‘There was an old woman who swallowed a fly..’ we read to our kids at bedtime. A cute, even silly story that winds its way through increasingly drastic steps to solve a problem, ending inevitably in
ad absurdum. We’re seeing this in policies right now. I hope the ending is better. To wit, one recent example: Iran ( and Ukraine) conflicts have led to a severe drop in diesel supply and a spike in prices. Truckers were massively negatively impacted. To help, the Administration suspended the restriction against truckers buying red dye diesel ( which, I just learned, means its gasoline-tax-free) . Good for truckers, bad for farmers, who are now facing competition for, and higher prices of, the tax-free fuel supply. The result? The recent announcement of an $8b assistance fund for farmers to offset the price pain. Secondary and tertiary impacts are underestimated and often ignored, but they play a major roleI’m trying to use less super sweetener in my coffee, even when I am watching football and soccer games being played on the latest Super Turf.
It sounds silly to just substitute a different word, no?
I have been reading and listening to thought pieces on AI and AGI, and the apparent sanguinity in the face of p(doom)—the probability, estimated by the heads of the major AI companies themselves, of AI destroying civilization—being higher than 10% baffles me. As is the push against federal or state regulation. The development of nuclear weapons led to the Atomic Energy Commission and the NRC. By comparison, would we be comfortable with no FAA or NTRB, with Delta and United self-inspecting? What about food companies self-policing the safety of their food product, or pharmaceutical companies regulating their own new drug trials and releases? Yet none of these hold nearly the same p(doom). Once again, money is talking, loudly.
**AI IS NOW A FINANCING STORY, AND SUPPLIERS ARE SOLD WITH BORROWERS.**I like Lam Research long against Oracle short. I’M WRONG IF high yield prints back near 290bp with the 10-year above 5.00%.**THE BARREL PRICES THE STRAIT AND GOLD FAILED AS A HAVEN.**I like Brent upside and gold short. I’M WRONG IF a Hormuz transit accord is signed, or gold closes above its 50-day near $4,358.**THE LONG END STILL CARRIES THE FISCAL BILL.**I like the 30-year from the short side. I’M WRONG IF the 10-year closes under 5.00%.**CREDIT IS MY HEAVIEST SCENARIO AND MY BOOK IS NET SHORT IT.**That is the concentration problem this morning. I’M WRONG IF CCC paper rallies back through 1,100bp.
Four companies beat last night, and all four were sold this morning: Meta, Tesla, IBM and Lam Research. The Kospi fell 3.52% and Nasdaq 100 futures 1.8%.
Did investors stop believing in AI demand overnight? I don’t think so. I think they started charging for the money.
THE SETUP.
The fixes keep generating the next problem, and each one is financed:
The hyperscalers fund the build. Meta raised 2026 capital spending to $135 to $145 billion and booked a $9.89 billion litigation charge in a quarter where revenue beat at $64.6 billion.
The builders without a hyperscaler’s cash flow borrow. AI companies have borrowed $466 billion this year to mid-October, against $101 billion at the same point of 2025 (Morgan Stanley, FT).
The lenders then buy insurance. Weekly protection traded on seven AI names ran above $9 billion in September, 4.5 times July. CoreWeave costs $1.09 million a year to insure $10 million of debt; Oracle’s protection has more than doubled this year.
The tokens are circular too. Among the top 1% of AI spenders, 80.4% are venture-backed (Ramp, Apollo), so a slice of the revenue the suppliers book is investor money passing through.
JPMorgan’s flows team puts the financing gap, after buybacks and dividends, at about $500 billion this year and $696 billion in 2027, split between bonds and loans, SPVs and private credit. That debt prices off a 5.13% 10-year. Tesla cut its capital spending outlook to about $20 billion from $25 billion, and Lam still guided 2027 equipment spending above $200 billion. The supplier is selling with the borrower.
THE EVIDENCE.
Credit moved first. High yield printed 333bp for 20 October, 4bp wider, on a day the Russell 2000 gained. CCC paper is at 1,310bp, a 2026 high, while BB is near 200bp.
