Read more

1 · This Week’s Action

The winners were sold — all of them, and hardest into Friday. It started with the chips and did not stay there. The chip fund SOXX fell 11.8% to 493.39, its eighth down day in a row on Friday. Software joined on Wednesday and again on Friday: the software fund IGV fell 7.4%, 4.8% of it on Friday alone; CrowdStrike −17.3%, Palo Alto −18.3%, Snowflake −16.8%, Palantir −11.0%. And the hyperscalers went with them: Oracle −9.3%, Amazon −4.7%, Alphabet −3.1%, Microsoft −1.1%, with Alphabet −2.5%, Meta −3.7% and Microsoft −1.9% on Friday alone. Nvidia fell 6.6% to 214.13. This was not one sector’s bad week. It was broad profit-taking in everything that had carried the year.

The index and the market under it. The S&P 500 fund fell 1.97% to 763.25. The Nasdaq 100 fund fell 5.52% to 709.80. But the equal-weight S&P rose 0.49% — the average stock was up in a week the index was down two percent and the Nasdaq five. The Nasdaq 100 without its tech stocks rose 0.82% to 98.95, its second week up; the tech half of the Nasdaq 100 fell 9.15%. Small caps fell 3.05%.

The tech bull channels broke. Since the 30 March low every dip in the tech funds stopped on a rising line, drawn from that low through the late-July low. This week QQQ closed under it — the line runs at about 729 to 732 now, the fund closed Friday at 709.80. Wednesday and Thursday it closed right on the line, at 728.95 and 728.00; Friday’s 2.5% drop took it through. The tech sector fund XLK closed at 183.47 against a line at about 193; the chip fund SMH at 539.79 against about 601. Software is the one that kept its spring line — IGV at 104.34 sits above the line from the April low through the June low, near 100 — but it fell through the steeper line it had followed since August.

The editor’s chart: QQQ over one year — the rising line from the 30 March low through the late-July low, broken on Friday at 709.80; horizontal lines at 746, 729, 704, 691, 674 and 660; the editor marks the June top as wave 1.

The levels under it, from the editor’s chart. 746 — last week’s confirmation line — was lost on Monday by twelve cents and never regained. 729 went on Friday. The next lines are 704, the area of the September low (704.54 on 15 September, 700.00 intraday the day after), then 691, 674 and 660 — the last the July low that the rising line was drawn through.

The eleven S&P 500 sectors

Seven of eleven up — in a week the Nasdaq fell 5.5%. Consumer staples +5.02% to 87.62, their highest close since 3 March. Energy +4.86%. Health care +3.45% to 176.71, after a one-year closing high of 177.19 on Thursday. Materials +3.01%, utilities +1.93%, financials +1.44%, real estate +1.32%. Communications −0.05%, consumer discretionary −0.35%, industrials −0.88% — and technology −7.70%, the only sector with a real loss and still +27.4% on the year.

The shape of it: defensive growers. Staples and health care are the companies that keep growing when the economy slows — food, soap, medicine. Friday’s Pulse put it in one line: out of chips, into pills and soap. Since the 6 October record health care is up 6.0% and staples 6.0%, technology down 6.3%, chips 12.7%. Against the S&P this week staples beat it by 7.0 points, energy by 6.8, health care by 5.4; technology trailed by 5.7.

Breadth — health care and energy strong, tech thin. In health care 51 of 59 members closed above their 5-day average and 42 above the 50-day, with 17 new five-day highs against one low. In energy 16 of 21 are above the 50-day. In financials 55 of 76 are above the 5-day, 30 above the 50-day — up from none above the 20-day two weeks ago. Technology is the mirror: 26 of 72 above the 50-day, 45 new five-day lows against 7 highs.

The Nasdaq 100, cut six ways

The week: the Nasdaq 100 without tech +0.82%, the S&P 500 without tech +1.04%, the S&P without the Magnificent Seven −1.65%, the S&P −1.97%, the Magnificent Seven fund −3.38%, the Nasdaq 100 −5.52%, the Nasdaq tech sector −9.15%. On the year: Nasdaq tech +30.8% (from +44.0% a week ago), the Nasdaq 100 +15.5%, the S&P +11.9%, the S&P without tech +6.7%, the Nasdaq without tech −0.4%. Thirteen points of the tech half’s year went in five sessions.

The counterpoint: the Nasdaq without tech is green. QQXT closed at 98.95 — its low of the year was 95.56 on 1 October, and it closed at 99.26 on Thursday, the best close since the low. In the week the tech half fell 9%, the other half of the same index rose. That is rotation inside the Nasdaq, not a sell-everything week: the money that left Nvidia, CrowdStrike and Oracle went to the drug makers, the retailers and the food companies in the same index.

The editor’s chart: the Nasdaq 100 without tech over one year — at 98.95, two weeks off its early-October low near 95.5 and still under its August high near 105.

Momentum against low volatility — the turn. Momentum fell 6.79%, low volatility rose 3.24%. The ratio dropped below its 50-day average — from an indexed 166.4 to 150.2 — and the factor board now reads “defensive rotation — low-vol leading, regime-change watch.” It is still in the 82nd percentile of its range. One week does not end a momentum regime; it does mark the first real crack in it.

The tech ETFs we carry

Nineteen of twenty-one red. XSD −12.48%, SMHX −12.14%, SOXX −11.80%, SMH −10.53%, cybersecurity −10.16%, software −7.39%, cloud −6.78%. Only defence tech (SHLD +2.69%) and gaming (ESPO +0.65%) rose. The software-over-semis ratio still rose 3.5% — software fell less — and is 48% above its June low, but this week software was not the shelter it was a week ago; it was the second wave of the selling.

The names in one line each. On the way down: Applied Optoelectronics −31.1%, Aehr −28.4%, Coherent −25.5%, Lumentum −25.3%, Nebius −25.2%, CoreWeave −25.2%, Fabrinet −25.3%, HPE −25.0%, Credo −21.6%, Okta −19.3%, Applied Materials −19.3%, Palo Alto −18.3%, CrowdStrike −17.3%, Dell −17.6%, Snowflake −16.8%, Micron −14.6%, AMD −13.1%. Up: Intuit +7.4%, Eli Lilly +6.6%, Stryker +6.3%, Chevron +6.1%, Exxon +5.9%, Salesforce +5.1%, Veeva +4.8%, Amgen +4.3%, Apple +2.0%. The software names that rose were the out-of-favour ones — Intuit is still down 48% on the year; the ones that had led the year were sold.

The four indices: Dow, S&P 500, Nasdaq 100, Russell 2000

The S&P 500 −1.97%, the Nasdaq 100 −5.52%, the Russell 2000 −3.05%. The Dow held up best: it gained 1.1% on Thursday and was still green at midday on Friday while the Nasdaq fell, because it holds more banks, health care and consumer names and fewer chips. The S&P is 2.0% under its 6 October record, the Nasdaq 6.6% under its own. The Russell is 6.4% under its 50-day average.

The Treasury curve, one month to thirty years

The ten-year yield fell from 5.24% to 5.10% on the week, the thirty-year to about 5.43%, the two-year to about 4.72%. It fell in a week of hot inflation numbers: consumer prices rose 3.7% on the year in September, core 2.6%, both above forecast (Wednesday); producer prices jumped 1.3% in the month (Thursday). The bond market looked past them — the ten-year touched 5.24% on Wednesday and fell to 5.10% on Thursday, as consumer spending outside food and energy stalled. The long-bond fund TLT rose 2.65% to 80.05, with a gap higher on Thursday on the heaviest volume of the month.

The editor’s read of the bond chart: TLT may have made an initial breakaway gap. On Thursday the fund opened at 80.47 after closing at 78.89 the day before, and the day’s low of 79.93 never came back to Wednesday’s high of 79.01 — a gap of almost a dollar, on 86.7 million shares, the heaviest volume of the month. Friday traded down to 79.67 and closed at 80.05, so the gap is still open. That close sits on the two lines on his three-year chart: the falling line from the 29 June high (87.45) and the 79.9 floor the fund held in 2024 and 2025 and first closed under on 24 September. 74 is the line below. His point: a breakaway gap starts a move; if the move is real, a second gap higher — a running gap — may follow soon, and the next test is a close clearly above the falling line. All four Treasury funds gapped higher on Thursday and held it on Friday: TLH (10–20 years) at 94.09 on its own falling line, IEF (7–10 years) at 90.42 back above its 89.9 line with 93.5 next, IEI (3–7 years) at 114.24.

The dollar

The dollar fund slipped 0.34% to 28.92 while the yields fell. A softer dollar and lower long yields in the same week is what a market pricing slower growth looks like.

2 · The State

The week, in order. Monday: Korea’s market, back from two days off, fell 3.5% to catch up with the US chip selling; in New York memory and optics were sold again — Lumentum −7.7% and SanDisk −6.5% at midday — while Meta rose 3.8%. Tuesday: the Kospi broke its 18-month uptrend; J&J and Citi fell after beating, JPMorgan recovered after missing; the chip fund closed back under its 50-day. Wednesday: hot inflation, the ten-year at 5.24%, the chip fund −3.7% and 20% under its June high, software −2.4%, Oracle −7.5%, AMD −4.9%; health care +2.0% the only shelter. ASML reported a record quarter and said it is sold out into 2028. Thursday: TSMC beat but called itself “prudent” on 2027; the chip fund −2.0%, Dell −7.7%; software was bought, ServiceNow +5.4%; the ten-year fell to 5.10%; health care +2.5% led again. Friday: the whole complex sold — chips −2.1% for the eighth day, software −4.3%, CrowdStrike, Snowflake and Palo Alto −7% — while the Dow stayed green and the money stayed in staples, banks and utilities.

The mechanism, named: profit-taking in the winners, not a flight from stocks. A sell-everything week takes the equal-weight S&P down with the index; this one did not. The S&P fell 2%, the average stock rose. The money went from the places it had made the most this year — chips +86% on the year before the week, the AI builder index +103% — to the places it had made the least: staples, health care, energy. Two things pushed it. The AI supply chain got its first hard question about 2027: TSMC prudent, an earlier report that put OpenAI’s revenue nearer 50 billion dollars than 70 billion. And a peak in rates with slower growth behind it — the ten-year off its 5.31% high of 5 October in a week of hot prices — favours companies that grow without a strong economy.

