is unchanged by it.

May 6 Setup

The Quarterly Refunding Announcement is the most important data point on the rates calendar between FOMC meetings. Treasury announces three things that move the long end. The composition of issuance over the next quarter, broken out by tenor. The guidance on coupon sizes through the rest of the year. And any updates to the buyback program for off-the-run securities.

We are watching four things.

First, the bills-versus-coupons mix. If Treasury continues to lean on bills disproportionately, the framework’s deferral observation gets stronger. Eventual coupon issuance grows, and the trajectory steepens. If Treasury rotates back toward coupons, the front-loaded duration accelerates and the long end has to absorb more, sooner.

Second, the language on average maturity. Treasury Borrowing Advisory Committee minutes and the Assistant Secretary’s statement carry signal about how Treasury thinks about the WAM trajectory. Any acknowledgment of WAM extension as a deliberate policy stance, or of WAM drift as a problem requiring correction, would push duration earlier in the issuance schedule.

Third, buyback program guidance. Buybacks remove off-the-run duration from the market, support liquidity in older issues, and at the margin shift duration absorption from real-money to dealer balance sheets. Expansion of buybacks would be incrementally bullish for back-end yields. Contraction or flatlining would be neutral.

Fourth, the dealer-tone tells. The TBAC minutes are released alongside the announcement and they carry the dealer community’s read on absorption capacity over the next quarter. Language about indirect demand softening, about dealer balance sheets bloating, or about specific tenors where issuance increases would be challenging is the kind of detail markets price within hours. The minutes are not flashy. They are the closest thing we have to a primary-source assessment of where the duration capacity actually is, written by the people who clear it.

The framework’s destination does not move on May 6. The pace can.

Invalidation

The framework can be wrong. Here is what would tell us so.

If term premium retraces below 40bps without a return to QE or a material foreign demand surge, the framework is wrong. The drivers we model would have to be overwhelmed by a force we have not identified, and that would mean the structural story is incomplete.

If the federal deficit narrows to below 4 percent of GDP through either revenue growth or expenditure restraint, the fiscal supply leg weakens materially. We do not consider this a base-case path under current policy, but a sustained move below 4 percent would force a reweight.

If Treasury executes a meaningful WAM extension over the next six to twelve months, accelerating coupon issuance and absorbing duration earlier than expected, the back-loaded supply story compresses. The framework still calls for higher term premium, but the path becomes faster and the destination potentially moves.

If inflation expectations re-anchor materially above target, with 5y5y breakevens persistently above 2.5 percent, then 150bps becomes a floor rather than a target, and the framework’s call understates the destination.

If the Fed restarts QE as a response to plumbing stress, the term premium repricing pauses or partially reverses for the duration of the program. The destination still holds. The path is interrupted.

We watch these. If any of them register, we revise.

Bottom Line

Term premium has reflated from a regime-driven anomaly to the floor of the prior distribution. The conditions facing the bond market in 2026 are not the conditions of 2005. The framework reads 150bps as the honest level given fiscal supply, duration mismatch, post-QT plumbing, and a re-anchored inflation regime with wider variance.

We are at 70bps. We see 150bps as the destination. May 6 does not move the destination. It sets the pace.

The reaction function we are watching is straightforward. A bills-heavy refunding with neutral coupon guidance and a steady buyback program is the path of least resistance, and it leaves term premium grinding higher week by week as supply hits a buyer base that is no longer subsidized by the Fed. A rotation toward coupons or an acknowledgment of WAM extension as deliberate policy is the path that pulls forward the back-end repricing, with the 30-year typically leading the move and the 10-year following within a session. A surprise announcement of expanded buybacks is the path that compresses term premium short term and gives the framework’s destination a slower glide. We are positioned for the first path and watching for the second.

The base case is repricing higher. The risk is that the pace is faster than the position size assumes.

That’s our view from the Watch. We’ll keep the light on...

This morning’s GDP advance came in at 2.0% annualized, a real upside surprise against GDPNow’s 1.2% nowcast and a meaningful reacceleration from Q4’s 0.5%. The internals tell the story the headline buries. Investment accelerated. Consumer spending decelerated. The cycle is not lifting. It is splitting.

Inside investment, the split sharpens further. The BEA flagged increases in equipment, intellectual property products, and private inventory, partly offset by decreases in residential and nonresidential structures. That decomposition is itself the AI capex fingerprint. Equipment and IP products are where information processing hardware and software live in the national accounts. Structures are everything else the investment economy traditionally builds. One side accelerated. The other rolled.

The AI build is the splitter. Most coverage runs through model releases, GPU backlogs, and earnings-call superlatives, all of which are downstream. Upstream, AI is a capital cycle, and capital cycles show up in standard macro data: investment shares, industrial production, relative prices, inventories, and the credit and liquidity that underwrite duration. The macro question is whether the shock is large enough to matter beyond tech.

Our read this morning is that it already does. Just not evenly.

The Data

Five charts carry the argument.

Subscribe now

Read more

The S&P closed at a record yesterday, 7,174. Thirty days ago it was below trend and our Market Structure Index sat at −1.99, deep in broken territory. Yesterday MSI printed +1.12. That is a 3.1 z-score swing in a month. We do not see swings like this outside of 2020.

Consensus is reading the rally as confirmation that the soft landing is intact. Earnings carried it. The AI complex carried it. Pre-FOMC blackout did the rest.

The framework reads it differently. The same readings that say structure is back also flag two divergences underneath. They both crossed our warning thresholds in the last two weeks. Most weeks one of them fires. This week both are firing at the same time.

Subscribe now

Read more

There are weeks where a lot of small things happen, and there are weeks where the whole economic story compresses into a single window. This is the second kind. Over the next ninety-six hours, we get the Federal Reserve decision, the four largest companies in America reporting earnings, and the three biggest data releases of the quarter. Most of it lands inside thirty hours starting Wednesday afternoon.

The reason it matters has been building for two months. The headline economy has looked fine. Earnings are growing at double digits for the sixth straight quarter. The unemployment rate has not moved. Consumer spending has held up. But underneath all of that, the data we watch most carefully has been telling a quieter story. Hiring has slowed to a level we have not seen outside recessions. Workers have stopped quitting their jobs at a rate that says they do not believe they could find a better one. Credit markets, the parts of finance that tend to flinch first when the cycle turns, have stayed completely calm.

That gap between what the surface says and what the flows say is the thing this week is going to test. Either the surface is right and the flows are noise, or the flows are right and the surface is about to follow. The data over the next four days will tell us a lot about which one it is.

The shape of the week

The architecture is unusually compressed.

Tuesday morning brings consumer confidence numbers. The University of Michigan series finalized at a record low for April. Conference Board confidence sat near 92 in March. Whether households are pulling back on spending, or whether this is sentiment running ahead of behavior, is one piece of the puzzle.

Wednesday is the spine. The Fed announces its rate decision at 2pm Eastern. Chair Powell holds his press conference at 2:30. Then between 4pm and 5:30, four of the five largest companies in America report earnings: Microsoft, Meta, Alphabet, and Amazon. Together they will tell the market how much they plan to spend on artificial intelligence infrastructure in 2026. That number, more than anything Powell says, will move the entire AI-adjacent economy on Thursday morning.

