Nobody’s Promise
morning musings 7.22.26-views from the island
Morning Musings | Wednesday, 22 July 2026 | morningmusing.com
“The truth is rarely pure and never simple.” Oscar Wilde
Three tornadoes came through here yesterday. Morris, Warren, Sussex. Hail, flash flooding, towns telling people to get to the lowest floor of the house.
The National Weather Service will not call a touchdown without a ground survey, and a survey takes days. So the official count this morning is two.
The number is three.
Brent printed $95.47 at the open. Then it printed $93.92. Same problem, different building. The war trade went on and came straight back off inside a few hours, while an eleventh consecutive night of American airstrikes was still being written up. Somebody paid ninety-five. Somebody else took the other side and took it hard. The touchdown is on the tape. Whether it gets confirmed is a separate question, and nobody answers that one until the settlement.
Four research products hit my desk this morning quoting Brent above ninety-five and calling the supply premium paid. Every one of them pulled the number at the top of the wick. Four sources agreeing on the same instant is not four sources. It is one source, quoted four times, and I very nearly published it.
Here is the part that should bother you more than the fade. Gold is sitting at eighty percent of its daily range. Brent is at sixty-three. In a supply shock the barrel leads and the metal follows, because the metal is hedging the second-order inflation and not the first-order barrel. Today the metal is leading.
That is not a war bid. That is a currency bid wearing a war costume.
Three tornadoes touched down in New Jersey yesterday. Four havens got tested this morning, and exactly one of them showed up.
TL;DR
The barrel printed the headline and refused to pay for it. Brent hit $95.47 and sits at $93.92, sixty-three percent of the day’s range and falling (FMP). Gold is at eighty percent of its range. In a war tape oil leads and gold follows. Today it is inverted, which means the bid is monetary, not geopolitical. One haven is working out of four, and it is the only one that is not somebody’s promise. The 10-year closed at 4.63 percent, the highest since 19 May, and every brief in my inbox calls it ‘sticky.’ Tonight’s Alphabet print is being traded as a revenue question. It is a funding question, and a capex raise is the bear case, not the bull case. Standing down the energy arm despite the trigger printing. Duration short stays on.
MY READ
Three things, and I hold all three against the crowd.
One. The oil trade is being sold, not bought. Brent traded $95.47 overnight, the P3 arm level, and is now $93.92 (FMP). Day low $91.33. It gave back thirty-seven percent of its own range inside a session on the eleventh consecutive night of American airstrikes, with Hormuz tanker crossings at ONE (BBG chart grid). One. Cargo ships at three. The seven-day tanker count is fourteen against a pre-war run rate north of four hundred.
Read that again. The physical waterway is effectively shut, the barrel is up three percent, and the market spent the morning selling it. That is not a market pricing a supply shock. That is a market that has decided this war is a headline generator and nothing more, and has been rewarded for that view two sessions running.
I think the market is right about today and wrong about the structure. But I am not paid to be right about the structure at a price I did not get, so the energy arm stands down. More on that behind the wall.
Two. Gold is the tell, and it is not the tell everybody thinks. Gold $4,132.80, up 1.38 percent, sitting at eighty percent of its daily range (FMP). Brent at sixty-three percent. The metal is outperforming the barrel intraday.
That ordering matters enormously. In a genuine geopolitical supply event, crude leads and bullion follows, because bullion is hedging the second-order inflation, not the first-order barrel. Invert the ordering and you are no longer looking at a war trade. You are looking at a debasement trade wearing a war costume.
Confirm it with the haven table. Gold bid. The 10-year completely inert at 4.63 on a day crude printed a six-week high. The dollar inert at 101.15 (BBG). The yen made a fresh four-decade low at 163.228 and needed a leaked BOJ story to stop the bleeding (BBG).
One haven bid out of four, and it is the only one that is not a liability of a government or a central bank. Treasuries are a promise from a Treasury running the deficit that produced a 5.13 percent thirty-year. The yen is a promise from a central bank that has spent four years explaining why 1 percent is enough. The dollar is a promise. Gold is nobody’s promise, and gold is the only thing catching a bid.
And while I am here: the file on my desk this morning called $4,125 gold ‘a fresh high’ and built a whole miner trade on ‘gold at a record while the seniors lag their January peak.’ Gold’s fifty-two-week high is $5,626.80 (FMP). Spot is twenty-six-point-six percent below it, below its fifty-day at $4,235.59, ten percent below its two-hundred-day at $4,594.03. Silver is at $59.73 against a fifty-two-week high of $121.30. Half. The metal is the thing lagging January. The trade may still work. The reason given for it was inverted, and an inverted reason is how you end up adding to a loser with total confidence.
