Friday, October 23, 2026, written at 9:10 a.m. Eastern, twenty minutes before the open. Every future, yield, currency and commodity level in this note carries that stamp, and the European and Asian figures are intraday or local closes rather than settled American marks. There is no PCE today. That print slipped to Thursday October 29, the day after the Fed decides, which means the committee goes into next week’s meeting without its preferred inflation gauge. What we get today is the S&P Global flash composite at 9:45 and final Michigan sentiment at 10:00.
Top Call: The AI Complex Lost Three Funding Channels in a Day
In roughly 24 hours, the artificial intelligence build lost access to all three of the markets that pay for it. Credit: CoreWeave’s five year default swap widened as much as 74.5 basis points on Thursday to a record 1,194.5, which implies better than a 60% chance of default inside five years, while Meta’s own swaps printed a record alongside Amazon, Alphabet and Nvidia. Private equity: OpenAI cut the valuation it is asking from new investors to $1.25tn from the $1.4tn it pitched three weeks ago, a $150bn haircut on a $30bn round it wants signed by the end of next week. Public equity: SK Hynix formally ended its review of a United States listing for Solidigm, a deal reported as recently as last week at up to a $150bn valuation and $10bn to $15bn of proceeds.
Three different instruments, three different jurisdictions, one message. For two weeks this desk has argued the binding constraint in this cycle is the price of capital rather than the level of demand. Yesterday Meta answered that by building an equity funding vehicle. Today the answer is blunter: in the space of a single session, the swap market repriced the weakest borrower in the complex toward distress, the strongest private company in it accepted a 11% markdown to get a round done, and the most valuable subsidiary in Korean memory gave up on going public altogether.
Our read is that this is the week the funding question stopped being an argument about disclosure and became a set of prices. A rating agency saying on Monday that it cannot see all the debt is a caution. A 1,194 basis point swap is a quote you can hit.
The Credit Print Is the One to Carry
If you keep one number from this note, keep the CoreWeave spread. It is the only instrument in the complex that marks the thing everybody is arguing about, which is whether leveraged compute capacity financed on depreciating collateral against contracts that pay in the 2030s is money good.
The composition matters as much as the level. Nvidia and Meta sit near 90 to 108 basis points. CoreWeave sits above 1,100. That is not a sector being repriced uniformly, it is a capital structure being sorted by who can refinance without asking permission. The hyperscalers are being charged a premium. The neoclouds are being charged a probability. That distinction is the whole reason we have separated the demand case, which keeps printing fine, from the funding case, which keeps getting worse.
There is a counterargument we should state rather than bury. A single name’s swap is a thin, dealer quoted market, and a 74 basis point move in one session on a company with roughly $2.6bn of exposure in question can be positioning as easily as information. CoreWeave has not missed a payment, its contracted backlog is real, and swap markets have been wrong about going concerns before. What makes us weight it anyway is that it did not move alone. The Zenith Arc and Galaxy Helios data centre bonds led declines, CoreWeave’s unsecured notes traded into distressed territory, and the broader risk off reading was the largest in more than a year. When the quote, the cash bonds and the adjacent paper all move the same way in the same week, the thin market excuse gets harder to use.
Western Digital: The Test Resolves Yes, and It Did Not Pay
We pre-committed yesterday that Western Digital would be the cleanest remaining read on whether the physical shortage argument we have leaned on since October 13 is still intact, and that the company needed to guide fiscal second quarter revenue above consensus with capacity committed beyond calendar 2026. It did that, and then some.
| Metric | Reported | Against expectations |
|---|---|---|
| Revenue | $4.38bn, up about 55% y/y | Above |
| Adjusted EPS | $4.64 | Beat the $4.11 consensus by $0.53 |
| Non-GAAP gross margin | 58.1%, operating margin 49.1% | Expanded sharply |
| Cloud mix | 88% of revenue | A record for the business |
| FQ2 revenue guide | $5.1bn midpoint, up about 69% y/y | Well above |
| FQ2 margin and EPS guide | Gross margin 60% to 61%, EPS $5.80 plus or minus $0.20 | Well above |
Our take. On the stated terms the test resolves yes, emphatically. Management went further than a guide. The CFO said capital spending will run modestly above the normal 4% to 6% of revenue band for the next two years, and specified that the money goes into HAMR heads and media technology and factory automation rather than into new capacity. That is a company telling you in public that it intends to keep supply tight and harvest price, which is the supply shortage thesis stated by the people who control the supply. The 40 terabyte ePMR platform moves to roughly 30% of nearline shipments next quarter and about 50% the quarter after, with first generation HAMR qualifications at two hyperscalers due in calendar Q1 2027.