Thursday morning is the data triple. At 8:30am the government releases first-quarter GDP, March inflation through the Fed’s preferred measure, and the Employment Cost Index, which tracks how fast wages are rising. Apple closes the earnings cluster after the bell. Friday’s manufacturing survey tidies up the week. The April jobs report and the March JOLTS print fall outside this window, releasing the following week. The convergence we care about is FOMC, hyperscaler capex, and the Thursday data triple. All of it lands inside thirty hours.

Five trading days. Four pieces of our framework getting tested in real time. We have spent the year mapping the cycle. Now the cycle reads back.

Powell’s last podium

Markets are pricing roughly a 98 percent probability the Fed holds rates steady on Wednesday. We agree. There is no path to a rate cut with headline inflation at 3.3 percent year over year, gasoline prices climbing on the back of the Iran supply shock, and crude back near triple digits.

What matters is the tone. The April meeting does not include the Fed’s quarterly economic projections, which means every word in the statement and every Powell answer carries more weight than usual.

Three things are worth listening for.

First, whether the statement keeps describing risks as roughly two-sided. The March meeting framed it that way, which gave the Fed room to move in either direction without committing to anything. We expect that frame to hold. It is the language of a central bank that wants to signal patience.

Second, whether Powell repeats the line he used at Harvard at the end of March. He explained that supply shocks like the oil disruption tend to fade by the time monetary policy could respond to them, and so the right move is to look through them. Here is exactly how he put it: “By the time the effects of a tightening in monetary policy take effect, the oil price shock is probably long gone, and you’re weighing on the economy at a time when it’s not appropriate. So the tendency is to look through any kind of a supply shock.” That sentence is the Fed’s permission slip to ignore the gasoline-driven inflation print. If Powell repeats some version of it, the market reads patience. If he hedges, expectations for cuts later this year fade.

Third, whether the Fed signals anything about its balance sheet. The technical name for what ended in December is quantitative tightening, but what it really means is that the Fed had been letting bonds roll off its books for two years and stopped doing so. Since December, it has been quietly buying around forty billion dollars of short-term Treasury bills per month, which is supposed to wind down in May. The plumbing of the financial system is calm but thin. Bank reserves have been rebuilt from a stress trough last October, and the buffer that used to absorb shocks has been functionally drained.

This is most likely Powell’s last press conference as Chair. His term ends May 15. Kevin Warsh has been named as his successor, with Senate confirmation pending. We do not expect Powell to use the moment for theater. He is institutionalist by temperament, and the institution is what he has been protecting all year.

The internal composition of the committee has already shifted under the surface. On one end, Cleveland’s Beth Hammack is the only voter publicly entertaining a hike. On the other, Michelle Bowman has emerged as the dovish anchor, with multiple cuts in her March projections submission. The most interesting voice is Austan Goolsbee, who has historically been the most cut-friendly member in the room. He told Semafor in mid-April that if the oil-shock inflation lingers, cuts get “pushed out of ‘26” entirely. When even Goolsbee gives up on cuts this year, patience is the path of least resistance for the chair.

The capex print is the macro print

We have written before that hyperscaler capital expenditures are now a macro variable, not a tech-sector variable. What that means in plain English is that the spending plans Microsoft, Meta, Alphabet, and Amazon announce on Wednesday night will move the entire industrial economy on Thursday morning. Semiconductors, power infrastructure, data center real estate, utilities, copper, natural gas, all of it.

The arithmetic is the easy part. The four hyperscalers spent roughly $230 billion on capital expenditures in 2024. They spent roughly $350 billion in 2025. Their guides for 2026 add up to somewhere between $610 billion and $640 billion. At that level, four companies are spending the equivalent of nearly two percent of US GDP on what is mostly AI infrastructure. There is no precedent for this in any prior tech cycle.

What we are watching Wednesday night is whether any of these guides flinch.

Amazon is the cleanest tell. Its cloud business, AWS, is the part of the company that powers most of the AI demand the others are also chasing. If AWS grows above thirty percent and Amazon raises its 2026 capex guide, the AI cycle extends. If AWS slows toward twenty-five percent and the capex guide stays flat, that is the first crack in the story.

Meta carries a different question. The company has built itself a $115 to $135 billion capex envelope for 2026, and the question is whether ad revenue can carry that without margins compressing. Reality Labs, the part of the company building the metaverse, lost more than nineteen billion dollars across 2025. Meta is the largest single-stock event of the night.

Microsoft and Alphabet are both spending at rates that are reshaping their financials. Microsoft has said capacity, not demand, is the binding constraint on its cloud business. Alphabet’s earnings per share are actually expected to decline this quarter despite revenue growth, because the depreciation schedule on all the new data center capacity is now hitting the income statement. Both companies are showing what it costs to be in the AI infrastructure race.

The capital intensity ratios make the story concrete in a different way. Capex as a percentage of sales now runs roughly 86 percent at Oracle, 54 percent at Meta, 47 percent at Microsoft, 46 percent at Alphabet, and 25 percent at Amazon. Bank of America’s credit team estimates that AI-related capex consumes roughly 94 cents of every dollar these companies generate after dividends and buybacks across 2025 and 2026, up from 76 cents in 2024. Barclays models Meta’s free cash flow falling roughly 90 percent in 2026. The cycle is being financed by what is left after capital returns to shareholders, and what is left is shrinking.

There is one more wrinkle to all of this that does not get enough attention. The double-digit earnings growth we mentioned at the top is real. It is also extremely concentrated. The S&P 500’s blended growth rate of fifteen percent gets driven by the Magnificent Seven at twenty-three percent. But strip out Nvidia alone and that group falls to six percent, which is below the rest of the index at ten. The “everything is great” earnings tape is, on closer reading, an “AI infrastructure is great, plus Nvidia” tape. The rest of the S&P 500 is far more sensitive to consumer and labor conditions than the platforms are. And the consumer data is not great.

Three releases, one signal

Thursday morning at 8:30am, the government releases GDP, the inflation print, and the wage data inside the same minute. The market gets one number for growth, one number for inflation, one number for wages. The triangulation between them is what tells us where the economy actually is.

GDP first. The Atlanta Fed’s running estimate for the first quarter has drifted from a healthy three percent in late February down to roughly one percent. The St. Louis Fed’s version of the same exercise tracks closer to two and a half percent. Most of the gap reflects different ways of treating the surge in imports we saw in early 2026, as companies front-loaded purchases ahead of tariff actions. A print near one and a half percent, with consumer spending holding up, is the soft-landing read. A print under one percent, paired with hot inflation, is the start of a different conversation entirely.

Inflation second. The measure the Fed actually targets, called core PCE, has held steady around three percent year over year. The piece of it that matters most for monetary policy is what gets called the “supercore” reading, which strips out housing because housing tends to lag everything else by twelve to eighteen months. Anything in the high 0.3s on the monthly supercore reprices the cuts curve hawkishly. Anything at 0.2 or below buys the Fed more time. There is a quieter measure published by the Dallas Fed that trims out the most volatile categories on either side, and it has been running closer to two and a half percent. That is the calmer read on inflation, and it is probably the one Powell will privately weight.