The channels, said plainly. The rising line from the 30 March low held every dip for six and a half months: in June, in July, in September. A break of it after the chips fell 25% from their June high is not a crash; it is the end of the easy part of the trend. What it changes is the stance — a dip that breaks the line is not bought the same way as a dip that holds it.

The chips — the second leg. SOXX 493.39: 24.7% under its 22 June high, 8.2% under its 50-day, 6.1% above the 465.00 July low. SMH 539.79, 19.3% under its June high; XSD 462.03, 28.7% under, back at its July level. The July low is the line the whole chip complex now trades against.

Software — from shelter to seller. IGV 104.34 — 7.4% under last Friday’s 112.67, which was its best close of 2026; 1.5% under its 50-day. CLOU 28.07, 6.8% under its 30.11 high of last Friday. CIBR 98.82, 10.2% lower on the week. A week ago software was making highs while the chips fell. This week it fell with them.

3 · The Outlook

The three-index read — capex and opex both sold, applications flat. Capex (Rubin Build-Out) −13.45%, 26.8% under its June high, +76.4% on the year. Opex (Agentic Ecosystem) −14.16%, from a record a week ago — not one of its 34 names rose. Applications (Agentic Winners 40) −0.50%, after −2.0% on Friday. The builders and the companies that run AI fell together; the applications held most of the week and gave a little back at the end. HALO — growth with no AI thesis — fell 0.5%.

Read more

1 · This Week’s Action

The index and the market under it. The S&P 500 fund rose 1.16% to 778.57 — a record close of 779.09 on Tuesday, through the August high of 777.88 this letter carried as the line above. The Nasdaq 100 fund rose only 0.23% to 751.27, after its own record close of 759.66 on Tuesday. The equal-weight S&P rose 1.58% — the average stock beat the index, and the index beat the Nasdaq. Small caps fell 0.92%.

Last week this letter wrote that the Nasdaq made its record while the market underneath made its low. This week the order turned: the Nasdaq 100 without its tech stocks rose 2.35% to 98.15, back above 96.92 — the first sign of a turn this letter named a week ago — while the tech half of the Nasdaq 100 fell 1.42%. The S&P 500 without tech rose 2.14%.

The macro — a hawkish week that traded like a peak. The news on rates was not friendly. On Wednesday the Fed’s September minutes pointed to another rate hike by year-end. On Monday the ten-year Treasury yield touched 5.35%, and the ten-year real yield — the TIPS yield, what the government pays above inflation — closed at 2.95%, its highest since November 2008. By Friday the ten-year was back at 5.24% and the real yield at 2.91%; the long-bond fund made a new low close on Monday and ended the week up 0.65%. The market read the week the other way from the minutes: it bought the sectors that live on interest rates and sold the ones that live on next year’s capital spending.

The eleven S&P 500 sectors

Ten of eleven green — and every one of the ten did better than technology. Utilities +3.97%, staples +3.60%, energy +3.60%, health care +2.79%, consumer discretionary +2.55%, financials +2.32%, real estate +1.96%, materials +1.17%, communications +0.05%. The two red: technology −0.52% and industrials −0.41%. Last week three sectors rose and technology was the best of them. This week technology was the worst.

The bottoms came in a sequence. The rate-sensitive sectors did not turn on the same day; they turned one after another over eight sessions, and the order is the read of the week:

• Utilities on 28 September — 39.25, now 41.41, +5.5% off the low and back above the 20-day average.

• Financials on 30 September — 53.40, now 54.73, +2.5%, every close since above it and Friday’s the highest since 25 September; the fund sits exactly on its 20-day average.

• Staples on 1 October — 80.33, now 83.43, +3.9%, with a 2.1% jump on Thursday.

• Health care on 2 October — 166.18, a higher low than the 165.36 of 11 September, now 170.81, +2.8%, back above its 50-day average.

• Real estate on 7 October — 40.57, now 41.61, +2.6%.

Defensives first, then the financials, then property. That is the order a market moves in when it starts to price the end of a rate rise — and it started while the central bank was still talking about the next one.

Consumer discretionary — the cap-weighted fund and the average. XLY rose 2.55% with Amazon +4.3% and Tesla +3.3% — but the move was not two names: the equal-weight discretionary fund rose 1.58% and the equal-weight retail fund 1.77%. XLY is still 9.4% under its January high and 1.4% under its 50-day. The one corner of the consumer that did not rise is the one most tied to mortgage rates (below).

Breadth — the financials came back to life. A week ago not one of the 76 financial stocks in the S&P closed above its 20-day average. On Friday 29 did — 38% — and 71% were above their 5-day. Above the 50-day it is still only 20%: a turn, not yet a trend. In energy 20 of 21 members closed above their 5-day average; in health care 44 of 59. In technology 68% of members are above the 50-day — the sector that fell this week is still the one in the best shape underneath.

What has not joined yet. Three groups that usually lead a true rate peak fell this week:

• Homebuilders: ITB −2.17% to 84.76, XHB −2.00% — 25.4% and 21.9% under their February highs, 8.7% and 7.1% under their 50-day averages.

• Banks: the equal-weight bank fund −1.91%, regional banks −2.50% — both made their low close on 7 October and sit on it; brokers rose 1.48%.

• Small caps: −0.92%, 4.3% under the 50-day.

The money bought the large, the defensive and the cash-rich. It did not yet buy what depends on a cheaper mortgage or a cheaper loan. That gap is the measure of how much of the peak the market believes.

The Nasdaq 100, cut six ways

The week: the Nasdaq 100 without tech +2.35%, the S&P 500 without tech +2.14%, the S&P +1.16%, the S&P 500 without the Magnificent Seven +1.05%, the Nasdaq 100 +0.23%, the Nasdaq 100 tech sector −1.42%. On the year the order is unchanged: Nasdaq tech +44.0%, the Nasdaq 100 +22.3%, the S&P +14.2%, the S&P without tech +5.6%, the Nasdaq without tech −1.2%. One week does not undo a year; it does show where the next dollar went.

Momentum against low volatility. Momentum fell 1.35%, low volatility rose 1.76%. The momentum/low-vol ratio fell 3.1% on the week — still above its 50-day and still in the 96th percentile of its range. The defensive end was paid; the aggressive end gave back a little of a very long lead.

The tech ETFs we carry

Software up, silicon down. Cybersecurity +5.06% to 110.00, cloud +4.51% to 30.11, software +3.91% to 112.67 — the last two at their highest closes of 2026. The chip funds: SOXX −5.01% to 559.40, XSD −4.91%, SMH −4.32%, the fabless fund −2.84%. The software-over-semis ratio rose 8.6% in one week and is now 43% above its June 22 low. Thursday did most of it: the chip index fell 3.4% on a report that put OpenAI’s revenue nearer 50 billion dollars than 70 billion, and software rose the same day and again on Friday.

The names in one line each. Zscaler +18.9%, GitLab +10.7%, Palantir +10.8%, Atlassian +10.1%, Okta +9.8%, Veeva +8.3%, Snowflake +8.2%, MongoDB +7.8% on the way up; Aehr −21.1%, IREN −15.7%, Arm −13.4%, Nebius −9.0%, CoreWeave −8.4%, SanDisk −8.0%, Coherent −7.3%, KLA −5.5%, Micron −4.3%, Nvidia −2.0% on the way down. The Magnificent Seven fund rose 1.63% — Microsoft +3.4%, Amazon +4.3%, Alphabet +2.4% — and closed at 73.77, a record close of 73.90 on Tuesday.

The four indices: Dow, S&P 500, Nasdaq 100, Russell 2000

The S&P 500 +1.15% to 7,812, the Nasdaq 100 +0.24% to 30,883, the Russell 2000 −0.94% to 2,806. The S&P and the Nasdaq made records on Tuesday; the Russell is 8.5% under its August high. The Dow sits between them on the chart. Two of the four at their highs, one in the middle, one still well under: the broadening is in the large caps, not yet in the small ones.

The Treasury curve, one month to thirty years

Over the month the whole curve rose — the 2-year to 4.79%, the 10-year to 5.24%, the 30-year to 5.62% on the vendor curve — and over the week the long end eased: the ten-year from its 5.35% touch on Monday to 5.24%, the thirty-year to 5.60% on the sovereign board. The real yield did the same in miniature: 2.95% on Monday, 2.91% on Friday. The level is the highest since 2008; the direction, for one week, was flat to lower.

The dollar

The dollar index rose 0.29% to 102.23, the dollar fund made a new high close of the year on Wednesday. A dollar at its high in a week the market priced a peak in rates is a contradiction worth watching: either the dollar is pricing growth that keeps rates up, or the stock market is early.

2 · The State

The mechanism, named: a peak priced in sector order. Put the week in order. Monday: the ten-year touched 5.35% and the real yield 2.95%, Brazil’s fund jumped 12.5% on its election, US stocks rose anyway and the average stock kept pace. Tuesday: the S&P 500 and the Nasdaq 100 at records, Marvell +7.6% on its investor day, AMD at a record; the storage makers fell — Seagate −9.2%, Western Digital −6.9% — and the chip-machine makers with them. Wednesday: small caps −1.3%, the power and cooling suppliers to AI data centres sold; the Fed’s minutes pointed to another hike. Thursday: Brent above 103 dollars, energy the strongest sector; the OpenAI revenue report — the chip index −3.4%, Broadcom −4.4%, Micron −4.8%, CoreWeave −7.8% — and software up; PepsiCo beat and cut its outlook; TSMC’s September sales +54.6% on the year. Friday: the cloud fund +2.9% for a second up day, Zscaler +6.4%, Datadog +5.7%; Delta cut its outlook on jet fuel; the S&P +0.6% with the average stock alongside.

Hawkish words and the highest real yields since 2008, and the sectors that suffer most from high rates were the best of the week. Whatever the minutes say, this is not the price action of a market that expects rates to keep rising sharply. The market is not reading the Fed’s next step; it is reading the step after it.

Why the peak trade and the chip sell-off are the same trade. A peak in rates helps the companies whose value sits in steady cash today — utilities, staples, insurers, health care — and it helps less the companies whose value sits in spending that has to be financed tomorrow. This week the second group got a second question on top: the OpenAI report asked whether AI’s biggest buyers will earn enough to keep buying. The chip fund SOXX is 14.6% under its June 22 high, the AI builder index lost 3.4% with 110 of its 126 names losing strength, while the applications index rose 4.9% with 33 of 40 names gaining. Money left capital spending and went to cash flow — inside tech and outside it.

