Friday, October 16, 2026, written at 9:25 a.m. Eastern, five minutes before the open and twenty minutes ahead of September industrial production. Every future, yield, currency and commodity level in this note carries that stamp. Monthly options expire today. The Federal Reserve enters its pre-meeting quiet period, so there is no official voice left between here and October 28.
Top Call: The Book Got Done, and the Clearing Price Is the Story
Volta Infrastructure’s $5.012bn financing priced yesterday at SOFR plus 650 basis points at 95 cents on the dollar. Talk was SOFR plus 625 to 650 at an original issue discount of 97 to 98. It cleared at the wide end of the spread and two to three points below the wide end of the price, which puts the all in yield above the 11.75% top of the indicated range and closer to 12%.
On Monday this desk pre-committed to a single test: the package prices at or wider than 11.75%, or with a cut tranche, or the funding thesis is wrong. Yesterday we carried the test open because nothing had printed. It has now printed and it resolved on our stated terms. We will take that.
What we want to be careful about is the half of the outcome we did not emphasise enough. It cleared. A $3.712bn best efforts GPU loan secured on roughly 36,000 Nvidia processors, left uncovered by its own arrangers, found buyers in the same week that CoreWeave’s five year default insurance set a record and Oracle’s did too. That is not a funding market closing. It is a funding market repricing, and the distinction decides what happens next. A failed book would have forced the whole neocloud build to be re-planned around equity. A book that clears five points cheap tells you the capital is still there and that the owners of the assets will earn less of the return on them than they expected eight weeks ago. Volta tried this at SOFR plus 500 in August and September and could not get it away. It is now paying roughly 150 basis points more plus a discount, with a two year non-call so the lenders keep the coupon even if conditions improve. The lender took the spread. The sponsor took the dilution of its own economics.
So the frame we have carried since October 7 stands, with one correction to its temperature. Credit is the binding constraint in this cycle rather than compute demand, but the constraint is a price constraint and not yet a quantity constraint. Those behave very differently. A price constraint compresses returns across the AI infrastructure complex quarter by quarter and shows up in equity multiples slowly. A quantity constraint stops projects. We are in the first one, and anybody positioning for the second needs the next book to fail, not merely to widen.
CoreWeave Is Still Where the Risk Is Marked
CoreWeave’s five year credit default swap widened as far as 1,194 basis points intraday and closed at 1,086. That is a record on both measures. The shares bounced about 5% on Thursday after Jensen Huang spoke supportively about the company’s financial position, and they are down about 3.9% again in the pre-market.
Alongside that, three co-founders sold roughly $3.3bn of stock. We are not going to over-read insider sales, which are usually scheduled and almost never timed. What we will say is that the combination is unhelpful: the cost of insuring a company’s debt at a record while its founders monetise $3.3bn of its equity is the kind of pairing that draws a cost of capital higher rather than lower, whatever the merits of each transaction taken alone.
Note what Huang’s comment actually did and did not do. It moved the equity 5%. It did not move the swap, which went to a new wide on the same day. When the two instruments on one company disagree that sharply, the one with a documented recovery assumption is usually doing the arithmetic and the other is doing the sentiment.
Warsh Said Nothing and the Market Called It Dovish
Chair Warsh spoke at the IMF meetings in Bangkok and gave no guidance. He said that since the September meeting the world has broadly conformed to the committee’s expectations, that he does not believe in prejudging meetings, and that he will make his judgments at the October and December meetings. He also described a massive investment surge from artificial intelligence and said the technology will hit demand before it hits supply, and allowed that he may be over-scarred by the financial crisis.
Traders priced the silence as dovish. October 28 hike odds fell to roughly 14% from about 26%, with venues spread between the high single digits and the high twenties depending on where you look. December remains the base case at 78% to 90%. Goldman expects a pause this month and a final hike in December, which is now close to the consensus of the brokerage desks.