Wages third. The Employment Cost Index has been cooling on a steady glide path for six quarters. The most recent print at 0.7 percent was the softest reading since 2021. A print at 0.7 again confirms wages are no longer adding to inflation pressure and gives the Fed cover to ease later this year. A print at 0.9 breaks the cooling trend and complicates everything.

The market reaction on Thursday morning is going to be asymmetric. A clean trio (growth around two percent, inflation cooperating, wages soft) is the soft landing. Equities hold the bid, the yield curve steepens a little, the dollar drifts. A bad trio (weak growth, hot inflation, hot wages) is the stagflation tape, and it will not be subtle. Yields rise on both ends, equities sell. The most likely outcome is somewhere in between: data that confirms the framework directionally without forcing the Fed’s hand or anyone else’s.

The flows tell the truth

We have written about this before, but it is the central argument of the year, so it bears repeating. Most of the time, the headline employment number does not lead the cycle. It lags it. The level of unemployment moves up only after the economy has already turned. What turns first is the flows: how often workers are getting hired into new jobs, how often they are quitting their current ones to look for better ones.

The flows have been telling a story all year that the headlines have not.

The hire rate, which measures how often workers are getting new jobs each month, sits at 3.1 percent. That is a level we have only seen at the worst moments of the global financial crisis and during the depths of the 2020 lockdown. It means the engine that absorbs displaced workers in normal recessions, the willingness of other employers to hire them, has slowed to crisis levels. Even though no one is being laid off in big numbers, the second half of the equation, the part that tells you how easy it is to find a new job, has effectively gone offline.

The quit rate tells the same story from the worker’s side. When workers feel confident, they quit jobs they do not like to find better ones. The quit rate at 1.9 percent has been sitting at or below 2.0 percent for eight straight months. Workers are staying put. They do not believe they could find a better job easily. That is not a level-of-employment story. That is a confidence story, and it tracks what would normally be the early warning signs of a turn.

So why has the unemployment rate not moved? Because layoffs have not started. Initial jobless claims, the weekly read on layoffs, are sitting at completely benign levels. The labor market has settled into what one Fed official called a “low-hire, low-fire” equilibrium. Nobody is hiring much, but nobody is firing much either.

The reason the framework worries about this state is that it is structurally fragile. When employers eventually pivot from “we are not hiring” to “we are reducing headcount,” there is no second job market for displaced workers to absorb into. The hire rate that absorbs them in normal recessions is not there. We can already see the early signature of that fragility in another piece of data: how long the average unemployed person stays unemployed. The share of unemployed people who have been out of work for more than six months is rising even as the headline unemployment rate is not. Duration is the early signal.

The announced layoffs that have not yet shown up in jobless claims also tell a story. Challenger reported roughly 60,000 announced job cuts in March, with AI cited as the reason for a quarter of them, up from ten percent in February. Tech-sector cuts ran roughly 18,700 in March alone, with the first quarter total up forty percent year over year. None of that has reached the weekly claims data yet because severance windows and retention payments smooth the timing. It will.

This is what we mean when we say flows lead and stocks lag. The flow data has been saying the labor cycle has turned for six months. The level data has not said it yet. The credit market, which prices off the level data, has stayed calm. That gap is the thing the framework is built to read.

The credit-labor gap

Investment-grade corporate bond spreads (the extra yield investors demand to lend money to companies versus to the government) sit near 80 basis points, the tightest level in roughly thirty years. High-yield spreads, the version for riskier borrowers, sit near 285 basis points, inside levels we have seen only twice in the last twenty years: May 2007 and July 2021. Both prior episodes turned out to be peak complacency.

How can credit be this calm with labor flows this weak? The answer is mostly technical. Credit spreads do not need labor to be strong. They need defaults to be low. Companies refinanced most of their debt in 2024 and 2025 at lower rates, and corporate balance sheets are sitting on plenty of cash. Near-term defaults look benign, and credit is correctly pricing that.

What credit is not pricing is the cyclical risk that follows once labor turns. The two-month lag between when labor flows weaken and when defaults begin to rise is the gap we are watching. Right now, one market is reading the level signals (everything is fine), and our framework is reading the flow signals (this looks late-cycle). Both can be defensible at the same time. They cannot both stay defensible forever.

The thinning of the consumer’s savings cushion is the same gap, viewed from a different angle. The headline personal savings rate sits at four percent, down from five and a half a year ago. Our “Two Economies” piece on April 20 walked through the composition of that aggregate. The top decile of US households is saving roughly eighteen percent of income off appreciating asset bases. The bottom sixty percent is saving 1.2 percent and rolling balances on revolving credit. The four percent headline is camouflage. The household sector has been spending against a thinning buffer for a year, and the cushion is mostly gone for the people who do not own stocks.

The Treasury market is too quiet

There is a number called the term premium, which sounds technical but is not. It is the extra yield investors demand for the risk of lending the government money for ten years instead of rolling shorter-dated debt over and over. When the term premium is high, it means investors see real risk in long-duration government debt. When it is low, they do not. Right now it sits near 70 basis points, up from 40 a year ago, but our framework points to a level closer to 150 basis points as the honest one given how much the federal government is borrowing, the heavy reliance on short-term bills inside that mix, and the gradual fading of foreign demand. The trajectory is right. The pace is the question.

The next catalyst is a Treasury announcement on May 6, the week after this one. The market is watching for one specific change: whether Treasury drops the language that has been telling investors coupon auction sizes will not increase “for at least the next several quarters.” If that line gets removed, long-term yields rise immediately as investors price in more supply ahead. If it stays, the long end stays anchored.

Bond market volatility, captured by what is called the MOVE Index, sits more than forty percent below its year-ago level. That is a remarkable level of calm given how much risk the framework sees brewing underneath. Some of that calm is real, and some of it is technical: vol-targeting strategies and Fed forward guidance both suppress measured volatility. The absence of a wobble in the Treasury market should not be confused with the absence of risk in it.

The oil overlay

Oil deserves a paragraph and not the spine. The 2026 supply disruption tied to the Iran conflict pushed crude back near triple digits in late February and has kept it there, with a steady geopolitical premium baked into the curve. Gasoline near four dollars a gallon is the single biggest reason headline inflation lifted to 3.3 percent in March. The Fed’s posture, articulated by Powell at Harvard in March, is to look through it. Oil is not the framework. Oil is the noise that makes the framework harder to read in real time. The signals we want sit in the inflation data on Thursday and in the wage print. If those land soft, the oil shock is doing what supply shocks do: showing up in headlines and not in the structure underneath. That is the read we expect.

What would change our mind

Honesty requires laying out what would force us to revise the framework. Three scenarios.

The first is a clean hyperscaler capex blow-out on Wednesday night. If Amazon raises its 2026 spending guide aggressively, Alphabet does the same, and Meta tightens to the upper half of its envelope, the AI capital cycle extends another four to six quarters. That spending feeds through the industrial economy, lifts the breakeven rate of payroll growth, and arguably defers the labor-flow reckoning we have been tracking. The bullish read on business conditions becomes the dominant one. Equities rally. The credit-labor gap stops being a mispricing and becomes a recognition that the cycle just got extended.

The second is hot wages paired with soft inflation on Thursday morning. If the Employment Cost Index reaccelerates while inflation cools, the wage-shelter loop we have been tracking breaks in a benign direction. The Fed gets cover to cut on the inflation read while wages support the consumer. That is the cleanest soft-landing scenario, and we cannot rule it out.

