US close, 21 October
Capital expenditure went up. So did the cost of funding it.
Technology led a third session of recovery, semiconductors rose 1.92% and volatility fell again. On the same day high yield widened to 333bp, the widest of this episode, and one of the largest platforms lifted the floor of its 2026 capital-expenditure guidance. We hold stage four, and make the way out easier.
What changed on the day
Where we sit in our risk framework
Divergence
Yields, oil and inflation deteriorate while equities and volatility stay calm.
Structure resilient
Positive gamma and contained credit. Credit still fails it.
Controlled de-rating
Held sixteen sessions, from 25 September until 19 October.
Broader deleveraging
Third session. Credit at its widest while equity mechanics repair.
The equity half of this market is mending and the financing half is not. Page four sets out what that does to our conditions, and makes the exit easier rather than waiting to be asked.
How the day changed the dashboard
| Indicator | 20 Oct | 21 Oct | Signal | What changed |
|---|---|---|---|---|
| Equities and breadth | ||||
| S&P 500 | 7,712.57 | 7,757.17 | Stable | +0.58%, and now above its flip for the first time |
| Equal-weighted S&P 500 | +1.16% | +0.33% | Watch | Narrowed again. Smaller companies +0.27% |
| Nasdaq ETF (QQQ) | 717.22 | 725.25 | Elevated | +1.12%, still 0.45% below its flip |
| Semiconductors (SMH) | 546.09 | 556.57 | Severe | +1.92%. Chips +2.01%, yet 12% below 6 October |
| Volatility and dispersion | ||||
| VIX | 17.31 | ~16.70 | Watch | A third decline, from 22.31 intraday last Friday |
| VVIX (volatility of volatility) | 97.15† | 94.55 | Watch | Below 100 for the first time this episode |
| Cboe DSPX dispersion | — | — | Elevated | Unverified a second session. We are not estimating it |
| Single-stock implied volatility | — | — | Elevated | Depends on the dispersion index. Also unavailable |
| Dealer positioning (net gamma) | ||||
| SPX | −$8.6bn | +$38.9bn | Stable | Positive, and 0.58% above its flip at 7,712 |
| SPY | +$0.21bn | −$9.6bn | Watch | Negative on a snapshot taken well before the close |
| QQQ | −$3.8bn | −$2.5bn | Elevated | Better, still below its 728.51 flip |
| SMH | −$683m | −$695m | Severe | Marginally worse, 6.4% below its flip |
| Macro and rates | ||||
| US 2-year Treasury | 4.785% | 4.823% | Elevated | +3.8bp, the largest move on the curve |
| US 10-year Treasury‡ | 5.116% | 5.145% | Severe | +2.9bp. Above 5% for a fourth week |
| US 30-year Treasury | 5.425% | 5.437% | Elevated | +1.2bp. The curve stays positively sloped |
| Brent crude* | ~US$109 | US$108.82 | Severe | Flat, and still inside our US$107–110 zone |
| Implied chance of an October rate rise | 17.2% | 15–16% | Watch | Lower. The meeting is 28 October |
| Credit | ||||
| ICE BofA US High Yield OAS§ | 329bp | 333bp | Severe | +4bp. The widest of this episode |
*Brent is the next ICE Brent futures contract, rolled on each contract's last trading day. †The 20 October volatility-of-volatility close, unavailable when we published yesterday, has since been reported at 97.15. ‡The 20 October 10-year close is the verified 5.116%, which supersedes the approximate 5.10% we published yesterday. §High-yield spreads are shown by observation date; 333bp is the 20 October observation, published on 21 October. Exchange-traded fund prices use a single-venue daily feed and may differ slightly from consolidated closing prints.
Options market structure
| Market | Snapshot spot | Closing level | Zero gamma | Close vs flip | Put / call wall | Net gamma |
|---|---|---|---|---|---|---|
| SPX | 7,757.85 | 7,757.17 | 7,712.30 | +0.58% | 7,650 / 7,900 | +$38.9bn |
| SPY | 766.91 | 773.13 | 772.26 | +0.11% | 760 / 800 | −$9.6bn |
| QQQ | 723.82 | 725.25 | 728.51 | −0.45% | 700 / 760 | −$2.5bn |
| SMH | 556.63 | 556.57 | 594.93 | −6.45% | 530 / 650 | −$695m |
We show both bases this session because they disagree. The snapshots for the index, the Nasdaq fund and semiconductors were captured within a fraction of their closes; the S&P ETF snapshot at 766.91 was taken 0.8% below where it finished, and the distances in the last three columns are measured against closing levels rather than the snapshot. Net gamma figures remain provider-model estimates calculated on the snapshot.
The S&P and its ETF disagree, and probably shouldn't
The index printed positive gamma of US$38.9bn while its exchange-traded fund printed −US$9.6bn. Two instruments on the same market, in opposite mechanical regimes.
The explanation is likelier to be timing than structure. The fund's snapshot was captured at 766.91 against a 773.13 close. Measured on the close it finished 0.11% above its flip, which agrees with the index. We would not trade the divergence until synchronised snapshots confirm it.
Breadth has flipped three times in three sessions
Monday was one sector up and six down. Tuesday the equal-weighted index rose 1.16% and advancers ran two to one. Wednesday it managed 0.33% against 0.58% for the headline, with four sectors lower.