Here is the part that matters more than the odds. The ten year sold off five to six basis points back to 5.11% and the two year rose four to 4.78%, giving back a meaningful share of Thursday’s rally, when the ten year fell 12.7 basis points to 5.06% and the two year 5.9 to 4.74%. Yesterday we pre-committed to a test: if this is a credit event rather than a growth scare, investment grade technology spreads should keep widening while the two year keeps falling. Spreads widened. The two year rose. That test came back half right, and we are marking it half wrong rather than quietly retiring the half that failed.
The honest reading is that the front end is not yet treating AI credit as a macro event. It is treating it as a sector event with a wide spread and a functioning primary market, which is exactly what yesterday’s Volta print described. If the two year will not fall on a day when the Chair declines to guide and October pricing collapses by twelve points, it is telling you that the policy path and the credit story are, for now, separate problems.
Everything That Moved Most This Week Moved Mechanically
The Kospi closed up 2.05% at 5,668.05 after trading as much as 4% higher intraday, its first gain in eight sessions and a thin recovery from Thursday’s 7.57% collapse. The Nikkei rose 1.39% to 65,324.29 and the Topix added about 1.9%. Hong Kong gained 1.21%, the CSI 300 about 0.8% and Taiwan 0.7%. MSCI Asia Pacific rose as much as 1.8% and is still heading for its worst week in more than three months.
The exception is the one worth your attention. Indonesia’s Jakarta Composite fell about 11%, its steepest single day decline since January 1998, after MSCI proposed reclassifying the country to standalone market status. There is no earnings news, no policy change and no credit event behind that. It is an index rule.
Put the week’s three largest dislocations side by side. Korea fell 21% across seven sessions, driven at the margin by domestic retail unwinding leveraged single stock products. Indonesia fell 11% on a proposed MSCI reclassification. Strategy is down about 3% in the pre-market on separate MSCI rules that could remove a bitcoin holder from the global indices, with the final list on November 11 and effect after the close on November 30. Not one of those three moves came from anybody revising a view on the underlying assets. They came from index construction and from borrowed money.
This is the most useful observation we can offer this morning, and it cuts against our own positioning. We spent the week writing about credit spreads as the analytical centre of the market. The analytical centre may be right and still be a minority of the realised volatility. When the biggest moves on the tape are produced by rulebooks and by margin, the information content of price is lower than it looks, and a desk that reads every 7% index move as a message about terminal value will be wrong roughly as often as it is right.
Rotation Held, and the Semiconductor Tape Did Not
Thursday closed with the S&P 500 up 0.33% at 7,744.06, the Dow up 1.08% at 52,177.69 and the Nasdaq Composite up only 0.08% at 26,713.03, with the Nasdaq 100 fractionally lower at 29,935.59 after recovering a steep intraday loss. The equal weighted S&P rose 1.36% and the Russell 2000 1.38%.
Health care set a record high, with the sector ETF up 1.86%, its first record since mid August, after RBC moved to overweight health care and market weight technology on valuation. Materials, consumer discretionary and financials all gained more than 1.3% while technology fell about 0.65%. The semiconductor ETF closed at $514.71 for an eighth consecutive loss and is down 12.6% from its October 2 level.
We described this on Thursday as a rotation rather than a liquidation, with capital leaving the levered end of the AI complex for assets that do not need a syndicate to fund next year’s capital plan. That held, and defensively it held well: a market that can take an eight session decline in its most crowded industry group while the index makes a small gain and the equal weighted version makes a large one is doing orderly work. The qualifier is that eight sessions is now long enough to be a trend in semiconductors rather than a reaction, and that today’s monthly expiry will clear a lot of the hedging that has shaped the last two weeks of tape. Positioning on Monday will be a cleaner read than anything we see today.