Gold failed. The metal fell 1.2% to $4,102 at 08:53 ET, with seven tankers reported damaged in Hormuz (AP via
Irish News), while Brent was up 2.9%. Silver fell 2.9%.FLIGHT CHECK: gold offered, the 10-year bid (5.08% indicated against a 5.13% close), the yen bid at 157.03, the dollar index offered at 101.53. Mixed havens with risk down reads as a rotation into duration; a de-gross would have taken Treasuries down too.
Harris Kupperman wrote his Q3 letter around this: funding gets cut first, the neoclouds go bankrupt next, then the frontier labs, “slowly, then all at once,” and he owns SOFR futures and call spreads ahead of what he calls ‘Project Zimbabwe’. Treasury Secretary Bessent told a charity dinner he sees no “systemic problem” in the boom (FT). One of them is badly wrong on timing, and my book leans toward the letter.
Who polices the build if Washington won’t? For now, the lenders do, through the price of protection, and they charge CoreWeave more than ten times what they charge Nvidia ($1.09 million a year per $10 million, versus $101,000). That is a regulator of sorts, and it moves faster than Congress.
THE CASE AGAINST.
Dario Perkins at TS Lombard wargamed the crash this week, and his version is the best argument against my timing:
At least a 30% fall in US stocks, about $20 trillion of market cap, and $125 to $200 billion of credit losses, with the neoclouds, Oracle and the standalone model labs in acute distress.
Only a mild recession, with today’s earnings still mostly ‘capex recycling’ in his words, because the Fed has room to cut and long yields fall in a demand shock. His working date for the end of the capex cycle is early 2028, and he would add bonds only once the strain shows.
Unless the funding market closes on a specific borrower this quarter, I am short the financed builder two years early. The overnight selloff could just as easily be a crowded long (a record $7 billion of retail money went into SOXL) shaking out on four headlines.
The red-team version is a ‘stagflationary repricing’ at 40%, oil and real yields up together, which hurts my credit view more than my long-end short.
I’d put the financing read at 60/40 against a positioning flush, and I’m less sure of the timing than of the direction.
THE BARREL.
US crude inventories, including the SPR, are near 700 million barrels, the lowest since 1983. Commercial crude fell 3.06 million barrels last week against 0.82 million expected, and gasoline fell 2.95 million.
Freight from the US Gulf to Asia runs near $38 a barrel, up from $3 to $6 before the war, so the delivered barrel costs more than the screen shows.
Brent touched $112.68 overnight, and the first settle above $110 still hasn’t printed.
Simon Hunt reads Lavrov’s Tehran visit as a refusal to drop Iran (”sanctions won’t stop us”) and expects the Houthis to move from Saudi oil infrastructure to the kingdom’s southern border. December Brent and WTI both made new highs, and one chartist I read has $120 on both.
Distillates built 1.68 million barrels, which is the one number in the report that argues against me.
The second-round effects are already in the food chain. Thai rice export prices are up about 37% year over year (Nomura), and tariffs have pushed non-oil goods inflation near 2% where it would otherwise be negative. Each relief program (red-dyed diesel for truckers, $8 billion for farmers) lands on top of that.
Claims came in at 194,000 against 200,000 expected (
Bloomberg), a three-month low, keeping the December hike path alive and keeping the front end from following the 10-year lower. Blackstone’s third-quarter distributable earnings fell 9.2% year over year on slower dealmaking (Bloomberg), and Huntington and Sallie Mae report this morning.**THE CONSUMER.**Michigan sentiment is at or near the lowest reading in its 70-plus years, and the two questions I care about most are both at records. The final October read lands Friday at 10:00.
The preliminary headline is 46.3, against 48.1 in September, 53.6 a year ago and a record low of 44.8 set in May, about 45% under the long-run mean of 83.6, and lower than at the start of any of the last six recessions.
Current conditions are 44.7, the lowest on record. Expectations edged up to 47.3.