Three, and this is the one I actually want to own: tonight is not a revenue question.
Everyone is watching Alphabet’s cloud line to see whether AI revenue is outrunning AI capex. Consensus is EPS $2.90 on revenue of $117.0bn with capex around $44.2bn, up from $35.7bn in Q1 (BBG).
Wrong line. Watch the funding.
BofA has hyperscaler capex at $412bn in 2025, $769bn in 2026, $967bn in 2027
(BofA/Hartnett)Aggregate hyperscaler free cash flow over the same window: $191bn, then minus $19bn, then minus $26bn
Slok has the cover ratio on hyperscaler bond issuance down from roughly five times in February to below two times in July
(Apollo)Nikkei puts off-balance-sheet obligations at the five big spenders at $1.65 trillion, exceeding the $1.35 trillion actually on the balance sheets. Meta alone at roughly $420bn, nearly triple its recorded debt
(Nikkei)
Three houses. Three different instruments. One mechanism. The AI buildout crosses from self-funded to debt-funded inside eighteen months, and the marginal buyer of that paper has already halved his bid.
Which flips the sign on tonight. Consensus logic runs ‘capex up means demand strong means buy the complex.’ My logic runs: capex up means the funding gap arrives sooner, into a credit market that started backing away five months ago while everybody was reading the revenue line.
A capex raise tonight is bearish. A capex cut is the bullish print. That is my variant and I have not seen it written anywhere.
Netscape was right about the internet. Every single thing Andreessen said about the web was correct, and the equity still went to zero, because the multiple did the damage, not the invention. The multiple is downstream of the funding. The funding is what is changing, quietly, in a place nobody is looking.
WHAT MATTERS TODAY
The 10-year is not sticky. 4.63 percent at Monday’s close, the highest print since 19 May (FMP Treasury). It was 4.38 on 29 June. That is twenty-five basis points in three weeks. The 2-year is 4.26, tied for its highest reading of the quarter. 2s10s +37bp. 2s30s +87bp against +76bp on 29 June, so the long end has steepened eleven basis points against the front while the whole curve shifted higher.
Every brief I read this morning describes this as ‘sticky near 4.60’ or ‘pinned.’ Sticky is what you call a number that is not moving. This one has moved every session since Friday and is at a two-month high on a day nobody mentioned it.
GOOGL, TSLA, IBM, TXN after the close. Capex line. Not the revenue line.
EIA crude inventories, 10:30. Consensus a 1.5 million barrel draw (technical desk). This is the only scheduled item that can settle whether the overnight fade was real.
Trump at Dover, 11:00, for the dignified transfer of the soldiers killed this weekend. Then Marietta, Georgia, for a high school rally. Trade Representative Greer sits before Senate Finance at 10:00 to explain a tariff schedule that changes on Friday.
Friday: the ten percent blanket global tariff expires and is set to be replaced with permanent duties. A twenty-five percent Brazil tariff went live today. A hundred percent generic drug tariff is being floated for 2028, doubling in 2029.
ECB tomorrow, 92 percent priced for a hold at 2.25, with September increasingly live (BBG).
MOVERS THAT MATTER
SMCI, up roughly fifteen to twenty percent on a record $60bn backlog and a gross margin guide doubled to 15-17 percent from 8.2-8.4 percent. This is the print every bull will cite tonight, and it is the wrong evidence. A backlog is an order book, not a cash flow. And a hardware assembler doubling its gross margin is not a demand signal, it is a scarcity rent, which is a tax the buyer pays. SMCI’s margin expansion is precisely the line item that pushes hyperscaler free cash flow to minus $19bn. So what: the bulls are quoting the supplier’s income statement as proof the customer can afford it.
Airbus +7 percent on a 2029 EBIT target of €12-13bn from €7.1bn in 2025, plus a €5bn buyback. So a European industrial just guided to a near-doubling of profit off a capital cycle with a visible order book, funded from cash, and it trades at a fraction of the AI complex’s multiple. So what: this is what a capex cycle with returns actually looks like on a page. Hold it up next to a hyperscaler guiding capex to $769bn against negative free cash flow and ask which one is the growth story.