Earnings: Nothing in the AI Complex Reported, Which Is Itself Worth Saying
This is a bank and insurance morning. No company in the compute, memory, optical or launch complex reported overnight, so there is no new operating data against the funding story that is driving the market. The two prints that carry information are below.
| Company | Consensus | Actual | Read |
|---|---|---|---|
| Alcoa | $1.54 adj EPS | $1.82 adj, revenue $3.7bn light | Beat on operations, missed on price |
| Truist | $1.19 adj EPS | $1.16 adj, $0.56 reported | Missed, with a $1.4bn bond loss |
Alcoa earned $483m, or $1.81 a share, against $407m in the second quarter, with adjusted earnings of $1.82 against roughly $1.54 expected and revenue near $3.7bn that came in about 2% light on weaker metal prices. Alumina third party revenue rose 7% to $680m on better realised prices and higher shipments, with a $95m segment EBITDA improvement credited to operations at Pinjarra. Free cash flow was $472m and the quarter closed with $4.4bn of cash, including $2.6bn raised for the South32 interests. Our take is that this is the better shape of beat: the upside came from the plants rather than the price deck, and a producer generating $472m of free cash while funding an acquisition is positioned for the one part of the AI build nobody can finance their way around, which is power and primary metal. The revenue miss is the honest warning that the metal price is not doing the work.
Truist adjusted $1.16 against $1.19 to $1.21 and reported $0.56, the gap being a $1.4bn loss on a bond sale, with revenue of $5.3bn slightly light and taxable equivalent net interest income just short. The shares are down about 2.1%. This is the sixth consecutive quarter in which a large American bank has taken a realised securities loss to reposition a portfolio built when the ten year was nowhere near 5.11%, and it is the same line item that cost JPMorgan 18 cents on Tuesday. The curve is not an abstraction for these balance sheets. It is a cash charge, taken a quarter at a time, by institutions that have decided it is cheaper to crystallise the loss than to carry the drag.
Elsewhere in the morning, Citizens Financial beat at $1.41 against $1.39 and raised its dividend 13%, Travelers fell 4.6%, Interactive Brokers beat on revenue at $1.914bn and raised its dividend but fell about 2% after hours, and J.B. Hunt missed on earnings while beating on revenue.
Tesla, and Where the Risk Appetite Went
Tesla is up 4.4% to 5% in the pre-market after unveiling a redesigned Roadster developed with SpaceX, with rocket style thrusters, a claimed zero to sixty below one second and production scheduled for 2028.
We have nothing to say about the vehicle. We have something to say about the tape. On a morning when the central question in the equity market is the cost of capital for companies that need to fund assets, the single largest move in a mega cap is a product announcement for a car that will not be built for two years. That is not a criticism of anybody’s positioning. It is a measurement. Risk appetite has not left the market, it has moved to places where the story does not have to be financed in the next twelve months. Hold that next to a $5bn loan clearing five points cheap and you have the whole of this week in two sentences.
Energy: The IEA Validated Half of Our Thesis
Brent is $105.21 and down 0.44%, WTI $97.35 and down about 0.75%. Both are set for weekly gains, Brent around 0.7% and WTI close to 5.9%, the gap explained by an 8.56m barrel American crude draw for the week to October 9.
The IEA agreed on Thursday to release more than 20m barrels of diesel from emergency stocks and members pledged to complete the roughly 100m barrels still outstanding from the March release. European diesel refining margins fell on the news and crude barely moved, which is precisely the split this desk has been arguing for a week: crude is a surplus story for 2027 and distillate is a shortage story right now. When an emergency release is called in the middle of a reported crude glut, the agency is telling you which barrel is actually scarce.
The constraint behind it is unchanged and getting worse. Hormuz flows averaged 11.62m barrels a day in the week to October 13 against a 17.15m baseline, roughly 68% of normal, with Iran stepping up strikes near the strait. A container ship was struck again. Pakistan has confirmed a military deployment to Saudi Arabia after Houthi attacks. Trump said Tehran has asked for a halt to American strikes, which is unconfirmed and unresolved, and Washington is still weighing sanctions relief for some shadow fleet tankers to bring freight down. None of that changes the 2027 balance. All of it raises the cost of moving a barrel today.
Europe Outperformed Again, and the FTSE Found a Domestic Problem
The Stoxx 600 is 643.72 and up 0.60%, heading for a third consecutive gain and its strongest week since August, with the DAX up about 0.9%, the CAC 40 modestly higher and media and autos leading. Health care is the only sector lower, which is the mirror image of Wednesday’s American session and a reminder that the rotation is a flow rather than a verdict.