The table, applied. Last Saturday’s Global letter set out four combinations of real yields and earnings. For most of 2026 the market sat in the second row: real yields rising because growth is strong, stocks rising with them. This week it traded the third row — real yields peaking, earnings resilient, the setup that can help put in a bottom — for the sectors that had fallen. It did so with the level still at its highest since 2008 and the direction only one week flat. The reporting season decides whether the earnings half of that row holds.

The confirmation held. QQQ’s lowest close of the week was 747.58 on Thursday — above 746.16, the line confirmed a week ago. The index absorbed a 5% chip week without giving up its base.

The chips — the dip. SOXX 559.40: 14.6% under the June high, 4.1% above its 50-day, 5.3% above 531, the first line under it. SMH 603.33, 9.8% under its June high; XSD 527.93, 18.6% under. This week they are the dip the season was expected to offer.

Software — through its lines. IGV 112.67 — through the 110.32 August high, its best close of 2026, 7.4% above its 50-day. CLOU 30.11 — through the 29.11 door, a year high. CIBR 110.00 — 10.9% above its 50-day. Last week software held its year-end line; this week it broke out above its year’s range.

3 · The Outlook

The three-index read — capex down, opex and applications up. Capex (Rubin Build-Out) −3.44% to 2,094.58, +103.8% on the year, 15.5% under its June high. Opex (Agentic Ecosystem) +2.76% to a record 1,786.36, +81.0%. Applications (Agentic Winners 40) +4.91% to 942.64, all nine sub-indices green, −3.7% on the year. The house AI board counts the flow behind it: 110 of 126 Rubin names losing strength, 22 of 34 Ecosystem names and 33 of 40 Winners names gaining. HALO — growth with no AI thesis — was flat at −0.23%.

The regime gauges. The Money Temperature board read 57, unchanged — mixed, transitional, not a signal. The momentum/low-vol pair turned toward low-vol for the week from the 96th percentile. Structural inflation reads “contained”, its direction flat — one reason a market can price a peak while the Fed talks about a hike.

4 · What May Lie Ahead

The fourth quarter — positive for the US, and broader. This letter’s frame for the rest of 2026, unchanged: seasonally the strongest stretch of the calendar, expected to be positive to year-end. What this week adds is breadth. A rally carried by one sector is a bet on that sector’s earnings; a rally in which utilities, staples, financials, health care and the consumer rise together is a bet on the economy and on rates. The second kind usually lasts longer.

Read more

1 · This Week’s Action

The tape, day by day. Five sessions: two up, two down, one up, and the index finished higher. Monday: the S&P +1.55% to 773.50 and the Nasdaq 100 fund +2.77% to 741.47 — Korea’s chip exports up 259% in twenty days, Meta +11.3% on its agent’s app-store rank, Arm +14% and Intel +13% at midday, bitcoin through 86,000. Tuesday: the Nasdaq 100 at 747.46, its highest close of the year, the Composite at a second record close, AMD into the trillion-dollar club; the S&P flat. Wednesday: −0.72% on the S&P and −0.84% on the Nasdaq — a flash business survey showing the fastest US output growth in more than five years, Fed Governor Barr saying more increases are likely needed, the ten-year back above five percent. Thursday: flat — the thirty-year to 5.45%, Oracle’s force-majeure notice, Arm −7.9%, Meta +4.5% after its Connect event. Friday: +0.54% and +0.46% — Microsoft +3.6% on Copilot, Meta −3.3%, oil −3.5% on US–Iran talk. The week: the S&P +1.27% to 771.35, the Nasdaq 100 +3.19% to 744.50, the equal-weight S&P −0.56%, small caps −0.75%.

The eleven S&P 500 sectors

Four green, seven red — technology on top, the rate-sensitive groups at the bottom. Technology +3.52%, communications +1.94% (Meta’s week), health care +1.37%, industrials +0.40%. Then materials −0.38%, discretionary −0.42%, staples −0.89%, financials −1.83%, real estate −2.28%, energy −3.53%, utilities −3.87%. A week with the ten-year up 18 basis points sorted the sectors by one question: does the business grow faster than its discount rate rises? Technology does; utilities, real estate and financials pay for rates rather than earn from them; energy fell with its barrel.

The sector rankings — three lenses. Against the S&P over the week, technology led by 2.2 points and communications by 0.7; utilities lagged by 5.1, energy by 4.8, real estate by 3.5 and financials by 3.1. Over twenty-one sessions technology is the only sector clearly ahead of the index, by 6.7 points — health care −2.3, communications −0.4, energy −1.3, and everything else five to ten points behind: utilities −9.9, real estate −8.5, materials −7.9, financials −6.6, discretionary −6.3, industrials −6.2, staples −5.6. Over sixty-three sessions energy still leads, by 9.4 points, then technology +2.6 and health care +0.7; utilities trail by 20.3. The month belongs to technology outright; the quarter still belongs to energy, and energy’s week was its worst relative week in the table.

Underneath the focus sectors — technology held its breadth, financials did not. The dispersion pages count members against their averages. Technology: 62.5% of its 72 members above their 50-day average, 70% above the 200-day, 22 five-day highs against 15 lows. Health care: 56% above the 50-day, 71% above the 100-day, 20 highs against 9 lows. Financials: only 6 of 76 members — 7.9% — above their 50-day average, and 16 five-day lows against 8 highs; no new highs on any window longer than a week. Industrials: 16% above the 50-day. Energy: 1 of 21 above the 20-day. The index rose on technology; the financial sector underneath it is as washed out as it has been this year.

The Nasdaq 100, cut six ways

The top made new highs; the rest did not; the non-tech part fell to its low. The Nasdaq 100 can be bought in cuts, and this week the cuts told the story better than the index. Nasdaq 100 technology (QTEC) +5.64% on the week and +43.6% on the year. The thirty largest Nasdaq stocks (QTOP) +3.58% and +23.4% — a new high for the year on Tuesday at 39.55. America’s twenty largest companies (TOPT) +2.49% and +13.2% — closing Friday at 35.34, level with its high of the year. The index itself +3.19% and +21.2%. Then the rest: the seventy smaller Nasdaq 100 members (QNXT) +2.74% and +12.8%, 4.2% under their August high; and the Nasdaq 100 without its tech stocks (QQXT) +0.19% on the week and −2.5% on the year.

QQXT is the chart of the month. The non-tech Nasdaq — the retailers, the drug makers, the media, the travel and the utilities inside the index — made its 2026 high at 104.67 on August 19 and closed Friday at 96.92: −7.4% in five weeks, and 1.0% above its March 27 low of 95.94. From the year’s high to within a percent of the year’s low in a month, while the thirty largest members of the same index made a new high. That is what “narrow” looks like in one index: the leaders and the tech weights carry it, and the other half has given back its whole year.

The editor’s charts put the two ends side by side — the dramatic decline of non-tech growth against the top of the market. QQXT has spent all of 2026 in a box between about 96.2 and 103.8; it broke out of the top in August, failed within three weeks, and has fallen straight back to the bottom of the box — Friday’s 96.92 is seventy cents above the floor. On five years, the rising line from the 2023 low that carried the fund for three years broke in September. QTOP, the thirty largest Nasdaq stocks, closed at 39.43, a new high, on the rising line from its 2025 low. Same index, same week: the top at a new high, the non-tech half on the floor of its year.

The axis, five weeks on — past the limit this letter set. Four weeks ago this letter set a rule: tech beating its ex-tech cut for more than four weeks running turns a rotation into a concentration. This is the fifth week: QTEC +5.64% against QQXT +0.19%; the S&P without tech −0.25% against the S&P +1.27%. The equal-weight Nasdaq 100 rose 2.72%, so the average Nasdaq stock did take part — it is the non-tech half that did not. The limit was stated so it could be scored, and it is scored: this is a concentration, and the letter reads the index from here as the leaders’ index, not the market’s.

The factor read — momentum at the 99th percentile. The factor-regime gauge — US momentum (SPMO) against US low volatility (SPLV) — closed the week at 170.3 indexed, above its 50-day at 156.7, in the 99th percentile of its range: “momentum leading — risk appetite building”. The legs sorted the normal way: momentum +2.49% on the week and +28.2% on the year; low volatility −1.76% and −0.2% on the year. The defensive factor is now flat on 2026.

The tech ETFs we carry

Fourteen green of twenty-one — the chips first, then everything that uses them. The fabless chip fund +7.78%, the equal-weight semis +6.73%, the chip majors +5.86%, the WisdomTree AI fund +4.28%, the internet-of-things fund +4.16%, quantum +4.01%, ARK +2.88%, the Global X AI fund +2.87%, cloud +2.05%, software +1.59%, internet +1.20%, cybersecurity +1.08%, robotics +0.83%, the grid fund +0.74%. The red seven: digital transformation −2.39%, lithium −2.10%, defense tech −1.95%, uranium −1.77%, fintech −1.58%, data centers −0.71%, gaming −0.50%. Last week software beat the chips by two points; this week the chips beat software by four. On the year: the chip index +90.2%, cybersecurity +41.3%, cloud +25.5%, software +0.3% — green again, by thirty basis points.

Software over silicon, as one line. The software fund divided by the chip majors fell 4.0% on the week to 0.1748 — still 33.9% above its June 22 low, but the week went to the chip.

The four indices: Dow, S&P 500, Nasdaq 100, Russell 2000

The big-tech index led, the old-economy index barely moved, the small caps fell. The Nasdaq 100 +3.25% to 30,608 — +21.2% on the year, 0.4% under Tuesday’s high of 30,732. The S&P 500 +1.21% to 7,743 — +13.1% on the year, 0.7% under its August 13 high. The Dow +0.28% to 51,829 — +7.8% on the year, 4.6% under its August 5 high. The Russell 2000 −0.80% to 2,838 — +14.3% on the year, 7.5% under its August 14 high. Two indices near their highs, two well under them, and the dividing line is the same as in the Nasdaq cuts: the companies that sell the AI economy against the companies that pay its interest rate. The small caps borrow at floating rates; the Dow carries the banks and the industrials; neither had a week in which the ten-year rose 18 basis points.