The third is high-yield spreads tightening further into the print week. If the credit market absorbs hot inflation, hawkish Powell tone, and weak GDP without flinching, our read on the gap is wrong, and what we see as complacency is actually correct pricing of a low-default regime. We would have to revise our credit framework.

We do not expect any of those three to materialize cleanly this week. The most likely outcome by Friday’s close is some version of the muddle: data that confirms the framework directionally without forcing anyone to reposition urgently. The market continues to price the level signals. We continue to price the flow signals. The gap stays open.

What we will know by Friday

Four reads land between now and the weekend. Each carries asymmetric information.

The Fed and Powell’s tone tell us how close the central bank is to acknowledging the labor flow story. We expect minimal acknowledgment with optionality preserved.

The hyperscaler capex aggregate tells us whether the AI cycle is extending or showing first cracks. We expect extension at the headline level, with widening dispersion in capital efficiency that becomes a 2027 problem rather than a 2026 one.

The Thursday data triple tells us whether stagflation pressure is intensifying or easing at the margin. We expect easing on growth and wages, sticky on inflation.

The manufacturing survey on Friday is the cleanest soft-data read. We expect it to print near a level consistent with mild expansion, with the prices-paid component staying elevated.

The framework does not ask the data to prove a forecast. It asks the data whether the cycle map we have drawn matches the terrain in front of us. By Friday’s close we will know whether the labor flow signal is getting confirmation from earnings, whether the credit signal is starting to wake up, and whether the Fed is moving any closer to acknowledging the asymmetry between what the surface says and what the flows say.

Conclusion

The cleanest test of a framework is when the data converges to a single window. This week is that window. Our read going in is that the labor flow signal is real, the credit market is pricing the wrong cycle variable, the hyperscaler capex super-cycle is real but increasingly capital-inefficient, and the Fed is correctly stuck. None of those views is contrarian taken individually. The combination is.

The wager underneath the framework is that flows beat stocks at the turn. The stock data this week, meaning the headline earnings number, the GDP level, the inflation level, will be loud. The flow data, meaning the trajectory of capital spending, the hire rate the following Tuesday, the wage print, the composition of inflation underneath the headline, will be quieter. If we are right, the noise resolves into the signal we have been tracking all year. If we are wrong, we will know it because the AI capex prints come in flat, the wage data surprises hot, and credit tightens further into all of it. We do not expect that. We are watching for it anyway.

By next Sunday we will have the Fed decision, four mega-cap reports, GDP, the inflation print, the wage print, and the manufacturing survey on the desk. The framework will be either confirmed or strained. Strained is fine. Strained is information. We will report what the data did to the cycle map. The discipline is not the conviction. The discipline is the willingness to repaint the map when the terrain changes.

The terrain has not changed yet.

The S&P 500 printed a record this week.

Every pillar in our twelve-pillar framework is flashing late cycle.

That gap is the story. Price is the last variable to turn. The flows under it turned quarters ago.

This edition runs the diagnostic layer pillar by pillar. Sixty charts across twelve pillars, tracking where the fragility is concentrated, where it is not, and how the internals are diagnosing the distribution in real time. Composite synthesis returns next Chartbook. For this one, we let the indicators do the work.

What the data says

Labor is freezing, not cracking. Quits at 2.0%. Hires at 3.2%. Layoffs still historically low at 1.1%. The market isn’t breaking, it’s seizing up. Temp help is 38 months inside the recession signal. At some point the firing means something.

Prices still above target. Headline CPI 3.3%, core PCE 3.0%. Trimmed mean 2.3%. Strip the tails and inflation still prints above 2%. The last mile math has not changed.

Subscribe now

Read more

The S&P 500 closed at 7,137.90 on Wednesday, a fresh record. JPM lifted year-end to 7,600 on Monday. The VIX just completed a 44% compression in three weeks, from 31.05 on March 27 to 17.48 on April 17, one of the sharpest on record. The tape reads clean.

The internals do not. Our Market Structure Index has moved from −1.03 on April 1 to +1.38 on April 22. Two and a half standard deviations in three weeks. Structure is chasing price, which is what distribution looks like.

Subscribe now

Read more

March retail sales rose 1.66% month over month, the biggest monthly gain in more than a year, and twelve of the report’s thirteen categories were green. The control group, the measure that feeds directly into GDP, climbed 0.76% against a 0.2% consensus. Bloomberg ran it as broad-based resilience. Benzinga ran it as consumers shrugging off the pump shock. The tape bought the number early, then gave most of it back into the close.

Headlines did their job. The composition is where the story actually lives.

Read more

The launch window on the Diagnostic Dozen Discount rate closes tomorrow, April 21, at 9:30 AM ET. This is the lowest you’ll be able to lock in the real-time application of the framework. Subscribers will get 200+ charts and 10,000+ words a month, plus the model portfolio, live-tracked and audited via PiTrade. They’ll also have first dibs on future rollouts. We already have some cool stuff brewing with Permutable AI and OpenBB. This is just the beginning, and I hope you’ll consider joining along for the ride. I strongly believe you’ll find it worthwhile.

A few things landed this week worth sharing before the window shuts.

The Diagnostic Dozen articles stay free. The Beacon stays free. Everything else, Beams, Chartbook, Horizon, model portfolio, moves behind the wall at tomorrow’s bell.

The Beacon · April 20, 2026

On Friday, we wrote the companion piece ahead of schedule. Three other writers had each moved the ball that week, and the conversation was ready for it. We promised the full chart pack would drop Sunday. A weekend hardware issue had other plans. Here it is Monday, with eighteen charts, the methodology, the sources, and where the framework sits today.

Executive Summary

The headline consumer data averages two populations that have almost nothing in common. One is spending down an asset base that keeps appreciating. The other is being repossessed. The aggregate numbers only make sense as statistics. As a description of any actual American household, they are fiction.

This Beacon gathers the eighteen charts we have been building to show that. It updates the January thread, folds in recent contributions from James at JB Macro, Jeff Horwich at the Minneapolis Fed, and Tim Pierotti at WealthVest, and closes with four threshold breaches that just printed in the underlying data.

The four breaches matter. So we lead with them.

The arithmetic of the "resilient consumer" works only because the top half of the distribution is spending a growing asset base and the bottom half is spending borrowed money it cannot repay. Neither is stable. They are only stable together, and only as long as asset prices keep rising and credit keeps extending. The labor data just told us which side of that pairing is cracking first.

Why We Are Writing This Piece Again

We laid out the two-economies framework in January, first in a thread on the 15th, then on Substack the week after. The argument was straightforward: the aggregate consumer data is a weighted average of two populations with almost nothing in common, and treating it as a single number hides the risk.

The conversation has caught up. James at JB Macro traced the equity wealth channel in February and followed up this week, noting that strong equity performance is doing what wages aren’t. Jeff Horwich at the Minneapolis Fed worked through Moody’s, Bank of America, the NY Fed, and the BLS Consumer Expenditure Survey in March and concluded the K-shape debate is unresolved because every source stratifies by income instead of wealth. Tim Pierotti at WealthVest Macro published on the generational wealth transfer this week, the leg of the wealth story almost nobody has quantified.