And the stock fell 6.4% in the session and another 4.9% after hours, which is where our own framing deserves criticism rather than a victory lap. We designed a test to find out whether the shortage was real. It is. We did not design a test to find out whether that still matters to a share price, and it is now reasonably clear that it does not. A company printing 58% gross margins into a sold out market, guiding up 69%, cannot hold a bid in a tape where the question has moved from who needs the drives to who is paying for the building they sit in. Anyone who traded our supply thesis this week was right about the world and wrong about the market.
The Tape: Rotation, Not Liquidation
Thursday was the worst day for the Nasdaq 100 since June, down almost 3%, and the equal weight S&P 500 rose 0.8% on the same session. That one pair of numbers is the entire market right now. Consumer staples are up roughly 12% in October and on track for their best month since October 1998. Money is not leaving equities. It is leaving the part of equities that has a financing problem attached.
The bond market is telling a compatible story with an awkward wrinkle. The ten year fell about 11 basis points on Thursday to roughly 5.02% and traded below 5% for the first time in nearly a month as investors bought duration against the tech selloff. This morning it has given four basis points back to about 4.99%, with the two year at 4.66% and the thirty year near 5.30%. Meanwhile December hike odds have drifted from about 86% at Thursday noon to roughly 75% to 78%, and the October 28 meeting is priced at only a 5% to 16% chance of a move, down from 27.7% a week ago.
Here is the wrinkle, and it is worth sitting with. Brent is at a five month high and hike odds are falling. Normally $115 crude pushes a hiking central bank further along, not less. The resolution we favour is that the market has decided the equity drawdown in the largest capital spending programme in corporate history is itself a tightening of financial conditions, and that it does part of the Fed’s job without a policy move. If that read is right, every further leg down in AI credit buys a little more dovishness. If it is wrong, and the committee looks at $115 Brent and a 3.4% headline PCE from August and decides the inflation problem is the only problem, then the December repricing has further to run against a market that has been buying duration as a hedge.
Earnings: Fine Quarters, Hostile Tape
| Company | Result | Reaction | Read |
|---|---|---|---|
| Western Digital | $4.64 vs $4.11, FQ2 guided far above | Down 6.4%, then 4.9% after hours | Shortage confirmed, nobody paid for it |
| American Express | $4.68 adj vs about $4.51, revenue $20.06bn vs $20.12bn | Firm | FY EPS guide raised, provisions fell |
| Kering | Gucci down 6% comparable, group revenue flat | Down about 10% to 12% | China and Japan traffic, not a turnaround yet |
| RingCentral | Q3 beat, full year adjusted EPS raised | Up about 16% | Small cap software still clears a low bar |
| Deckers | Full year sales guidance cut | Down about 7% | Second consumer discretionary warning this week |
| Volvo Group | SEK 5.29 vs SEK 5.54 expected | Lower | Margin up, earnings light, industrial demand soft |
American Express is the quarter worth holding against the macro. Earnings per share of $5.42 reported, $4.68 adjusted after stripping the Global Business Travel stake sale, against roughly $4.51 expected. Billed business rose 8% to $454.0bn. The line that matters is the provision for credit losses falling to $1.0bn from $1.3bn a year ago, and the full year adjusted guide going up to $17.60 to $17.90 while the revenue guide held near 10%. That is the premium American consumer still spending and still paying, in a week where Deckers cut and the equity market is pricing a capital spending recession. The two coexist, and the reconciliation is that this cycle’s stress sits in corporate balance sheets financing compute, not in household ones.
Kering is the other direction. Gucci sales of EUR 1.29bn were down 6% on a comparable basis against roughly 5.4% expected, with own store sales down 9% on weaker traffic in mainland China and Japan while North America held up best. The group is sticking with the de Meo plan and accelerating closures toward 125 by year end. Shares are down about 10% today for a 12 month loss near 43%. Luxury has now delivered three consecutive disappointments across LVMH, Hermès and Kering, which makes the Chinese consumption problem a thesis rather than a quarter.