We used breadth as supporting evidence on Monday and withdrew it on Tuesday. It has now reversed again, so we are demoting it: from today it is an observation we report, not an input to the stage.
Rebound without repair
Semiconductors rose 1.92% and chips 2.01%, the strongest two-day stretch of the month. The sector is still about 12% below where it stood on 6 October and 6.4% below its own flip.
Its positioning barely moved: −US$695m against −US$683m, with downside concentration still anchored at 530. Price has recovered faster than structure, which is the same gap we flagged on Tuesday.
Credit made a new wide
High yield widened 4bp to 333bp, the widest print of this episode and 73bp above where it sat at the end of August. The sequence is 325, 331, 329, 333.
Tuesday's 2bp improvement looked like the start of a turn. It was not. Two of the last four observations have been new wides, and the direction over the month is unambiguous.
The spending and the cost of it
One of the largest platforms raised the floor of its 2026 capital-expenditure guidance, leaving a US$135–145bn range, and results elsewhere in the complex were taken as confirmation of demand.
That is the demand question answered and the funding question sharpened. Committing to spend at that scale while the cost of high-yield money makes new highs is precisely the tension we said last week had replaced the growth debate.
The whole curve moved up
The 2-year rose 3.8bp to 4.823%, the 10-year 2.9bp to 5.145% and the 30-year 1.2bp to 5.437%. A long-dated government auction cleared at an elevated yield.
A fourth week above 5% on the 10-year is the constraint that does not move. For assets whose value sits in cash flows a decade out, the discount rate is doing more damage than the news flow is repairing.
Where stage four stands, and a gap in our own drafting
| Condition | Level | Now | Status |
|---|---|---|---|
| High yield above | 325bp | 333bp | Met. By 8bp, and the widest of this episode |
| Gamma negative across the complex | All | Two of four | No longer met. The index and its ETF closed above their flips |
| VIX above | 20 | ~16.7 | Not met for a third session |
| Breadth narrowing | — | Mixed | Withdrawn as an input. Reported, not weighted |
The gap. Entry to stage four required credit above 325bp with gamma negative across the complex. The exit required two credit prints below 325bp and the complex back in positive gamma. Today sits between them: credit is comfortably through the line while gamma has largely repaired, so neither the entry nor the exit is satisfied. That is a drafting fault in our own conditions and we would rather name it than let the stage drift on inertia.
The fix, effective from the next note. The gamma limb of the exit is now substantially met, so it should stop being a second hurdle. From tomorrow, a single high-yield observation below 325bp, with the index complex at or above its flips, returns us to stage three. One print, not two. We note that this makes de-escalation easier while we are the ones holding the escalation, which is the only direction in which we are comfortable relaxing a rule we are currently relying on.
AI compute fundamentals overlay
Unchanged from the 7 October reading: frontier coverage 2.03× against our 1.20× level. Thirteen sessions old.
| Generation | Rate | Coverage | Read |
|---|---|---|---|
| Oldest frontier | US$2.85/hr | 1.19× | On our 1.20× level when last observed |
| Middle generation | US$5.29/hr | 2.03× | Had fallen 12.6% from 3 October |
| Newest generation | US$7.56/hr | 2.38× | Softer, 3.2% below its 3 October level |
The funding-cost companion we described on Tuesday now has its first input, and it is the high-yield spread itself at 333bp. A sector committing to US$135–145bn of annual capital expenditure is a sector whose equity value depends on the price of money, and this table says nothing about that. We will publish the companion before we lean on this overlay again.
Tomorrow's board
- SPX flip / put wall7,712 / 7,650
- SPY flip / put wall772.3 / 760
- QQQ flip / put wall728.5 / 700
- SMH flip / put wall595 / 530
- VIX16, then 20
- High yield, next print325bp
- US 10-year5.20%
What would move us
Stated before the data arrives.
- One credit print below 325bpStage 3
- With the complex at its flipsRequired
- Credit through 350bpDeepens stage 4
- SPX loses7,712
How we see the next one to five days
The stage-four conditions, with one of three met and one now failing:
These weightings are our current assessment and will change with the data.
What this means for our outlook
Wednesday answered the question the market asked in September and sharpened the one it asks now. Results across the technology complex were read as confirming that demand for computing is real, and one of the largest platforms raised the floor of its 2026 capital-expenditure guidance to a US$135–145bn range. Semiconductors rose 1.92% and volatility fell for a third session. The same day, high yield widened to 333bp — the widest of this episode.
Those two facts belong together. Committing to spend at that scale is committing to raise money at whatever the market charges, and it has just charged more than at any point in this episode. We wrote last week that the question had moved from how fast demand would grow to who can fund building it. Wednesday answered both halves, and they pointed opposite ways.
We hold stage four for a third session, and the conditions do not sit neatly. The index closed above its flip with positive gamma of US$38.9bn, and its ETF did too once the stale snapshot is corrected, so gamma negative across the complex no longer holds. Credit through 325bp does. Our entry and exit tests leave a gap and today falls into it. Page four names that fault and closes it in the direction that lets us out sooner: from tomorrow one credit print below 325bp, not two.
We are also retiring breadth as a stage input. We offered it on Monday, withdrew it on Tuesday and watched it reverse again on Wednesday. A measure that changes sign three sessions running describes rotation, not regime, and we were wrong to lean on it. What remains is harder to argue with: the cost of equity capital is falling and the cost of debt capital is rising, in a sector that has just committed to a great deal of both.