Each found a channel. James found equity. Tim found inheritance. Horwich found the measurement gap that lets both hide. We would add a fourth piece they have left out: the credit-distress data on the other side of the same K, and the labor-flow data that is now confirming it in real time.

This is the chart-pack version. Eighteen figures, four threshold breaches, one framework. Where Friday was the narrative update, today is the receipts.

One frame to hold in your head as you read: the consumer is a lagging indicator pretending to be coincident. By the time retail sales roll, labor cracked 6-9 months prior, credit stress was visible in delinquency data, and confidence had been deteriorating for two quarters. The consumer does not predict. It validates. This Beacon is about what it is about to validate.

The Wealth Story

The top 10% of US households by wealth own 67% of all household net worth. The top 1% alone own more than the entire bottom 50% combined, by a factor of roughly twelve

This is not new. It has been true for a while. What is new is the degree to which asset prices over the last five years have done the top decile’s saving for them. The S&P 500 is near all-time highs. Home prices for the top tier of the market are at all-time highs. When your balance sheet grows 15% a year unrealized, you do not need to save 8% of your income. Your wealth is saving for you.

James’s JB Macro piece captured this cleanly. The US saving rate fell from 6.1% in Q1 2024 to 4.0% in February 2026 while the UK’s rose, despite both populations facing similar cost-of-living shocks and elevated rates. He estimates that without the drawdown, US nominal GDP would have been roughly 1.5% smaller than it is today. In his follow-up, James noted that the buffer for US households is equity wealth, which is pro-cyclical. Strong equity performance pushes the saving ratio down through the wealth-effect channel.

He is right. But the mechanism is narrower than it first appears, because the 4.0% aggregate is not distributed evenly.

The Savings Story

The consumer cycle runs the same way every time. We call it the Three-Stage Stress Sequence. Stage 1 is Savings Depletion: income growth slows but spending habits persist, so households draw down their buffer. Stage 2 is Credit Substitution: the buffer is exhausted, so credit takes its place. Stage 3 is Spending Collapse: credit dries up or becomes unaffordable, and the spending data finally rolls. Stage 1 is complete. Stage 2 is where we live now. Stage 3 is what the labor print just started to warn about.

The pandemic-era cushion is gone. The SF Fed put aggregate excess savings at roughly $2.1 trillion at peak in 2021. That pool was fully depleted by Q1 2024. Since then, households have been financing consumption by running the saving rate below its 2000-2019 baseline, a stretch now approaching three years.

That is the aggregate. The distributional picture is where the story sharpens.

The top 20% of US households still carries roughly $480 billion of positive excess savings, the residual of pandemic windfalls that mostly landed at the top of the distribution. Every cohort below the top 20% is now underwater versus the pre-pandemic trajectory, with the bottom 20% down $140 billion on its own. The bottom 80% is collectively $380 billion below where its savings trajectory would have been without the last four years. The pandemic did not create a universal cushion. It created a cushion for the top 20% and a liability for everyone else.

That asymmetry shows up in the saving rate by cohort. The top 10% saves roughly 18% of income and is actually saving more than it did pre-pandemic, because the wealth-effect cushion means they do not need the income flow to service their lifestyle. The middle 60% saves close to the aggregate, around 4.8%, down about 2 percentage points from where it ran before 2020. The bottom 60% saves approximately 1.2%, functionally the floor. You cannot save less than your grocery bill.

That distribution tells you that when the aggregate saving rate falls from 7% to 4%, it is not a story about the average American cutting back. It is a story about the top decile drawing on a wealth cushion that grew faster than their income, while the bottom 60% continues running at a subsistence saving rate because there is no slack left.

Look at the relationship between the saving rate and consumer credit growth. The saving rate sits at 4.0%, well below the 8.5% historical average. Consumer credit year-over-year growth has stabilized near 3.2%, but the level has marched higher every single month. What you are looking at is a consumer who has stopped saving and started borrowing to fund the same basket of purchases. Spending growth, at the margin, is being funded by the balance sheet, not the income statement.

The debt service ratio, required debt payments as a share of disposable income, is the receipt for this substitution. It has climbed to 11.3%, above our 10% stretched threshold. The driver is not new borrowing, it is the mechanical repricing of existing variable-rate debt at higher rates. Credit card APRs averaging 22.3% for balance-carriers mean even a flat balance generates a higher payment. More of every paycheck is pre-committed before discretionary spending gets a dollar.

The composition of that spending growth matters more than the total. And that composition, as the next section shows, is lopsided.

Horwich’s Minneapolis Fed piece worked through four major data sources (Moody’s, Bank of America, the NY Fed’s Economic Heterogeneity series, and the BLS Consumer Expenditure Survey) and concluded the K-shape story was ambiguous at best. Moody’s showed a steep K. The NY Fed showed almost none. The BLS CE showed no K, and in 2024 lower-income households actually posted the fastest spending growth of any quintile.

His closing observation is the one to focus on: “data on spending-by-wealth might provide a complementary view of the scale, and the shape, of any divergence in household spending.”

That is the whole answer. The K-shape debate he reviews is inconclusive because every source he examines stratifies by income. Income-stratified data will always look less bifurcated than wealth-stratified data, because wealth in the US is roughly three times more concentrated than income. The top 10% earn about 45% of income. The top 10% own 67% of wealth. Of course the two pictures do not match.

Stratify by balance sheet instead of paycheck and the bifurcation is not ambiguous. It is clean.

There is one more layer before the credit story, and it is the one almost nobody shows you.

Inflation does not hit every income cohort the same way. The top 20%, because a smaller share of its budget goes to shelter, food, and energy, has experienced effective inflation closer to 3.2%. The bottom 20%, which spends a much higher share on those three categories, has faced inflation closer to 6.1%. The spread is 280 basis points. When the headline CPI prints at 3.3% and the Fed declares victory, the top of the distribution has already seen the benefit, while the bottom is still paying rent increases and grocery bills that have not stopped climbing. Inflation is a regressive tax and it has been grinding for three years.

Combine a negative real income story for the bottom 60% with an above-aggregate inflation experience for that same cohort, and the credit data stops being surprising.

The Credit Story

While the top decile spends off asset appreciation, the bottom 60% is running out of runway. The credit data says so directly.

The aggregate auto delinquency rate across all commercial banks sits at 2.70%, well below the GFC peak of 8.2%. The headline looks contained. That is the whole trap. The aggregate averages prime and subprime together, and prime is fine. Subprime is not. Fitch’s Subprime Auto ABS Index puts 60+ day delinquency at 6.9%, the highest since the series began in 1994 and above the 2008-2009 peak. The distinction matters because the stress is not distributed evenly across the borrower pool. It is concentrated in the bottom quartile, the same cohort already visible in the credit-card and student-loan data.

Pull the comparison to the 2008 peak directly and it sharpens. Every subprime auto stress metric we track is now above its GFC peak. 60+ day delinquency: 6.9% now versus 5.2% at the 2008 peak. Repo rate: 3.8% now versus 2.9% then. Negative equity (borrowers who owe more than the car is worth): 24% now versus 22% then. The subprime auto market is in worse shape today than it was heading into the financial crisis, and that is with the S&P at record highs and the unemployment rate still under 5%.