Tencent minus 7.1 percent, worst session in over a year, dragging Hang Seng Tech down three percent. Beijing has mobilised state funds, regulators, insurers and asset managers to stop an AI and chip selloff ahead of the CXMT listing. Chinese quant fund NAVs are down more than twenty percent at some managers on forced deleveraging. Vietnam fell 3.6 percent on margin calls, now thirteen percent off its peak (BBG). So what: the leverage in this cycle is Asian and retail, not Western and institutional, and it is already unwinding. Korean CFD balances have tripled in two years at up to 2.5x (BBG via Macro Charts). That is the transmission channel, and nobody in the US is marking it.
A note on discipline: I had a sector rotation table for Tuesday that showed Technology down 1.06 percent on a session the Nasdaq 100 rose 1.93 percent. Those cannot both be true. Rather than pick the one I liked, I cut both. Same reason I am not quoting a KOSPI number today: I have one source saying plus 0.7 percent and another saying plus six percent on a short squeeze, and the difference decides whether the semi bounce was demand or forced covering. I do not know, so I am not going to pretend.
FINANCIALS
The group’s whole argument this morning lives in two numbers that are twelve and twenty-two basis points away.
The curve. 2s10s +37.3bp. 10Y 4.63, 2Y 4.26, 30Y 5.13, and the 20-year at 5.14 trading above the thirty, which is a long-end distortion that has been sitting there for months and which nobody wants to explain (FMP Treasury, 21 Jul).
Now the two levels. The AOCI threshold where available-for-sale marks start biting the regional cohort sits around 4.75 on the ten-year. We are twelve basis points away. The Financials Daily’s own stated invalidation for the whole re-rating thesis is a sustained 10Y above 4.85. Twenty-two basis points away. And the KRE-versus-QQQ relative-value trigger on the tactical sheet needs 2s10s at +60bp. Twenty-three basis points away, in the other direction.
So: the regional bank thesis needs a steeper curve, and what it is getting is a higher curve. Those are not the same trade. A ten-year at 4.80 driven by term premium hands the group a book value problem, not a net interest margin gift. A ten-year at 4.80 driven by growth would be the gift. The 2-year at 4.26 with nine of eighteen Fed officials projecting a hike this year (Apollo) tells you which one is happening.
Where the dispersion actually is:
IB and brokerage, the long. Goldman printed a record Q2: revenue $20.34bn, EPS $20.98, investment banking fees $3.40bn up 55 percent, equity underwriting up 130 percent, north of a trillion dollars of M&A year to date. Then the upgrade cluster, Jefferies to $1,299, Wells to $1,325 (Financials Daily). None of that revenue has a duration. Advisory and underwriting fees do not reprice off the ten-year. In a tape where the discount rate is the problem, own the fee line.
Regionals, the short. KRE has been running near fifty-two week highs on a net interest margin story that is thoroughly well owned. Deposit costs are easing, RF at 3.66 guiding toward 3.70. But office CRE is still roughly twenty-four percent criticised at MTB, and the group is priced for a steepening that is not arriving. Entry near the highs on a thesis whose own falsifier is twelve basis points away is not a risk-reward, it is a hope.
BDCs and private credit, the other short, and the one that matters. Non-accruals jumped roughly forty percent sequentially in Q1 to 2.01 percent at cost, with Octus adjusting to 3.24. BCRED cut its July distribution to $0.18, the second cut in nine months. NAV $23.94, down 3.4 percent year to date. It returned about $3.7bn in Q1, which is 7.9 percent of assets against a five percent gate (Financials Daily). Ares Strategic Income and BCRED both breached gates earlier this year.
Set that against index credit: high yield option-adjusted spread to sovereign at 266.6bp, investment grade at 77.1, and US high yield is the only bond index positive month to date at plus 0.06 percent while the US Aggregate is minus 0.88 and the Euro Aggregate minus 1.45 (BBG). Credit did not widen a basis point on a six-week high in crude.
Everybody reads that as ‘credit is calm, therefore fine.’ I read it as the location of the problem. Index credit is calm because the stress is not in the index. It is in the retail-perpetual vehicles, where the gate is the price discovery mechanism and the gate is already being used. You do not get a warning from the OAS. You get it from a distribution cut, and you have had two in nine months.
The consumer leg, which is about to get interesting. Card net charge-offs around 3.8 percent, COF at 4.5 and SYF at 4.8 on delinquency, and buy-now-pay-later late payments at forty-seven percent of users, up from forty-one in 2025 and thirty-four in 2024 (Financials Daily). BNPL late-pay leads bank card losses by four to six weeks.