The Treasury curve, one month to thirty years

The whole curve moved up, and over the month the short end moved most. The yields at Friday’s close, with the week’s change: 1-month 3.96% (+8 bp), 3-month 4.18% (+10), 6-month 4.37% (+10), 1-year 4.48% (+7), 2-year 4.91% (+16), 3-year 4.94% (+10), 5-year 5.03% (+17), 7-year 5.08% (+16), 10-year 5.17% (+18), 20-year 5.56% (+18), 30-year 5.47% on the government-bond series, 5.50% on the CBOE index (+14). Over the four weeks since August 28: the 2-year +56 basis points, the 10-year +46, the 30-year +26 — the curve flatter after the Fed’s hike, the front end pricing the second increase Governor Barr talked about. Two shapes to note: the 20-year yields more than the 30-year, a hump that widened this month; and the 5-year crossed 5% this week, which puts every maturity from five years out above five percent.

The veto, scored at the close — the seventh failure on the belly, and the long end lost its line. IEF 91.15, 91.16, 90.19, 89.69 — a new fifty-two-week-low close on Thursday — 90.00. Seventh consecutive week under the 93.17 reclaim line. TLT 81.80, 81.75, 80.46, 79.42, 79.32 — the 81.2 line it held by a nickel last Friday lost on Wednesday, two new lows to finish. The editor’s five-year chart of the long-bond fund shows the floor near 79.9 that held in 2022, 2024 and 2025 breaking for the first time since the autumn of 2023; the next level is 74, the October 2023 low, 6.7% under Friday’s close. Last week the letter called the bond veto engaged on the belly and a dissent on the long end; this week it is engaged on both, and the equity leaders rose through it.

The dollar

A stronger dollar, on a stronger economy. The dollar index closed at 101.13, up 0.9% on the week, and the dollar fund made its highest close of the past year on Thursday — +5.9% on 2026. In a week in which Japan’s ten-year rose to 3.07% and France’s spread over Germany widened to 113 basis points, the US paid more because it is growing and the others paid more for other reasons; the currency took the growth reading. A stronger dollar tightens conditions for everyone who borrows in it, and it is the reason the Asian chip exporters’ dollar funds and Europe’s dollar funds rose less than their home markets; Saturday’s Global letter has the geography.

The cross-asset split, at tape altitude — bitcoin with the leaders, gold with the bonds. Bitcoin closed Monday at 86,603, through the 83,000 line this letter carried last week, and finished the week near 84,000; the bitcoin fund +3.37%. Gold’s fund −1.93% to 393.41, silver −2.99%. In a rising-rate, rising-dollar week, the coin traded like a growth asset and the metal like a bond.

2 · The State

The mechanism, named: a growth rate rise, not a squeeze — and the market paid the leaders for it. The yields rose on a business survey and a Fed governor’s sentence, not on an inflation scare or a funding accident. That is the kind of rise equities can carry, and this week the ones that carried it best were the ones with the most growth to discount: the chips, the megacaps with a new product to sell, the thirty largest Nasdaq names. The ones that could not carry it were the ones that owe the rate rather than out-grow it: the small caps, the banks, the utilities, the non-tech Nasdaq.

The applications turned up in the prices — Meta and Microsoft. Meta +12.9% on the week — +11.3% on Monday when its Muse agent reached No. 1 on Apple’s US App Store, then its Connect event on Wednesday naming Walmart, Best Buy, Gap, Sephora and Wayfair as shopping partners for the agent after Instacart on Monday, with new glasses at $449, $349 and $249 and a $1,299 headset; Amazon has blocked the agent from its store. Microsoft +4.5% — +3.6% on Friday on reports of a large Copilot expansion: coding tools, autonomous agents and direct access to Word, Excel and Outlook, after Stifel’s upgrade to Buy on Wednesday. At sector altitude, the reading is this: the first AI products the market is paying for as products, not as spending plans, belong to the platforms. The software fund rose 1.6%; the applications index that holds the companies selling seats on top of the platforms fell 0.1% on the week and 5.7% on the month. Intuit fell 9.0%, Adobe 5.4%, Salesforce 1.6%.

The Magnificent Seven broke out. The fund made a record close at 72.97 on Monday and held above 72 all week, 72.64 on Friday, +3.09%. The editor’s three-year chart puts it in its frame: a breakout from a ten-month box between roughly 53 and 69, through the 71.4 summer high. The seven inside it: Meta +12.9%, Microsoft +4.5%, Tesla +2.2%, Apple +1.5%, Nvidia +1.3%; Alphabet −1.6%, Amazon −1.6%. The hyperscaler cohort — Microsoft, Amazon, Alphabet — was −0.1% on the week as a group; the record was made by the consumer AI names.

The leaders, in the editor’s 2×2 — cloud, security, Meta, Microsoft. Cloud (CLOU) at 28.39, on a steep line from its spring low, with 29.3 and 31.7 the next marks above; cybersecurity (CIBR) at 100.95, through the 94.6 line it had pressed against since July; Meta at 751.66, out of the falling line inside the box it has traded in since early 2025, with 731 now the floor and the top of the box near 800; Microsoft at 516.17, back above 510, with 549, the 2025 high, the next line. Four charts, four breakouts or near-breakouts, and every one of them a company or a fund that sells the AI economy to its users.

Oracle — the one credit fact of the week. Oracle invoked force majeure on a data-center campus in New Mexico on Thursday; the stock fell 7.1% on the week to 137.10 — three red sessions to end it — and Arm fell 7.9% on the day it was reported. Oracle is the seller of compute this letter scored “sold” in its September 15 print window, at −8.3%, on a $664 billion backlog. A force-majeure notice is a contractual statement that something outside the company’s control prevents it from performing; which party it protects, and from what, is what the market will want to know. The builders sold on it for a day; the platforms did not.

The count — through the confirmation line on a close, and back under it on the week. The Nasdaq 100 fund closed Tuesday at 747.46, through the 746.16 confirmation line this letter has carried since August, for the first time on a close. It closed the week at 744.50, 1.66 under the line and 4.5% above its 50-day average at 712.65. 694 remains the kill-switch, 6.8% below; the 704 shelf that held twice last week is now 5.4% below. One daily close through confirmation is not a weekly confirmation; the letter’s standard has always been the weekly close.

3 · The Outlook

Read more

Read more

1 · This Week’s Action

The tape, day by day. Another quiet index week sat on top of a bond rout, an oil shock and a jobs report — and it moved a tenth of a percent. Monday −0.30% as gold’s Friday break hung over August’s last session and the dollar caught its bid. Tuesday −0.69%, the widest down day: Japan’s 10-year closed above three percent for the first time since 1996, the Bund at a 2011 level, the US 10-year through 4.78% on a sixth straight rising session — and, overnight, the US and Iran exchanged fire with two tankers reported hit near Hormuz, WTI +5% in two sessions. Wednesday +0.44% with the 10-year at 4.81%, its highest since November 2023, and oil through 90. Thursday +1.05%, the widest move of the week: Waller cut the September hike odds to a coin flip, Nvidia bought Hugging Face, the belly of the curve steadied. Friday −0.39% on a payrolls print of 162,000 against roughly 53,000 expected — the first jobs report of the Warsh chairmanship, three times the forecast, with 55,000 of upward revisions behind it — which put the September hike back above 58% odds, spiked the two-year to 4.425%, and split the tape by layer in the opposite direction from the whole month: chips +3, software −2.

Five sessions, and the S&P finished +0.11% at 770.19, the Nasdaq 100 +0.35% at 718.96, still above the 50-day (711.08) it reclaimed last week, and the equal-weight S&P −0.77% — the second consecutive week cap-weight beat equal-weight. Everything that happened this week happened underneath the index again. This week, for the first time in three, what happened underneath favoured the silicon.

The cross-asset board — oil owns it. The best line on the board is not an equity: USO +9.45% on the week, +23.6% on the month, +105% on the year — the Hormuz exchange put WTI above 90 and Brent near 96 by Friday. The coin fund +3.03% (IBIT 45.23) and the two top-heavy US cuts (+1.22%, +1.13%) are the only other green of consequence. The hard assets were sold: gold −0.52% to 406.77 after Thursday’s high at 410.22 and Friday’s payrolls-day fall; silver −0.33%; copper miners −3.98%, the worst line. Both Treasury funds red — IEF −0.65% to 92.25, TLT −0.81% to 82.21 — with the belly making a new year-to-date low on Tuesday. Saturday’s letter carries the sovereign board and the hard-asset charts in full; the one-sentence version for the tape: gold and bitcoin both rose to their resistance and the first attempt to break failed — the read stays constructive, and it stays unconfirmed.

The sector read — energy took the board, and the inversion stopped. Five green of eleven: energy +2.20%, technology +0.86%, utilities +0.82%, health care +0.17%, financials flat. Six red: communications −0.85%, staples −1.02%, industrials −1.06%, real estate −1.24%, materials −1.39%, discretionary −1.96%, the worst. For two weeks this letter recorded full inversions — each week’s leaders became the next week’s casualties. This week the pattern broke: technology, top-three last week, stayed green; health care, bottom-five last week, came back to flat; energy, quietly strong for a month on both lenses, finally led outright. A tape that has been auditioning leaders and dismissing them after five sessions kept two of them this time.

The sector rankings — relative strength, three lenses. Our sector-RS board reads the rotation’s speed, and this week it reads a leader by a distance: energy +11.7 points against the index over 21 days, +6.7 over 63 — the month’s leader, and the only sector whose month is stronger than its quarter by that margin. Health care keeps the quarter crown (+7.6 over 63) and repaired its month (+4.4) after fading last week. Technology: +0.7 on the month, −0.6 on the quarter — the tired leader that was −4.5 on the quarter a fortnight ago has had its quarter forgiven; it is now flat on every lens, which is not leadership but is no longer the drag. Financials +0.1 and +6.7: flat month, strong quarter. And the bottom two moved: industrials −6.0 over 21 days is now the worst line on the board, while utilities, last on every lens a week ago, improved to −1.4 on the month as the long end rallied. The rotation slowed. Last week it ran in both directions at once; this week it ran in one — toward the sector that a Hormuz headline prices first.