Vehicle repossessions hit 1.73 million in 2024. The 2009 peak was 1.77 million. We are effectively at post-crisis levels in a non-crisis macro environment. Auto loans are a particularly clean signal because Americans generally pay the car loan last. It is the vehicle to the job. When repossessions rise, it means households have exhausted credit cards, skipped rent, maxed out buy-now-pay-later, and finally defaulted on the bill they try hardest to protect.

It gets more specific.

Across the consumer credit stack, delinquency rates are rising. Credit cards lead at 2.94%, up 33 basis points versus Q4 2019. Consumer loans broadly are at 2.62%, up 31 bps. Auto loan delinquencies at 2.70%. Mortgage delinquencies, where underwriting tightened materially after 2008, are at 1.78% and essentially flat versus pre-pandemic. Business loan delinquencies are also subdued at 1.34%, which is a clue. The stress is concentrated on the consumer, not the business, side of the balance sheet.

Relative to the Q4 2019 pre-pandemic baseline, every loan category is showing stress, but student loans dominate. New delinquency transitions on student loans are 691 basis points above 2019. Credit cards are 174 bps above. Auto is 79 bps above. Mortgages are 35 bps above. Basically flat.

The stress is building from the bottom of the household balance sheet up. Unsecured consumer credit is where it shows first because it is where the working-age cohort has the most exposure and the thinnest collateral.

Tim’s piece argues that “the money center banks all said basically the same thing about consumer spending and credit quality, which is that they see no slowdown and no signs of emerging credit weakness.” That is accurate for bank balance sheets. Prime consumer portfolios are fine. The weakness does not show up in JPMorgan’s 10-Q because JPMorgan does not carry subprime auto paper or the bulk of the student loan book. It shows up in Experian, Cox Automotive, and the NY Fed Consumer Credit Panel. Different data, same population.

The Labor Story

You would expect a labor market signal to precede credit stress of this magnitude. It is there. You just have to look underneath the BLS headline. And as of the data we pulled this weekend, it is no longer quiet.

The quits rate just broke through 2.0%. JOLTS February print: 1.9%, down from 2.0% a year ago. This is the first time this cycle we have been below our 2.0% pre-recessionary threshold. The quits rate is the cleanest read on worker confidence there is. Workers quit when they believe they can find a better job. They stay when they are scared. They just stopped quitting.

Long-term unemployed share at 25.2%. Above our 22% structural fragility threshold, up 4.2 percentage points in twelve months. The pool of people who have been unemployed 27 weeks or longer is now a quarter of the unemployed population. These are the workers the labor market has the hardest time reabsorbing. They are also, disproportionately, the bottom-60% cohort whose credit data we just walked through.

The sharpest read on this dynamic comes from ADP’s Main Street Macro piece last week. Turnover at employers with fewer than 50 people fell to 3.9% in March, the lowest reading in the nine years ADP has tracked it. The last-three-year baseline for all private employers has run near 4.7%. A separation rate that low is not a sign of health. It is the signature of workers who have concluded there is nowhere better to go.

Two things matter here. First, the composition. Nela Richardson notes that in February and March, small employers accounted for nearly all net private-sector job gains, even as overall turnover collapsed. That is a bifurcation inside the labor market: stasis at the worker level, dynamism concentrated at the one segment of employers most vulnerable to macro shocks. Second, the direction. Low turnover supports spending in the short run because paychecks keep flowing. Over time it starves the labor market of the reallocation that drives wage growth and productivity. The short-run comfort is exactly what sets up the long-run problem.

The long-term unemployment share is the slow-moving part of the labor data. It does not spike on a single jobs report. It climbs. And it has been climbing for eighteen months. A worker who has been unemployed for six months is statistically unlikely to return to the labor force at prior wages. A growing share of long-term unemployed is a growing share of the labor market being written off, quietly, in the background of headline unemployment numbers that still read fine.

And the composition of that long-term unemployment is not random. Workers 55 and older carry a 31% long-term unemployment share, the highest of any demographic cohort. When older workers lose a job, they usually do not get another one at the same wage. That 31% figure is not a queue. It is a cliff. Black workers, separately, carry a 7.5% U-3 rate and a 28% long-term share, both among the highest in the data. The headline unemployment rate of 4% averages all of this together into a number that, by cohort, nobody is actually experiencing.

And then the income math.

Aggregate weekly payrolls, the product of employment times hours times wages, is growing +3.7% nominal year-over-year. Real, after deflating by CPI at +3.3%, that is +0.4%. Near zero. The tailwind that funded consumer spending through 2023 and 2024 is effectively gone. The top quintile does not care because the top quintile’s income does not come primarily from wages. The bottom 60% does care, because wages are close to the whole thing.

Average weekly hours are flat at 34.2. Temp help payrolls are down 2.1% year-over-year, approaching the -3% threshold we watch as a recession precursor. Initial claims four-week moving average sits at 209,750, still below the 230,000 stress level, but continuing claims are sticky at 1.82 million.

And then the revisions. The January 2026 benchmark revisions subtracted 898,000 jobs from the 2025 totals. The average monthly gain for 2025, after revisions, was just 15,000. Not the 150,000-to-200,000 that was reported in real time. Fifteen thousand. That is not a labor market growing. That is a labor market that was not growing and nobody knew it until the benchmark caught up to the truth. The current month’s print, whatever it says, has to be read through that lens now. The prior year was softer than it looked, and the current trend is softer still.

Horwich notes, correctly, that the K-shape in spending data is inconsistent across sources. But the labor market tells us something the spending data cannot. The capacity of the bottom cohort to keep spending is being eroded right now, even if the current month’s card-swipe data has not fully reflected it.

This is what the Labor Fragility Index (LFI) is built to measure: z-scores of long-term unemployment share, the inverse of the quits rate, and the inverse of the hires-to-quits ratio. All three components moved the wrong way in the February print. The LFI is rising. The credit data has been screaming for a year. The labor data is now confirming it.

Which brings us to the cleanest signal in the pack.

The Credit-Labor Gap

We define the Credit-Labor Gap as z(HY OAS) - z(LFI). It answers a single question: does the credit market agree with the labor market about how fragile the economy is?

Right now, it does not.

High-yield option-adjusted spreads closed April 16 at 286 bps, through our 300 bps complacency threshold. Spreads tightened 116 bps over the past twelve months despite what just happened underneath. The Labor Fragility Index moved in the opposite direction over the same period.

When we compute the gap, the result is -1.68. Our threshold for “credit ignoring fundamentals” is -1.0. We are 68 basis points of z-score past that.

One of these two markets is wrong. Credit is pricing a soft landing that the labor data no longer supports. If the labor data is right, credit will reprice. If credit is right, the labor data has to reverse sharply, which means quits have to rise, long-term unemployment has to fall, and small-firm employment has to recover, all simultaneously and in the next print or two. We can see the path for credit to be right. We do not see the data.

This is the sibling signal to the Credit-Growth Gap we introduced on Friday. Same mechanic, different pair. Credit spreads have tightened into a macro environment the rest of the data is not confirming. Spread-widening when the correction comes will be faster than the tightening was.

The Spending Story

The macro cohort data flows directly into the consumer tape, and the consumer tape confirms the bifurcation.