Now add today’s inputs. Retail gasoline is back to $4.06 and diesel to $5.18 (AAA). J.P. Morgan is tracking a contraction in July retail sales, headline and control both negative (JPM Research). Evercore was writing four days ago that the driving public was starting to enjoy sub-$4 gasoline. It lasted about a week.
The soft-landing consumer is the assumption in this sector with the freshest disconfirming print and the least attention on it.
And the politics, since we are in sharp mode. The House passed the Main Street Capital Access Act 270-155, a twenty-five bill package raising the asset thresholds that trigger regulatory scrutiny and expediting merger reviews. Basel endgame’s easing bias is intact and the capital relief is real. So Congress is handing the banking system capital relief and faster merger approval in the same quarter that private credit non-accruals jump forty percent and a flagship retail perpetual vehicle gates its investors for the second time.
Deregulating into the front end of a credit cycle does not prevent the problem. It enlarges the balance sheet the problem eventually lands on. We have run this experiment. Twice.
So what, for positioning: long the fee cycle, short the rate cycle, short the retail-perpetual credit complex, and do not express any of it as XLF. Sector beta is not the idea. The dispersion is the idea.
FLOWS & POSITIONING
The crowding is not where you think it is. UBS has hedge funds cutting momentum and semiconductor longs by more than five percent of gross market value, taking net exposure back toward April levels (UBS via technical desk). Institutional positioning in the AI complex is already reduced.
Against that: margin debt at a record $1.53 trillion, up 51 percent year over year. I will flag that this figure appears in three of my own documents and one underlying source, so treat it as one data point repeated, not three confirmations.
So the leverage is retail and Asian, and the institutions have already moved. That changes the shape of a bad print. It does not produce an orderly institutional de-gross into waiting bids. It produces margin calls in Seoul and Ho Chi Minh City at two-thirty in the morning New York time, which is roughly what Vietnam did yesterday.
The vol setup is worse than it looks. VIX 17.61 (FMP). Asset managers carrying significant short vol exposure (Macro Charts). Nobody is paying for protection into the single largest binary of the quarter, on the day the physical Strait of Hormuz is running one tanker.
Gamma. The S&P’s positive gamma magnet sits at 7,510 with about $12.9bn, which is why the index has been welded to that number. Below it, 7,450 carries minus $2.9bn. The NDX has no comparable stabiliser near spot; its positive gamma support is all the way down at 28,550, and 28,000 is negative (technical desk). Translation: the S&P is pinned and the Nasdaq is not. Downside there is faster than people expect because there is nothing under it.
Breadth. SPY at 98.4 percent of its fifty-two week high, on 223 advancers against 262 decliners, 57.6 percent of members above the fifty-day, and net new highs of plus five. The index is at the top of its range and the participation is not (technical desk).
Concentration. The largest twenty-four names are now more than half of S&P market capitalisation, below the pre-bust low of roughly thirty-two in 2000 (Simon White/Daily Chartbook). Twenty-four stocks, and four of them report inside the next eight days.
Correlation. Variant Perception’s leading indicator for stock-bond correlation has turned higher again, pointing to positive expected correlation ahead (Variant Perception). Bonds are not going to cushion the equity drawdown. That is not a forecast, it is the whole reason the duration short and the equity hedge are not offsetting each other.
Not used: a roughly $6.5bn market-on-close imbalance is circulating this morning with the direction unverified. It has already been promoted to a ‘sell imbalance’ in one product I read. It stays in the flagged column until somebody shows me the sign.
THE BRIDGE
Behind the wall this morning: why I am standing down an energy trigger that actually printed, and the exact level that re-arms it. The full book marked to live with distance-to-trigger on every leg, including the seventeen basis points that fire the duration add and the twenty-three that stop it out. Two financials expressions built on subsector dispersion rather than sector beta, with the netting problem that kills one of the ideas on my desk. The revised activation condition for the AI basket, which now requires a capex guide down rather than a beat, and why the consensus trigger is backwards. A four-branch decision tree for 16:30 tonight. Conviction ranking, and the watchlist including the two ideas whose premises I had to invert before they were usable.
morningmusing.com
Keep reading with a 7-day free trial
Subscribe to Morning Musings- Thoughts from Dauber Island to keep reading this post and get 7 days of free access to the full post archives.