Underneath the four focus sectors — the breadth split along the same line as the tape. The dispersion pages show the members behind each ETF. Technology: 51% of members above their 50-day — still only half — but 28 five-day highs against 15 lows, a +13 net after last week’s balanced 14-against-15. The chip names printed highs on Friday while the software names printed lows: the sector’s internals firmed exactly where its ETF did. Health care: 76% above the 50-day, 83% above the 100-day — and 2 five-day highs against 16 lows, −14 net after −20 last week. The medium-term structure is intact; the week’s tape has now been against it twice. Financials 12 highs against 7 lows with 92% above the 150-day — the strongest structure on the board. Energy the paradox: 90% of members above the 50-day, and yet 2 highs against 11 lows on Friday — the sector led on the week and was sold on payrolls day. Industrials the mirror: only 24% above the 50-day, but 27 highs against 15 lows — a broken sector bouncing broadly. One week ago the internals rotated as fast as the surface. This week they sorted by the same axis the surface did: silicon over code.

The factor read — the repair landed. The factor-regime gauge retired last week’s label. For three weeks it read momentum repair attempt below trend, with the percentile bleeding from 94.6 to 92.3 to 89.7. This week: the momentum-over-low-vol spread indexed at 159.08 against a 50-day at 155.92 — above trend — at the 94.18th percentile, and the state reads “momentum leading, risk appetite building”. The pair behind it: SPMO +2.00% against SPLV −0.45%, the twenty-day rates of change now +1.99 against −1.93, the sixty-day +3.02. Saturday’s letter found the international pair less decided — both ends rose, min-vol made a high on Thursday and gave it back. At home the style axis, which went quiet during the violent layer rotations of the last fortnight, moved this week — and it moved toward risk, on a hawkish jobs number, in the week the belly of the curve made a new low. That is the factor market saying the growth in the payrolls print matters more to it than the discount rate in the two-year. It is a reading, not a verdict; the gauge has retired labels before.

The axis, one week on — the everything-else trade trailed again, and the tech trade split down the middle. Two weeks ago the ex-tech cuts led at three-year highs; last week they gave back; this week they trailed a second time: the Nasdaq 100 ex-technology −1.39% against the Nasdaq 100’s +0.35%; the S&P ex-tech −0.39% against the S&P’s +0.11%. The year-to-date order still reads the other way, and the second pillar’s count — the number of consecutive weeks tech beats its ex-tech cuts — stands at two, against a stated limit of four. Run the pairs and one more thing shows: the Nasdaq 100’s own equal-weight technology sleeve fell 1.23% while the cap-weighted index rose. Tech beat everything-else this week, and inside tech, three names did the beating.

Because the money did not go where last week’s rule sent it. Last week the rule read “the market pays the layer, not the beat”, and it paid software: IGV +5.93%, through 107.70, green on the year. This week: software −4.50%, the worst line on the tech board but one — and the silicon it had been beating for a fortnight took the top of the board: SMH +2.51%, SOXX +2.21%, XSD +1.99%, SMHX +1.57%. IGV’s week, session by session, is the whole letter: Monday 109.98, holding the breakout; Tuesday −3.46% to 106.18, back under the 107.70 line it had broken the week before; Wednesday 103.42, under the year; Thursday +3.41% to 106.95 — green on 2026 again, intraday and at the close; Friday −2.23% to 104.57, red on the year by 1.06%. A breakout that lasted one weekly close, a retest that failed on the second session, a year regained on Thursday and given back on Friday. The pair board dates the turn: the software-to-semis ratio fell 6.84% in the week — from 52% above its June 22 low to 41% above it, a full month of the ratio’s rise surrendered in five sessions, most of it on Friday.

Underneath it, the participation statistic held for a second week — and changed its composition entirely: the Magnificent-7 basket +0.54% against equal-weight’s −0.77%, cap-weight above equal-weight in both indices again. But last week’s re-concentration was into the balance sheets — Microsoft +6.3%, Amazon +3.0%, Alphabet +0.5%. This week those three were the basket’s worst lines — Microsoft −2.69%, Amazon −2.97%, Alphabet −2.35% — and the basket was carried by two names with a chip thesis: Nvidia +5.89% and Meta +6.70%. The re-concentration is still on, and it is no longer a defensive crowding into size. It is a crowding into the two largest expressions of the silicon trade.

Inside tech — the split flipped, and it flipped to the chip. Thirteen green lines of twenty-one, led by the digital-asset fund +7.45% (the coin’s week), the AI-themed basket +3.20%, and the four semiconductor wrappers — SMH +2.51%, XSD +1.99%, SMHX +1.57% — with nuclear, fintech and the internet-of-things fund green behind them. Eight red, and the layer above the chip is all of them: defense tech −4.85%, software −4.50%, cybersecurity −4.03%, cloud −3.64%, then lithium, internet, quantum and grid. Last week: five green, all the layer above the chip; sixteen red, led by the silicon. This week: the exact inversion at board level. The three-week software-over-semis trend that this letter promoted from “tell” to “axis” seven days ago has been interrupted in one week — not overturned on the year, not overturned in the ratio’s two-month structure, but interrupted in the most emphatic single session of the summer.

2 · The State

The sorting rule broke — and the replacement is sharper: the market pays the number, not the beat and not the layer. Fourteen scoring windows closed this week, the record’s largest single week of the season. Seven paid: Nvidia +5.3% (the first paid Nvidia window since May 2024), Synopsys +7.2%, CrowdStrike +22.1%, Salesforce +25.3%, Elastic +10.3%, Dell +23.3%, GitLab +10.5%. Six sold: Marvell −12.9%, Autodesk −8.5%, IREN −9.2%, Rubrik −17.4%, MongoDB −15.1%, Palo Alto −8.0%. One flat: Workday +2.5%, one dollar under the paid line. Sort those by layer and the rule of last week fails on the first pass: three hardware and infrastructure names were paid (Nvidia, Dell, Synopsys), four software names were sold (Autodesk, Rubrik, MongoDB, Palo Alto). Sort them by the size of the beat and it fails again: MongoDB’s was its largest EPS beat in the covered record, Rubrik’s a raised guide, Palo Alto’s a seventh straight double beat — all sold. What sorts them is acceleration: whether the print showed a number still speeding up against a price that had already paid for the last speed. Dell’s $7.04 against $4.91, a 43% surprise on revenue up to $46.97 billion — paid 23%. Broadcom’s faultless quarter with a $34.8 billion guide that merely met the higher of two consensus numbers — sold 6.8% at Thursday’s low, closed −2.7%. Friday’s chart pick wrote the rule in one line: the market is not paying for beats; it is paying for growth that accelerates, and punishing anything that merely met a price already paid.

Broadcom is the rule’s exhibit for the week ahead, and its window is open. EPS $3.32 against $3.22, revenue $29.59 billion against $29.24 billion, AI-semiconductor revenue $16.7 billion, up 221% from a year earlier, against its own $16 billion target — and the stock was sold on the guide until the call put AI revenue at $115 billion for fiscal 2027 and $230 billion for 2028. It closed Thursday at 357.16, a dollar above the level that grades its window as sold, and Friday at 357.89, −2.95% on the week. The window scores at Tuesday’s close from a 367.24 anchor: Broadcom needs +2.6% by Tuesday to score flat, and the tape has to do it through Labor Day. A card that has flipped sign every quarter for a year — +10.0%, −16.0%, +8.9%, −17.2% — arrives at its next verdict with the whole chip complex having just risen 3.5% in a day.

The jobs number, taken seriously — because the tape took it two ways at once. Payrolls +162,000 against roughly 53,000; July revised from −23,000 to +21,000; unemployment 4.1% with participation rising to 61.6%. The rate market read it the obvious way: hike odds for September 16 back above 58% from Thursday’s coin flip, two-year 4.425%, dollar up, gold −1.4% toward $4,400, five of the seven largest technology stocks down. That is the discount-rate effect, and it hit everything whose value sits in the distant future — software and cloud −2, the long-duration growth lines, the ex-tech cuts. But a jobs report is also a demand statement, and the market put semiconductors on the cyclical side of that line: the Philadelphia index +3%, Nvidia toward its record, Applied Materials +5%, Astera Labs +11%. The same number raised the discount rate on the code and raised the near-term earnings expectation on the silicon. Friday’s split is not a rotation between two AI layers. It is the market grading one number twice, and the grades disagreeing by layer.

The veto, scored at the close — no fourth attempt; the belly was sold outright. Three weeks running IEF pushed above 93.17 midweek and was sold back by Friday, each push from higher than the last. This week it made no push at all: 92.74 Monday, 92.10 Tuesday — a new year-to-date low — 92.18, 92.28, 92.25 Friday. The 10-year touched 4.81% on Wednesday, its highest since November 2023; the 30-year finished at 5.247% from 5.207% with its buyback window starting Tuesday. TLT held its floor by a smaller margin — 82.21 against 81.2. Saturday’s letter carries the Pacific half, and it is the more interesting half this week: Japan’s 10-year closed above three percent for the first time since 1996 on Tuesday and was back under it by Friday, and Japan’s 30-year rallied — 3.970% from 4.125% — as the BoJ’s Takata talked “continuous” hikes and the yen went from through-160 to 156 in three sessions. The bond market has now declined to sign the equity resolution four weeks running. This week it did not even show up to argue.

The macro print was a number, and the number turned the September meeting into a hike question again. On Thursday Waller had the market at a coin flip for September 16. On Friday the payrolls put it back above 58%. Between them the yield curve did its own work: the belly at a year low, the long end selling and then steadying, oil through 90 on a shooting exchange in the Gulf that Saturday’s letter owns. The rate-hike-era pattern this letter identified on Warsh’s keynote — balance sheets bid, capex charged, hard assets sold — did not print on Friday. The balance sheets were sold (Microsoft, Amazon, Alphabet all red on the week), the capex layer was bid (Rubin +2.5% on Friday alone), and the hard assets were sold. One session is one session; but this session contradicted the template of the last one, and the only consistent reading across both is the one the print record gave: the market pays what is accelerating, and on Friday the accelerating thing was the economy.

The count — above the average, both lines live, neither approached. The Nasdaq 100 closed 718.96 against a 50-day at 711.08 — 1.1% above, from 0.6% last week. The map is unchanged: 1-2 off the April low, wave 3 ongoing, 746.16 the confirmation (3.8% above), 694 the kill-switch (3.5% below). The week traded a 707.64 low on Tuesday and a 718.96 high on Friday’s close; neither line was in play. Saturday’s letter carries the five-year count on the world ex-US, which took its record close on Friday — wave 3 progressing, the three-week postponement over. The two counts still agree, and the domestic one is now the laggard: the world ex-US confirmed, the Nasdaq 100 is still 3.8% from doing so.