Energy: Brent at a Five Month High on a Claim Nobody Can Verify
Brent trades $114.80, up about 1.4% and the highest since May, on track for a third consecutive weekly gain of roughly $6. WTI is $101.22. The move came after the Revolutionary Guard said seven vessels were hit or mined near the Strait of Hormuz, and after Houthi missile fire toward Riyadh. The Associated Press could not verify the IRGC claim and President Trump says Iran is inventing the incidents, which is the single most important qualifier in this section.
The supply cushion is what makes an unverifiable claim move the price $1.50. IEA data put observed inventories about 509 million barrels below end February levels, and effective OPEC+ spare capacity at 0.27 million barrels a day. At that spare level the market cannot absorb a real interruption, so it pays up for the possibility of one. Aramco has adapted, raising East-West pipeline throughput to 7 million barrels a day from 2 million, which routes around Hormuz but not around the Red Sea, and those flows were themselves halted on October 5 after a Houthi strike on a pumping station.
Set against that, the New York Times reports American and Israeli planners have prepared options for renewed strikes on Iran after the November 3 midterms, with no operation approved. Our split holds: a 2027 crude surplus underneath a present distillate and freight shortage, with the risk premium attached to transit rather than production. What has changed this week is that the premium now has a date on the other side of it.
Europe: France Waits for a Verdict It Has Already Been Given
Moody’s reviews France after the American close, around 2000 GMT, and a one notch cut to A1 from Aa3 with a negative outlook is widely expected. The market has done the work already. The ten year sits near 5.06% after touching 5.14% on Thursday, its highest since 2002, and the spread over Germany opened at a fresh post 2011 high of about 176 basis points.
The item that deserves more attention than the rating is the data. France’s October flash composite PMI fell to 48.4, with both services and manufacturing in contraction, against a eurozone composite still expanding at 51.9 and a United Kingdom composite rising to 53.1. A sovereign with a 5.4% deficit, a paralysed budget, a national day of action booked for November 5 and a contracting private sector is not a rating story, it is a growth and arithmetic story. A downgrade tonight changes the label on something the bond market priced two weeks ago. What would actually matter is an outlook that signals further cuts, because that is the line that forces index and mandate driven selling rather than discretionary selling.
Equities are soft with the Stoxx 600 down 0.5% at 634.75, the CAC off about 0.8%, the DAX down about 0.5% and the FTSE 100 down 0.3%. The euro was the weakest major currency on the week.
Asia: Korea Takes the Governance Discount and Moves On
The Solidigm withdrawal is the Asian story and it resolves something we flagged on Monday. SK Hynix said it has stopped reviewing a pre-IPO round and a listing, citing financial market volatility and the chip correction, and stated plainly that the decision does not reflect any change in Solidigm’s growth prospects or its investment needs. Translated: the asset still needs money, the listing is no longer the cheapest way to get it.
We wrote on Monday that the Financial Supervisory Service questioning whether a Korean company can monetise its most valuable subsidiary was a statement about who owns the upside in an asset already held, and that Korean memory was being priced as a jurisdiction rather than a demand proxy. The company has now conceded the point. What is notable is the share reaction: down as much as 4% at the open, then positive by the close, with the Kospi finishing up 0.3% at 5,550.30 after being down 4.1% intraday. Holders appear to prefer a parent that keeps the asset whole over a parent that sells a stake into a bad tape, which is a vote for governance clarity over near term cash.
Elsewhere the Nikkei closed down 0.34% at 62,693.11 with Japanese core inflation at 2.1%, Hong Kong rose about 0.7% above 23,000, and the CSI 300 bounced 1.71% to 4,105.33, which still leaves it lower for an eleventh consecutive week and inside the bear market it entered on Thursday. We would not read the bounce as a turn. An eleven week losing streak interrupted by one up day is a streak with a pause in it.