Inflation expectations rose for a second month: 4.7% a year out and 3.5% long run, both the highest since May.
Business conditions against a year ago: 10% better and 68% worse in August (index 42), after 75% worse in April. Grim, and still nowhere near November 2008, when 96% said worse (index 6).
Faith in government economic policy is the record: 12% good, 14% fair, 74% poor in August (index 38), and May’s 77% poor (index 34) is the worst in the series. By party, on three-month averages, Republicans are at 119, independents at 33 and Democrats at 8.
My read, with moderate confidence: independents decide which way the headline moves. Michigan said as much about October, where gains among Democrats and Republicans were canceled out by a drop among independents.
**THE POLLS.**The president’s approval is at the low of either term, and his approval on the economy runs 4 to 10 points under the overall number:
The RealClearPolitics averages are 38.3/60.5 overall and 32.5/64.4 on the economy, about 10 points net worse on the economy. Nate Silver’s average hit a new low of minus 24.3 net on 21 October.
NYT/Siena has 69% disapproving of the cost of living. Gallup rates the CDC good or excellent at a record-low 26% (
Axios).Samples differ: adult samples run 2 to 3 points worse for the president than registered or likely voters, which is the 31-to-34 split inside the Reuters/Ipsos poll.
Why does a sentiment record belong in a market letter? Because the spending hasn’t followed it down.
Empirical Research puts the equity wealth effect at about 125bp of consumer spending growth over the past year, against a 45bp average, so the top decile is carrying the bottom 60%, and a crack in the AI trade takes that away faster than lower rates give it back.
An electorate this grumpy by 3 November is a fiscal story, and I’d see it first in the long end. Goldman’s social-media gauge and the Augur hourly index agree with Michigan.
Housing is the transmission I can see. Pending home sales fell 7.3% in September to 66.0, the lowest since the series began in 2001, with all four regions down, and builder sentiment at 29 runs under every starts reading since 2012. The 30-year mortgage rate eased to 7.41% from 7.50% (MBA), and purchase applications still fell 7.3%.
Two movers pointed opposite ways: SpaceX was paid for AI revenue and Tesla was punished for cutting the spend that buys it.
**SPACEX AND TESLA.**SpaceX grew revenue 161% to $14.6 billion, and AI is now 52% of it, ahead of connectivity at 36%. Musk called a merger with Tesla “a good idea” the same day Tesla took its capex outlook down a fifth and slid Optimus. A combined company would carry the cash flow and the spending in one balance sheet; SpaceX’s five-year protection is at a record $226,000 per $10 million (FT).
**LAM RESEARCH.**Beat, guided 2027 wafer equipment above $200 billion, indicated 8.8% lower at $237.79, 46% under its high; Applied Materials fell about 5% in sympathy with no news of its own. I like it from the long side against Oracle, which has taken less of the neocloud markdown (CoreWeave and Nebius are down about 35% in two weeks).LOCKHEED MARTIN.$7.97 against $7.52 on $20.7 billion of revenue, into an eighth month of war, with Saudi airfields and an Aramco field under Houthi fire. I’m not long defense; my energy line already owns the escalation.
Behind the wall this morning:
THE BOOK: 15 live lines, 3 new today, each with the factor it monetizes, its size and the condition that kills it.
FLOWS AND POSITIONING: who funds the basis trade, what a nonbank fire sale does to bank capital, and where the CTAs flip on small caps.
FINANCIALS: buybacks paused on the long end, a regional resuming them anyway, and the two financials trades I want on today.
Everything above this line is free and always will be. Below it is where my own capital goes: $450 a year at marketmusing.com, less than one bad lunch a month.
Disclaimer: Opinions are mine alone and may change. It does not constitute an offer to buy or sell or a solicitation of an offer to buy or sell any security, loan or asset or to participate in any trading strategy. It is not intended to form the basis of any investment decision, should not be considered a recommendation, and does not constitute an offer or solicitation with respect to the purchase or sale of any investment, nor is it a confirmation of such terms.