The same chart at four focal lengths. The house grid runs the count through its windows: the five-year view carries the wave structure — 1 and 2 complete, wave 3 ongoing inside its channel; the year-to-date view shows this summer’s 1-2 sequence against the 746 line, the 2 low held; the one-month view frames the index under the descending line from the mid-August high — Friday closed exactly on that line at 718.96, with the 701 shelf untested below; and the five-day view shows the week’s own micro-structure: Tuesday’s bond-rout gap down to 707, Thursday’s jump back to 718, and two sessions pressing against a flat 718 top. Four windows, one reading: an advance pausing at descending focal lengths — and the two smallest windows now sit at their line, not below it. A close above 718 on Tuesday resolves the month’s line upward; the count above does not need it, but the week ahead will read it.

3 · The Outlook

The four indices — the physical layer separated from the pack, and it did it on Friday. The family printed the week’s split in one row: Rubin Build-Out +0.40% (the capex layer, +88.1% on the year — the only green line, and green only because of Friday’s +2.5% session: 1886.14 to 1933.47), HALO −2.02% (broad growth, no AI thesis; +4.0%), Agentic Winners −2.30% (the applications; back under water on the year at −2.4%, from −0.12% a week ago, one week after this letter wrote “one ordinary week from a green year”), Agentic Ecosystem −2.39% (the opex layer; +57.1%). Across the Atlantic Euro-AI −3.4% through Thursday, its defense cluster −6.9% and its semiconductor cluster the only green sector at +0.2%. Last week the lightest layer led for a third week and the capex layer was the casualty. This week the lightest layer was the casualty and the capex layer led — by one session.

The control group did not flip this time, and that is the week’s most important internal. For two weeks HALO and the AI layers alternated: an AI unwind, then an AI bid inside a growth sell. This week HALO −2.02%, applications −2.30%, opex −2.39% — all three within forty basis points of each other, all three red — and only the physical layer apart. This was not an AI-versus-growth week. It was a duration week: everything priced on next year’s earnings fell about two percent together, and the one layer priced on this year’s shipments rose. Inside HALO the ladder says the same: Latin America +2.18%, space +2.08%, longevity +1.80% green; destination economy −6.27%, autonomous defense −5.84%, surgical −3.96% red — four green of twenty-three.

Inside the indices — where the week actually happened, and where the month is happening. Inside Rubin the top five are the physical stack, every one: Storage +11.04% (+357% on the year), AI Factory Systems +5.51%, DC Power & Electrical +5.17%, AI Factory +4.93%, Thermal Management +4.79% — twenty-two of thirty-six sub-indices green. The bottom is the buildout’s software: EDA & chip IP −7.76% — last week’s best line — high-speed interconnects −6.12%, wafer processing −4.94%, design −2.15%. Widen the window to the month and the old leaders are back on top of the whole tracker: Storage +17.1%, HBM Memory +15.8%, AI Factory Systems +11.4% lead the thirty-six on a one-month view, with wafer processing −20.3%, fab subsystems −14.2% and interconnects −10.3% at the bottom. The memory-and-storage complex that spent August draining from the highest flow scores on the board — SanDisk, Kioxia, Micron — is the month’s leader by price. Inside the Agentic Ecosystem the bid went to the operators for the first time in four weeks: compute operators +3.69%, the layer’s best line, Asia constituents +2.20%; at the bottom, operations & observability −7.93% — the worst sub-index in the entire four-family system, edge −3.99%, data & memory −3.82%, govern & secure −3.62%. Inside Agentic Winners, two green of nine: Asia +0.95%, megacap gateway +0.61%; control plane −4.36% and Europe −4.36% worst. Thursday’s Daily Pulse argued the next scarcity is permission to act — the governed, observable, reversible path from an enterprise’s own data to a transaction. The market spent this week selling exactly the sleeves that thesis names: observability, control plane, governance. A thesis and a tape disagreeing this cleanly is either an entry or a warning; the print record above says which names the market is still willing to pay in that layer, and it is a short list.

Best and worst week, by name — the mirror, fourth week running, and it favoured the silicon. Across the family’s constituents, the five best: IREN +26.0%, Bloom Energy +20.0%, Soitec +18.8%, SanDisk +17.2%, Dell +14.9% — then Appier +14.5%, Kioxia +13.7%, Viasat +11.8%, Modine +9.6%, Vertiv +9.1%, Micron +9.0%. Nine of the eleven are capex names — storage, power, cooling, the operators — the exact inversion of last week’s list, which was nine software and application names with not one chip. The five worst, with Amphenol excluded from the list as a split artefact in the worker’s data this week: Credo −26.7%, Unimicron −18.7%, MongoDB −17.4%, UiPath −16.3%, Axon −14.2% — then Cadence −14.0%, JFrog −11.3%, Synopsys −11.0% (after its paid window closed Monday), Fastly −10.6%, Palo Alto −10.3%, PTC −10.3%. Fourth consecutive week in which the prior week’s hero list leads the casualty list — Synopsys and Cadence were the EDA leaders seven days ago; IREN was the third-worst name in the system. The medians: capex +0.72%, opex −2.09%, HALO −2.15%, applications −2.38%. Green counts: 67 of 126 in the buildout, 30 of 96 in HALO, 12 of 34 in the opex layer, 11 of 40 in the applications.

The AI Handoff Board — the up-stack run stopped, and the operators turned. The handoff ratios had three consecutive weeks of the stack handing value up. This week: use-against-build −2.70% to 1.5967, beyond-gateways −4.10%, use-against-operate flat (+0.09%) — the applications gave back against the builders. The sharpest line is inside the buildout: verification-against-design +9.52% — test and metrology re-rating against the EDA names that led last week. And the line this letter has watched for a month reversed: operators-against-suppliers +3.26% after −4.79% a week ago — the neoclouds rising against the chipmakers that supply them, for the first time since the capacity trade was marked down in August. Trust-against-execution −0.23%: the security spike of last week went flat, and its names (Palo Alto, Zscaler) were sold.

The hyperscaler cohort — through the shelf on Thursday, back under on Friday. The Mag Pulse board reads the basket at 69.44, +0.54% on the week — 0.09% below the 69.5 shelf that has capped it since spring. It closed through the shelf on Thursday at 70.43 — the first close above it since spring — and was sold 1.41% on Friday to six cents under. Underneath, the composition inverted from last week: Meta +6.70%, Nvidia +5.89%, Tesla +1.53%, Apple +0.08% — and the three hyperscalers red: Alphabet −2.35%, Microsoft −2.69%, Amazon −2.97%. Last week’s re-concentration was the rate-hike-era playbook — the companies that fund themselves. This week the same basket beat equal-weight by 1.3 points while its self-funding trio fell: the money went to the chip and the chip’s largest customer. The shelf test is now a daily-close test; a weekly close through 69.5 has been attempted once and refused once. The December pivot at 62.56 is not on the map.

Compute tightness — the operators’ first green week in four, on the number. For three weeks this letter recorded the neoclouds marked down against the most supply-constrained guide of the cycle. This week: Nebius +8.23%, CoreWeave +3.65%, IREN +26.04% — the best name in the entire system — the compute-operators sub-index the best line in its layer at +3.69%, and the operators-against-suppliers ratio +3.26%. The demand statement did not change; the funding backdrop, on paper, got worse (the belly at a year low, hike odds up). What changed is the payrolls read: a stronger economy is more usage, more inference, more rented compute. The market paid the capacity traders on the cyclical reading of the same number that sold the software. That is one week against three. The book’s tilt watches the ratio, not the week.

The financing architecture underneath it — the special analysis is in production. Last week this letter announced a separate Closelook analysis of the August 10 platforms — the $500 billion of third-party capital that Nvidia, Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR are mobilizing to make AI factories an investable asset class, with Nvidia reportedly standing behind residual GPU values for up to a quarter of a project. The week added the other half of the picture: Nvidia agreed on Thursday to buy Hugging Face for $12.93 billion — roughly 86 times its revenue — nine months after paying about $20 billion for Groq’s assets, and disclosed supply and capacity commitments of $279 billion, from $119 billion three months earlier. Friday’s Daily Pulse read the two purchases as one ecosystem: the chips that serve existing models and the platform where open models are shared, bolted onto the supplier that finances the factories. The analysis publishes separately on closelook.net; this letter’s interest is narrower — a supplier that owns the model marketplace, the inference silicon and the financing vehicle is a supplier whose customers’ equity is being priced around its balance sheet, and this week the market priced those customers up.

Structural inflation — contained, unchanged, while the curve did everything else. The house composite in the macro lab reads 49, “Contained”, direction flat, structural character “Mixed” — unchanged from last week to the point. A supply-shock impulse registers (oil, Hormuz) and the gauge’s own note says it is not spreading to breadth or expectations. Hold that against a two-year at 4.425% and a September hike above 58%: the market is pricing the Fed on growth, not on inflation, and the gauge agrees with the market’s decomposition. If CPI on Friday disagrees, this paragraph changes first.

Breadth and temperature — narrowing, cooling, then warming into Friday. Equal-weight lost 0.77% against cap-weight’s +0.11% — the second cap-weight week — while the Money Temperature composite went 57 to 49 midweek to 55 on Friday, transition band, the state reading “risk-on rally”. A gauge that cooled eight points into Wednesday’s bond rout and recovered six of them on a hawkish jobs print is reading the same thing the factor pair read: appetite returned on Friday, and it returned narrowly — into the silicon, into the operators, into two megacaps.

The count, restated for the week ahead. Above the 50-day by 1.1%, 3.8% from confirmation, 3.5% from the kill-switch, with a holiday Monday, five scoring windows Tuesday and Wednesday, Oracle Tuesday night, Adobe Thursday night, the ECB Thursday, CPI Friday and the Fed the following Wednesday. The referees are booked, and this week they are macro referees.