Luxury retail (Tapestry, LVMH, Ferrari) is up 17% since January 2022 on an indexed basis. Dollar stores (Dollar General, Dollar Tree) are down 9% over the same period. The two lines have diverged monotonically. Ferrari just raised guidance. Dollar General just cut. Walmart is gaining wallet share from Target because middle-income consumers are trading down. LVMH’s US revenue is growing. Dollar Tree’s traffic is breaking lower.

The aggregate PCE number of +2.6% real year-over-year masks this. Real PCE at the top quintile is probably running +4% or higher. Real PCE at the bottom quintile is flat to negative. The average describes an economy that does not exist for most participants.

Under the PCE hood, the composition is already rolling. Durable goods spending is contracting at -1.8% YoY. Services are still growing at +2.4%. Durables turn first at cycle inflections because consumers defer big-ticket purchases before they cut groceries. You do not need a new car this quarter. You do need to eat. Durables negative while services positive is the classic late-cycle handoff, and the spread has now been widening for two quarters.

Confidence is the part of the picture that rhymes with the spending composition. UMich sentiment sits at 57.3, deep in weak territory but off its December low of 53. The Conference Board’s Expectations Index, which is the sharper recession signal historically, has collapsed to 65.1. Our threshold is 80. The CB Expectations Index dropping below 80 has preceded every recession since 1970 by 6 to 12 months. We are 15 points past it. Present conditions still look okay. Forward expectations are not.

The Inheritance Story

And now to Tim’s point, which is the most important one nobody has been properly measuring.

Cerulli estimates roughly $2 trillion a year is transferring between generations right now. That is not a future event. It is a present-tense income stream for a specific cohort of younger Americans.

Tim’s specific contribution, and it is a real one, is quantifying this channel’s impact on housing.

Per Redfin, up to 40% of first-time homebuyers used a gift or inheritance to make the down payment. Zelman puts the figure for all buyers at roughly 30%. Tim’s back-of-envelope math suggests down-payment assistance alone adds about 50 basis points to discretionary spending, because the income that would have gone to saving for the down payment now goes to consumption instead.

That is the channel James did not reach and Horwich flagged but could not measure. It is not just top-decile equity wealth funding consumption directly. It is also intergenerational transfer. Wealth migrating down a generation, landing disproportionately on Millennials and Gen X, and converting into home purchases and consumption that income-stratified data has no way to attribute to its source.

Tim’s framing is sharper than ours would be: “betting against the US consumer is like betting on the Jets to beat the Patriots. It almost never works.” He is right on the aggregate, and for the same reason James is right on the saving rate. The top half of the distribution is subsidizing the headline.

But his essay stops one step short of the full picture. Tim is describing the wealth-advantaged subset of the “bottom 60% by income,” the Millennials with parents who own a home and a 401(k). That cohort is real and growing. What his piece does not address is the rest of that bottom 60%. The cohort without wealth to inherit, that is funding consumption off credit instead of capital, and whose repo volume just matched post-GFC levels.

The two cohorts look identical on an income chart. They look very different on a balance sheet. The aggregate consumer data averages them together and calls it resilience.

The Housing Bridge

Housing mirrors the rest of the story and lets you see the two economies side by side in a single asset class. Luxury homes (top 10% of the price distribution) are up 2.4% year-over-year with an average 45 days on market. Entry-level homes are down 8.6% year-over-year with 95 days on market. The luxury tier is running 50 days faster through inventory than the bottom tier, and the price direction is opposite. These are not two ends of the same market. They are two markets.

The mechanism is straightforward. Top-tier buyers are cash-rich, often funding purchases from equity portfolios or inheritance transfers. Bottom-tier buyers are rate-sensitive and credit-constrained. The 30-year mortgage rate at 6.3% effectively locks out the bottom half of the first-time buyer market, while the top half does not feel it. The frozen equilibrium our Housing pillar describes is frozen more at the bottom than the top.

This is the channel that connects the K-shape to the real economy with the shortest lag. If the top-tier housing market rolls over, the wealth-effect channel James identified stops working. The saving rate does not just stabilize, it has to rise, because the top decile’s asset cushion stops compounding. And the aggregate PCE number that has been running +2.6% real falls hard, because the top quintile accounts for roughly 40% of total consumer spending.

Synthesis

James saw the equity wealth channel. Horwich saw the measurement gap. Tim saw the generational transfer channel. We would add the credit-distress data on the other side of the same K, and the labor-flow data that is now confirming it in real time.

The common feature across the first three: all are describing the same population. Wealth-advantaged households whose spending is decoupled from their current income. Whether that wealth arrives via equity gains, home appreciation, inheritance, or expected inheritance is secondary. The common feature is that current income is not the binding constraint.

Simultaneously, the bottom 60% of households by wealth are showing the credit distress you would expect in a mid-recession. Repo volume matching post-GFC levels. Student loan delinquency transitions seven times pre-pandemic levels. Quits rate through 2.0%. Long-term unemployment share through 22%. These households cannot quit spending entirely because rent and groceries are not optional, but they also have no slack left to absorb another shock.

Both things are the same thing. The aggregate looks “resilient” because the top half is spending a growing asset base and the bottom half is spending borrowed money it is increasingly unable to repay. Neither is a stable state, but they are stable together as long as asset prices keep rising and the credit system keeps extending.

The labor data just moved. Not decisively, but it moved. Quits through threshold. LT unemployed through threshold. Real aggregate payrolls near zero. The Credit-Labor Gap at -1.68.

One wildcard sits on top of all of this: the tariff pass-through. Businesses absorbed roughly 80% of tariff costs through 2025, which protected the CPI print but compressed corporate margins. That absorption is projected to shrink to roughly 20% as the effective tariff rate, now at 11.8% (Yale Budget Lab, April 6), works through pricing decisions later this year. The arithmetic flips from margin pressure to price pressure, and it lands disproportionately on the cohort with the least room to absorb it. A bottom-60% consumer already running a 1.2% saving rate and a 6.1% effective inflation rate does not have the balance sheet to eat a second leg of goods inflation. That is the scenario where Stage 2 goes straight into Stage 3 without waiting.

If the Fed had a labor mandate free of political crosswinds, this is the data print they would start cutting into. Whether they do is a separate question.

The Framework Behind The Charts

For readers new to the framework, this is Pillar 5 (Consumer) of our Diagnostic Dozen, integrated with Pillar 1 (Labor) and Pillar 9 (Financial). The full Consumer post is here.

The Consumer Conditions Index (CCI) is a seven-component composite. Each component is z-scored against its own history, then weighted by its measured contribution to forward PCE. The components are real PCE year-over-year, the personal saving rate, retail sales control, credit card delinquency (inverted), UMich sentiment, real disposable personal income year-over-year, and the household debt service ratio (inverted). Spending momentum and the saving rate do most of the work. Sentiment and the debt service ratio sit at the bottom of the stack because neither leads forward PCE cleanly enough to give them more room.

The Labor Fragility Index (LFI) is built from three inputs: the z-scored share of long-term unemployed, the inverse of the quits rate, and the inverse of the hires-to-quits ratio. The three are weighted roughly in balance, with a modest lean toward the flow measures.