Also On The Tape
- SpaceX is in talks with tower companies for access to at least 80,000 United States cell sites to build a fourth national carrier, roughly half the purpose built towers in the country. Crown Castle is up about 7% to 9% and American Tower about 6%, while AT&T and T-Mobile are down 4% and Verizon 3.8%. Crown Castle’s chief executive would not name customers on the call
- OpenAI wants signatures on its round by the end of next week, with UAE linked investors and Coatue leading, against a March raise of $122bn at an $852bn valuation. Morgan Stanley reportedly saw weaker demand from wealthy clients than in previous rounds
- Hedge funds had a broad losing session Thursday, with multi-strategy books down 1.9%, following two weeks of record net selling in global technology
- Gold is roughly flat near $4,028 at a three month low, with silver at $54.05 and down 1.1%, which keeps the pattern from Thursday where crude rose while the metals fell
- Bitcoin is at about $81,600, up 0.5%, and has stopped falling alongside equities for the first time this week
- Fed pricing has the October 28 meeting at a 5% to 16% chance of a hike and December at roughly 75% to 78%, down from about 86% on Thursday
- Earnings today include SLB, Eni, Infosys, First Citizens, Booz Allen, Gentex, Autoliv and Rogers Communications
The Record
One test resolves, one is marked unresolved as promised, and two new ones go on the board.
Western Digital resolves yes. We said the company had to guide fiscal second quarter revenue above consensus and reiterate committed capacity. It guided $5.1bn against a far lower bar, took gross margin guidance to 60% to 61%, and told the market it will spend above its normal capital band on technology rather than capacity specifically to keep supply tight. The supply shortage leg of our argument is intact and we are not folding it into a sentence about demand. What we got wrong is the inference we were going to draw from it. We built the test to tell us whether to stay long the physical bottleneck, and the answer the tape gave is that the bottleneck is real and the trade is dead for now, because the marginal buyer is not asking about exabytes. That is a worse outcome for us than being wrong about the fundamentals, because being right and losing money is the failure mode that is hardest to correct.
The Volta secondary test is unresolved and we are claiming nothing. We said on October 16 that the loan should trade at or below its 95 cent issue price through its first week of secondary trading, with a print above 96 meaning we had overstated how expensive this funding had become, and we set today as the deadline. There are still no reliable secondary quotes we can source on an unrated deal that CLOs largely cannot buy, so it is marked unresolved, as pre-committed, and we are not back filling it with adjacent evidence. We will note honestly that the question the test was asking got answered elsewhere this week, through CoreWeave’s unsecured notes trading into distressed territory and the Zenith Arc and Galaxy Helios bonds leading data centre declines, and that the answer went our way. That is context, not a score.
One amendment to yesterday. We wrote that the market priced the Meta share class filing correctly within hours. The CDS prints that came out Thursday say the credit market had been marking this for longer than we gave it credit for, and that equity was the late arrival. Small, but we had the sequencing backwards.
New test one, a week out. If CoreWeave’s five year default swap is still above 1,000 basis points at the close on Friday October 30, then the funding stress we have been describing is structural and the dealer quoted thin market objection is dead. If it is back below 800 without a new financing, we overweighted a single illiquid instrument and we will say so.
New test two, Wednesday. Our claim above is that the AI equity drawdown is functioning as a tightening of financial conditions that substitutes for policy. The test: if the Fed holds on October 28 and December hike odds end that week below 60% with the ten year still under 5.10%, the substitution read holds. If December odds are back above 80% with the ten year above 5.15% after a $115 Brent print, then this is a sector event and not a macro one, our financial conditions framing is wrong, and the duration hedge everyone bought on Thursday is the crowded trade to worry about.
Key Events Today
- 9:45 a.m. S&P Global flash PMI, manufacturing expected near 55.5 to 55.6 after 55.9 and services near 57.9 to 58.0 after 58.8
- 10:00 a.m. Final University of Michigan consumer sentiment
- Pre-market American Express reported, with SLB, Eni, Infosys, First Citizens, Booz Allen, Autoliv, Gentex and Rogers Communications
- After the American close, about 4:00 p.m. Moody’s verdict on France, a one notch cut to A1 widely expected, with the outlook language the part that matters
- Monday Verizon before the open, Electronic Arts and Southern Copper
- October 27 and 28 FOMC, a hold priced at better than 84%, with no September PCE in hand
- October 29 September PCE and third quarter GDP, the day after the decision, plus the ECB
- Watching CoreWeave and hyperscaler default swaps, whether OpenAI gets its round signed by Friday, Hormuz transit, and whether a second large employer follows Barclays in attributing headcount cuts to artificial intelligence