4 · What May Lie Ahead

Read more

1 · This Week’s Action

The tape, day by day. The quietest index week of the summer sat on top of the loudest rotation. Monday −0.29% while the hard assets started giving back their Warsh-week gains. Tuesday +0.32% — the day the world ex-US index printed a record close and both min-vol funds made highs they would not hold. Wednesday +0.02%, the market flat to the cent into Nvidia’s print after the bell. Thursday +0.66%, the widest move of the week, as the supply-capped guide was paid: Nvidia +8.74%, the software beats from Wednesday night paid alongside it. Friday −0.23% on Warsh’s first keynote as chair — the hard assets sold, Thursday’s hardware beats sold harder, and the old-guard megacaps caught a bid against the whole tape.

Five sessions inside a one-percent band: the S&P +0.47% to 769.35, the Nasdaq 100 +0.42% to 716.43 — back above the 50-day average it closed on to the cent last week — and the equal-weight S&P −0.44%. The VIX fell 4.6% to 14.43 in a week the market sold clean beats. Everything that happened this week happened underneath the index again. For the first time in five weeks, what happened underneath favoured the top of it.

The sector read — the inversion, second week running. Three green of eleven, and they are last week’s bottom: communications +1.43%, technology +1.30%, financials +1.08%. Last week’s green three — health care, energy, materials — are in this week’s bottom five: XLV −1.98%, XLE −1.51%, XLB −0.67%. Utilities flat, staples and discretionary red, real estate −1.33%. Last week the sentence was “a board with the capex taken out and the non-tech growth put in”. This week the board put half the tech back in — the half that writes code — and took the non-tech growth back out. Two full inversions in two weeks is not leadership. It is a tape auditioning leaders and dismissing each after five sessions.

The sector rankings — relative strength, three lenses. Our sector-RS board reads the rotation’s speed. Technology, tagged Weakening last week with a negative 21-day read, has turned: +1.9% against the index over 21 days — the tired leader is repairing — though still −4.5% over 63: the quarter has not forgiven it yet. Health care keeps the quarter crown (+12.8% over 63 days) but the month is fading (+0.9%) — the same shape its breadth prints below. Energy +2.6% and +9.7% stays quietly strong on both lenses; financials −1.8% on the month against +10.9% on the quarter; materials near flat on both. And utilities is collapsing on every lens: −8.1% over 21 days, −5.5% over 63 — last on the board, in the week the long end rallied. The rotation is not just fast. It is now fast in both directions at once — tech repairing while its replacement leaders fade.

Underneath the four focus sectors — the breadth flipped with the boards. The dispersion pages show the members behind each ETF, and this week they show the rotation reaching the internals. Health care: still 76% of members above their 50-day — but 4 five-day highs against 24 five-day lows, a −20 differential, the exact inversion of last week’s 23-against-3. The sector that was “broad, not carried” spent this week broadly retreating. Technology: 50% above the 50-day, 14 five-day highs against 15 lows — balanced, no longer the narrowest sector on the board, its members firming as the ETF led. Materials 3 highs against 8 lows; financials 11 against 18 with 88% still above the 100-day — the medium-term structure intact, the week’s tape against it. One week ago the strong sectors were strong underneath and the largest sector was narrow underneath. This week the largest sector is the balanced one and last week’s leaders are printing lows. The internals rotate as fast as the surface now.

The factor read — repair attempt, below trend, percentile still bleeding. The factor-regime gauge holds its label from last week — momentum repair attempt below trend — and the numbers say the repair has not landed: the momentum-over-low-vol spread indexed at 155.25 against a 50-day at 157.22, still below trend, at the 89.7th percentile of its distribution — down from 92.3rd last week and 94.6th two weeks ago. The pair behind it went quiet rather than violent: SPMO −1.31% against SPLV −0.31%, the twenty-day rates of change now +2.05 against −1.51 — momentum winning the month — while the sixty-day still reads −5.36 against +4.71, the defensive leg owning the quarter. Saturday’s letter found the identical shape internationally: momentum three weeks under its failed breakout, min-vol making highs on Tuesday and fading them by Friday. On both continents, both ends of the factor spectrum keep refusing the baton. In a week when sectors and layers rotated violently, the style axis went still — which says the rotation is running through what companies do, not how their charts behave.

The axis, one week on — it inverted, and the destination is the story. Last week this letter re-drew the tape’s axis from software-vs-semis to tech-vs-everything-else, with the ex-tech cuts at three-year highs. One week later the axis inverted: the Nasdaq 100 ex-technology gave back 1.35% from that high while the Nasdaq 100 gained 0.42%; the S&P ex-tech lost 0.44% against the S&P’s +0.47%. The everything-else trade lasted exactly as long as every other leadership this tape has auditioned — five sessions.

Run the pairs explicitly, because each one carries its own sentence. Nasdaq 100 ex-tech against the Nasdaq 100 tech sleeve: −1.35% against +0.62% — the ex-tech cut that made a three-year high last week trailed its own tech sleeve by two points this week. S&P 500 ex-tech against the tech sector: −0.44% against +1.30% — same signature at index scale. The year-to-date order still reads the other way (the ex-tech cuts trail their parents on the year), which means this week did to the ex-tech trade exactly what the previous week did to tech: interrupted it without overturning it.

But the money did not go back to where it left. It went to two specific addresses inside tech, and to neither of the others. Software: IGV +5.93%, the best line on the tech board, through the 107.70 resistance that had capped it since June 1 — and, at +3.60%, green on the year for the first time since the winter. The last major sleeve underwater for 2026 surfaced this week. The pair board dates the move: the software-to-semis ratio bottomed on June 22 and is 52% off that low — a two-month trend this week extended, not a one-week bounce. And the balance sheets: the Magnificent-7 basket +2.66% against the equal-weight’s −0.44%. The silicon got neither half: XSD −3.29%, the worst tech line a second straight week, SMH −1.30%, the chip index within one percent of its floor. The axis is no longer tech-vs-everything-else, and it never went back to software-vs-semis as an internal affair. It is now the layer axis: code and balance sheets in; silicon, operators and last month’s defensive growth out.

Underneath it, the participation statistic that defined the last month broke: the defaults beat the average stock for the first time in five weeks — the Mag-7 basket +2.66% against equal-weight −0.44%, cap-weight above equal-weight in both indices. For four consecutive weeks this letter tracked the largest names trailing a resilient average stock and asked whether it was rotation or distribution. This week supplied the answer neither reading predicted: re-concentration. The money that left the defaults came back — on a hawkish keynote, into the names that finance themselves.

Inside tech — the split is now the whole board. Five green lines, all of them the layer above the chip: software +5.93%, cybersecurity +3.91%, cloud +2.82%, internet +2.69%, AI applications +1.25%. Sixteen red, led down by the silicon and its adjacents: XSD −3.29%, DAPP −2.21% (with the coin’s fade), ARKK −1.88%, fintech −1.60%, nuclear −1.52%, quantum −1.49%. Two weeks ago the split was one week old and this letter called it a tell. It is now a three-week trend with a dated bottom, a broken resistance, and the year’s first green print on the software sleeve. The layer axis runs through this board most visibly of all.

2 · The State

The sorting rule found its axis: the market pays the layer, not the beat. Last week the rule narrowed to “nothing gets paid”. This week it split cleanly in two — by layer, not by quality. The software prints were paid, and paid large: Okta +23.0% on the week on a double beat with raised guidance, Salesforce +22.4%, Elastic +16.3% (+19.3% in Friday daylight alone), CrowdStrike +13.8%, ServiceNow +12.6%, Workday +5.8% on its own beat-and-raise. The hardware and infrastructure prints were charged, whatever they reported: Rubrik, a clean beat-and-raise, sold 13.1% in Friday daylight; Marvell −10.3%; IREN −12.5% — and Fabrinet, the entry-premium exhibit of two weeks ago, bled a further 5.1% without printing anything. Same week, same tape, same quality of report; opposite verdicts, sorted entirely by what layer of the stack the reporter lives in.

Nvidia’s print is the rule’s largest exhibit. The eleventh consecutive double beat — revenue $96.0 billion against roughly $92 billion — with a fiscal-2028 outlook the CEO framed as supply-capped: demand above 70% growth, shipping at 70%. The market’s answer, in two sessions: paid +8.74% on Thursday to 227.98, half refunded on Friday, −4.57% to 217.55. A +1.32% week for the guide that would have been worth ten percent in April. The scoring window closes at Monday’s close, and it closes on a number the market has already argued itself into and half out of. On this record the printed columns were foregone; the two-day reaction shape — pay, then refund half on the macro — is the new information, and it rhymes with everything else the week did.

The 70% number, taken seriously. Fiscal 2028 runs roughly across calendar 2027, so what Wednesday’s call delivered was a growth forecast for next year, given tonight — and its construction is the unusual part. Huang’s exact framing: “Even though our demand is much greater than 70%, our supply allows us to confidently deliver 70%.” That is not a demand forecast with execution risk; it is a shipping schedule with demand overflow — the cap is the fabs, the packaging and the power, not the order book. The CFO added that customer forecasts “point to our growth doubling next year.” Read against this letter’s four layers: a supply-capped 70% year at the supplier is the strongest statement the capex thesis has ever received — and the market’s response was to pay the applications layer, half-refund the supplier, and keep selling the operators who rent the constrained compute. Either the market is looking through 2027 to what comes after the catch-up, or it is repricing the discount rate on everything that needs financing before it gets there. Friday’s keynote argues for the second reading; Monday’s scoring close and Wednesday’s Broadcom print will grade it.

The chart pick, scored and staged. Robinhood, published two Fridays ago at 95.10, scored +9.6% at its one-week checkpoint — recorded in the edition’s own scoring note, next checkpoint late September. Friday’s third edition published Synopsys pre-open with the machine block showing its Directional Flow still negative and accelerating down against a seven-driver fundamental repair — and the stock promptly fell 4.8% in the daylight session, the entry printing below the reference. The record keeps both: the pick that ran and the pick that opened red. That is what a published track record is for. And one lane of the thesis got independent confirmation this week: the EDA sub-index inside the buildout was its best line at +5.98% — the design-software lane leading the index it sits in, the same week its flagship name was charged. Layer paid, name charged: the rule again, one level down.