Bloomberg attributes European outperformance to a defensive tilt away from the AI trade, which is the compositional argument we have made for a week and conceded yesterday had stopped working. It worked again today. We will restate it carefully: these indices hold banks, insurers, defence and luxury rather than memory and neoclouds, so they outperform when the problem is specific to leveraged American technology and they do not when the problem becomes the global cost of capital. Yesterday looked like the second. Today looks like the first again. One session does not settle which it is.
The FTSE 100 is the laggard, with Lloyds down about 2.3% and NatWest 2.1% on reports that the Chancellor plans to raise the bank surcharge ahead of the October 28 Budget. That is a clean domestic policy risk rather than a read on the sector, and it arrives in the same month that European bank earnings have been the strongest part of the regional tape.
Also On The Tape
- Proofpoint, owned by Thoma Bravo, agreed to buy Varonis for about $6.3bn in cash at $49.75 a share, with Varonis up 2.3%
- China proposed export controls on advanced artificial intelligence technology while Taiwan said it will curb AI chip exports to China, a two sided tightening in the same week
- Taiwan is reportedly planning billions of dollars of American industrial parks to support Nvidia and other AI hardware makers, starting with a Texas complex valued at up to $4bn
- Mercedes-Benz and Nvidia extended their automated driving partnership
- Netflix is down about 2.5% and is replacing its chief content officer, with Disney’s Peter Rice succeeding Bela Bajaria on October 26
- Anthropic investors are now discussing a listing near $1.5tn against the $2tn previously circulated, unchanged from yesterday
- International Seaways agreed to buy Bruton in a deal valuing it near $370m, tanker consolidation inside a freight squeeze
- The Philadelphia Fed index fell to 20.9 in October from 37.8, jobless claims were 199,000 and continuing claims 1,710,000
- Upgrades drove Mobileye up 4%, Semtech 4.4% and Zeta 4.6%; Nordex rose 15% and Aixtron 10% in Europe
The Record
The Volta test resolved on the terms we set on Monday and carried open yesterday. It priced at the wide end of the spread and below the wide end of the price, for an all in yield above 11.75%. We claim that one. We also note what we got wrong inside a call we got right: we framed the test as pass or fail on price and never seriously modelled the case where the book clears cheap, which is what happened and which is a materially less bearish outcome than the one our language implied. A thesis that only specifies the direction of the price is not fully specified.
Yesterday’s second test came back half right. Spreads widened, with CoreWeave setting a record on both intraday and closing default insurance. The two year did not fall. It rose four basis points on a morning when the Chair declined to guide and October hike pricing collapsed. We said that if the two year reversed higher while spreads widened we had over-read the credit event, and we are holding ourselves to the softer version of that: the credit repricing is real and sector specific, and we were too quick to promote it to a macro event.
On Korea we retired our near term overreaction call yesterday after the index fell a further 8%. It rose 2% today and traded 4% higher intraday. The call was wrong on the mechanism, which is why we retired it, and the retirement was badly timed, which is worth saying plainly rather than hiding behind the mechanism. We do not get to collect the bounce we had just stopped forecasting.
One new test for Monday. If the AI credit repricing is structural rather than a positioning event, Volta’s loan should trade at or below its 95 issue price in the secondary market through the first week. If it trades above 96, the five point discount was a concession the market absorbed easily and we have over-stated how expensive this funding has become.
Key Events Today
- 9:15 a.m. September industrial production, consensus +0.4% against 0.0%, capacity utilisation 76.5%
- Through the morning Import and export prices, with import inflation running near 7.9% year on year
- 1:00 p.m. Baker Hughes rig count
- At the close Monthly options expiry, with SPX pricing a 0.54% move
- From today Federal Reserve quiet period begins ahead of the October 27 and 28 meeting
- Watching Volta secondary quotes, CoreWeave default insurance, and a ninth straight loss in semiconductors