The Credit-Labor Gap takes a definitional shape:

CLG = z(HY OAS) - z(LFI)

A difference of two z-scores, not an optimized composite. When the number goes materially negative, credit is pricing a softer labor market than the labor data describes.

Charts in this pack were built from the Lighthouse Master Database (~2,100 series, refreshed daily from FRED, BLS, BEA, NY Fed, OFR, Yahoo Finance, AAII, Zillow, and TradingView). Wealth-distribution data comes from the Fed’s Distributional Financial Accounts. Household credit data comes from the NY Fed’s Quarterly Household Debt and Credit Report. Subprime auto data comes from Fitch Ratings’ ABS tracker. Repossession data comes from Cox Automotive. Luxury and discount retail splits come from company filings and Yahoo Finance price data.

The component architecture is public. The exact weights, z-score windows, and signal-cleanup rules for every composite are proprietary to paid subscribers.

What Would Change Our Mind

This framework has clear invalidation criteria.

The wealth-effect channel dies when asset prices stop rising. If the S&P 500 breaks its 200-day moving average decisively, and Zillow’s top-tier home price index goes negative year-over-year, the top decile’s wealth cushion stops compounding. James’s nominal GDP math runs in reverse. This is a quarters-level event when it happens, not weeks.

The credit-distress channel inverts if small-firm employment turns up. If ADP shows small-business hiring recovering (not just stabilizing), the squeeze on the bottom 60% eases, delinquencies peak, and the two-economies frame loses half its force.

The transfer channel weakens if equity markets and housing both correct at the same time. Tim’s +50 bp discretionary spending contribution assumes the wealth being transferred is broadly stable. In a simultaneous equity-plus-housing drawdown, the transfer intensity falls.

The labor signal reverses if quits print back above 2.0% and LT unemployed share falls back below 22% in the next JOLTS and household survey releases. A single month’s move is not a trend. We need confirmation.

The Credit-Labor Gap closes if HY OAS widens meaningfully above 350 bps within the next 30 days. That would tell us credit has caught up to the labor signal, and the gap has closed the honest way.

If all three consumer channels unwind simultaneously, the “resilient consumer” becomes the “sudden consumer” in a hurry, because each of these channels is currently doing the work that weak real income growth is not.

The Bottom Line

The consumer cannot quit for the same reason risk assets cannot quit. Asset prices are doing the work that wages are not. Tim is right that betting against the aggregate is betting on the Jets. We would add: the game the Jets are actually playing is two different games at once, and only one of them is being scored.

The 4.0% saving rate is not a number about the American consumer. It is the arithmetic mean of a top decile saving 18% off a mountain of appreciating wealth, and a bottom 60% saving 1.2% because 1.2% is the floor. An average with that much underlying variance is not a data point. It is camouflage.

The consumer is the margin. Real PCE is growing at +2.6%, contributing roughly +1.4 percentage points to GDP. If consumer spending decelerates from +2.6% to +1.5%, the PCE contribution to GDP falls by roughly 0.7 points and GDP drops from +2.5% toward +2.0%. If consumer stagnates at +0.5%, GDP falls to +1.0%. The 68% of GDP that is consumer spending is the difference between a soft landing and a recession, and that 68% is already running on a funding mix that has shifted from income to credit.

What changed this week is that the labor data started to move. Quits below 2.0%. Long-term unemployment above 22%. Real aggregate payrolls near zero. Credit-Labor Gap at -1.68.

James, Tim, and Horwich have each moved the ball this spring. What we would add, and what the January piece was already pointing at, is that the headline is not hiding the distribution by accident. It is hiding it because we stopped publishing the distribution decades ago. The BEA does not release a saving rate by wealth decile. The BLS does not break spending out by net worth. We built the Diagnostic Dozen, and specifically the Consumer Conditions Index and the Labor Fragility Index, to put that distribution back into view.

The consumer does not predict. It validates. What it is about to validate is what labor, credit, and confidence have been saying for months. The two economies have been there the whole time. Now there are charts.

Get 36% off for 1 year

The S&P 500 closed April 15 at 7,022.96. A new all-time high. A ten percent rally off the late-March lows. The narrative is clean. Ceasefire, vol crushed, risk on, systematic funds re-leveraging into the bid. We’re seeing flows everywhere. We’re not seeing conviction.

The ceasefire was announced April 7. Two weeks. In principle. Iran has already re-closed the Strait of Hormuz once, after Israeli strikes on Lebanon. WTI sits near $92, well off the $119 war peak but still forty percent above the pre-war $67. Gold is holding near $4,800. The ten-year is 4.27%, roughly where it was before the ceasefire. Stocks have fully priced a clean exit. Bonds, gold, and oil have not.

That’s the first thing we notice. Equities are pricing resolution. Duration is still pricing uncertainty. Gold is still pricing a structurally higher inflation floor. One of these markets is wrong.

Credit agrees with equities, which is the uncomfortable part of the setup. High yield option-adjusted spreads closed April 14 at 284 basis points, through our 300 bps complacency threshold after peaking at 346 during the war. Spreads have tightened despite what’s underneath. Q4 2025 GDP was revised to 0.5% annualized. March CPI landed at 3.3% year-over-year, the highest print since April 2024. Real retail sales growth has halved from its early-2025 pace. February printed +0.8% YoY versus +2.5% last March, stalled in a +0.3% to +1.8% band for seven months now. Nominal spending keeps pushing higher. Real spending is barely breathing.

This is the Credit-Labor Gap thesis in a different asset. Call it the Credit-Growth Gap. Spreads are pricing a soft landing that the growth data hasn’t delivered. And the Fed is anchored. Fed funds futures show roughly eight basis points of cuts priced through year-end. No rate-cut cavalry is coming. The next CPI print is May 12. If energy passthrough broadens into goods and services, Powell has no cover to ease.

Here’s the contrarian twist worth sitting with. The AAII bull-bear spread for the week ending April 15 came in at minus 11.1%. Bulls 31.7%, bears 42.8%. The index is printing record highs and retail sentiment is more bearish than it was a week ago. Those two conditions almost never coexist. You usually get euphoria at tops. The wall of worry is still standing.

That ambiguity is why this tape deserves respect, not conviction. Trend and breadth have inflected. Price is above the 200-day. Market structure has repaired. Systematic funds are re-leveraging into a bid that is mechanical, not considered. But the sentiment read says the marginal buyer isn’t a believer. If the ceasefire extension fails, if the Strait closes again, if May CPI comes in hot, those flows reverse as fast as they arrived.

So what do we do with it. Positioning doesn’t change much from here. The Core Book stays with quality and defensives that held through the drawdown and haven’t given back their relative strength. The Technical Overlay can participate where trend, momentum, and relative strength align, but with tight stops, because the macro catalyst is a two-week announcement. Gold stays. Duration stays. We respect the tape without chasing it.

We’ll know we’re wrong if the ceasefire formalizes into a durable deal, if the Strait of Hormuz reopens on a sustained basis, and if oil trades back under $75. That path is the real all-clear, and on that path the framework will say add risk. We’re not there.

What we have right now is a relief valve, not a resolution. Stocks priced a ceasefire that isn’t finished. Bonds and gold are telling us to wait.

Get 36% off for 1 year

Bob Sheehan, CFA, CMT | Founder & Chief Investment Officer
Lighthouse Macro | Research | @LHMacro