The veto, scored at the close: failed a third time — from the highest push yet. IEF spent midweek comfortably above the line — 93.51 on Tuesday, the strongest reclaim attempt of the three — and closed Friday at 92.85. Three consecutive weeks now: push through 93.17 midweek, sold back by Friday, each push from higher than the last. TLT held its floor again (82.88 against 81.2, +1.01% — the best line on the cross-asset board). The long end actually rallied into the buyback window — the 30-year finishing at 5.207% from 5.275% — while the belly kept getting sold on Fridays. The bond market has now declined to sign the equity resolution three weeks running, each refusal more emphatic in its intraweek shape, and Saturday’s letter adds the Pacific half: the yen closed through its wire for the first time, and Japan’s 30-year printed another era high. The veto stands, and its company got worse.

The macro print was a podium, not a datapoint. Warsh’s first Jackson Hole keynote as chair: the 2% vow, prices first, the bet that AI aids labor supply — with Hammack calling for hikes the same day and the July minutes’ three dissents behind them both. The direction had exactly one session to trade, and that session sold gold 1.9%, the miners near four, and bid the dollar and the balance-sheet megacaps. One session is one session; the question it opens — blip or September template — is section 4’s, and the market’s first answer was the rate-hike-era playbook to the letter: Amazon +3.97%, Alphabet +1.74%, Microsoft +1.68%, Apple +1.63%, Meta +1.21%, while Nvidia gave back 4.57% and the beat-and-raises were sold. When the discount rate leans up, the market buys the companies that fund themselves and sells the ones that need the capex cycle. Friday was a clean specimen.

The count — back above the average, both lines still live. The Nasdaq 100 reclaimed the 50-day it closed on to the cent last week: 716.43 against 712.06. The map is unchanged — 1-2 off the April low, wave 3 ongoing, 746.16 the confirmation (4.1% above), 694 the kill-switch (3.1% below) — and the week moved the index from the average to just above it without approaching either line. Saturday’s letter carries the five-year count on the world ex-US — wave 3 still progressing, the stall reading as a pause — and the two counts agree: postponement inside an advance, not a topping structure. Both counts also carry the same caveat: the falsifiers are close enough to answer quickly.

The same chart at four focal lengths. The house grid runs the count through its windows: the three-year view carries the wave structure — 1 and 2 complete, wave 3 ongoing inside its channel; the year-to-date view shows this summer’s 1-2 sequence against the 746 line; the one-month view frames the index between 746 above and the 701 August shelf below; and the five-day view shows the week’s own micro-structure — an ascending triangle pressing against the 713 line, resolved upward into Friday’s close at 716. Four windows, one reading: an advance pausing at descending focal lengths, with the smallest window the only one that resolved this week — and it resolved up.

3 · The Outlook

The four indices — the money went back into the thesis, and picked its lightest layer. The family printed the layer axis in one row: Agentic Winners +4.68% (the applications — all nine sub-indices green, and at −0.12% YTD one ordinary week from going green on the year for the first time since January), Agentic Ecosystem +1.59% (the opex layer; +60.9%), Rubin Build-Out −1.77% (the capex layer; +87.1%), HALO −2.25% (broad growth, no AI thesis; +6.0%) — with Euro-AI +0.4% across the Atlantic, +32.9% on its year.

The control group flipped, and that is the week’s most important internal. Last week HALO held flat while the AI layers bled — an AI unwind, not a growth unwind. This week HALO fell 2.25% while the applications rose 4.68% and the opex layer rose too: an AI bid inside a growth sell — the exact mirror. Two weeks, two opposite verdicts, same one-week cadence as every other board. What was bought in both weeks: the application layer. What was sold in both: nothing — which is the sideways market in one sentence. The index carrying no AI thesis, which last week was the house expression of the axis, spent this week as its casualty: longevity −3.2% after leading, speculative −7.4%, space −7.3%, only Asia-Pacific and payments green.

Inside the indices — where the week actually happened. Inside Rubin the best line was EDA & chip IP +5.98% — the buildout’s software — with design +3.55% and the chip architects +2.34% behind it, while HBM Memory, last week’s only green sub-index, fell 4.01% into the bottom five in the same week Korea’s fund confirmed its breakout in dollars. The bottom is unchanged in character: DC construction −6.69%, storage −6.18%, fab subsystems −5.15%, testing & metrology −5.04%. Inside the Agentic Ecosystem the security stack got paid outright: identity, trust & governance +13.66% — the best sub-index in the entire four-family system — operations & observability +5.84%, govern & secure +5.74%, agentic security +3.72%; at the bottom, compute operators −4.02% and foundation models −3.40% against their +723% year. Inside Agentic Winners, everything: control plane +7.87%, enterprise +5.84%, application leaders +5.51%, and megacap gateway +3.38%, green again after two red weeks, still the only sub-index positive on the year. No rotation inside the layer. The whole layer was bought.

Best and worst week, by name — the mirror, third week running. Across the family’s constituents, the five best: Okta +23.0%, Salesforce +22.4%, freee +17.3%, Elastic +16.3%, CrowdStrike +13.8% — then ServiceNow +12.6%, Ferrotec +11.9%, Atlassian +11.7%, Veeva +11.6%, Synopsys +11.2%. Nine of the ten are software or application names; not one chip, not one cloud operator, not one HALO name — the exact inversion of last week’s list, which was nine HALO-and-applications names with no software. The five worst: Aehr −20.6%, AST SpaceMobile −15.4%, IREN −15.4%, Applied Optoelectronics −14.9%, Everpure −14.0% — then Tempus AI −11.9%, last week’s best name in the entire system at +39.5%, now sixth-worst. Third consecutive week in which the prior week’s hero list leads the casualty list. The medians say it without names: applications +2.93%, opex +1.65%, capex −1.62%, HALO −2.18%. Green counts: 34 of 40 in the applications, 45 of 126 in the buildout, 26 of 99 in HALO.

The AI Handoff Board — a third week up-stack, with a security spike. The handoff ratios: use-against-build +3.85% to 1.6007, use-against-operate +3.04%, beyond-gateways +3.19% — the applications pulling away from builders and operators for a third consecutive week. The sharpest line is new: trust-against-execution +4.48% — the security layer re-rating against the workflow layer it protects, the Okta-CrowdStrike week expressed as a ratio. And the one going the other way is the operators again: operators-against-suppliers −4.79% — the neoclouds falling against the chipmakers that supply them, which is what a capacity trade looks like when the market stops paying for capacity.

The hyperscaler cohort — the streak broke at the shelf’s door. The Mag Pulse board reads the basket at 69.07, +2.66% on the week — 0.62% below the 69.5 shelf that has capped it since spring, from 3.2% below a week ago. Underneath: Microsoft +6.27%, Meta +5.11%, Apple +3.35%, Amazon +3.02%, Oracle +2.99% (+18.3% on the month), Nvidia +1.32% through its own print, Alphabet +0.51%. The hyperscaler cohort beat QQQ by 2.24 points. Four weeks of “the defaults trail the average stock” ended in one: the basket beat equal-weight by over three points, and the question this letter has carried — rotation or distribution — received the third answer: re-concentration into the balance sheets, on a hawkish podium. The shelf at 69.5 is now the working test: a weekly close through it would say the defaults are leading again for the first time since spring; a rejection would file this week as a defensive crowding into size. Either way, the December pivot at 62.56 has stopped being the live end of the map.

Compute tightness — the operators’ third red week, against a supply-capped guide. The demand signal could not have been louder: the chip supplier itself capped next year’s growth at 70% by shipping capacity, with demand above it. The operators who rent that compute were sold anyway — Nebius −4.5%, CoreWeave −4.1%, IREN −15.4% — their third consecutive red week, and the operators-against-suppliers ratio fell 4.8%. Price and thesis now disagree in the sharpest form yet: the tightest supply statement of the cycle, from its most authoritative source, against a funding market that keeps marking the capacity traders down. One of them is early. The book’s tilt has already chosen which.

The financing architecture underneath it — and a coming deep-dive. The operators’ bleed is, at bottom, a funding question, and the funding side moved structurally this month. On August 10 Nvidia announced compute-financing platforms with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR — over $500 billion of third-party capital to be mobilized for AI factories, with the explicit aim of turning Nvidia compute and full-stack AI infrastructure into an investable asset class — long-duration, usage-linked revenue streams packaged for the capital that owns pipelines and toll roads, with Nvidia reportedly standing behind residual GPU values for up to a quarter of a project’s value, the exact risk that has kept banks away from lending against chips. BlackRock’s chief executive compared it to the birth of mortgage-backed securities in the 1970s. Hold that against this week’s tape: the market is repricing the equity of the capacity traders in the same weeks the industry is building the machinery to move that capacity onto credit — off the neoclouds’ balance sheets and into structured vehicles. Whether that machinery rescues the operators or disintermediates them is, in this letter’s view, the most consequential open question in the whole stack — and it gets its own treatment: a special Closelook analysis on the August 10 platforms — how AI infrastructure and GPUs become a financeable asset class, who carries the residual risk, and what the mortgage-securitization analogy implies for both directions — is in preparation and will publish separately on closelook.net.

Structural inflation — contained, while the long end rallied. The house composite in the macro lab reads 49, “Contained”, direction flat — unchanged. The 30-year rallied to 5.207% into the buyback window while the belly failed its third reclaim; the gauge keeps saying the long end’s problem is supply, not inflation, and this week the long end traded like a market that agrees — until Friday’s keynote reminded it that the short end has its own opinion. Warsh’s prices-first framing against a Contained structural gauge is the fall’s macro argument in miniature.

Breadth and temperature — the reversal week. Equal-weight lost 0.44% against cap-weight’s +0.47% — the first cap-weight week in five — while the Money Temperature composite warmed a second consecutive week, 57 from 53, still transition. Equities crept, hard assets fell, the gauge warmed anyway: what it is reading is rotation-as-appetite — money moving between layers at full speed rather than leaving. Participation narrowed this week by choice, not by damage: the average stock fell half a percent while the top of the index absorbed the inflow.

The count, restated for the week ahead. Above the 50-day, 4.1% from confirmation, 3.1% from the kill-switch, with five scoring windows Monday, six Tuesday, Broadcom Wednesday and payrolls Friday. The count has its referees booked daily.

4 · What May Lie Ahead

Read more

Edition of 2026-08-23 · from the C+ archive

1 · This Week’s Action

The tape, day by day. The week had five different sessions and one direction. Monday the semis reclaimed their 50-day lines at midday and the S&P slipped 0.47% anyway. Tuesday the dollar-yen wire broke overnight and the semis were rejected at the overhead line — SOXX −4.96%, QQQ −1.69% — the sharpest session of the week. Wednesday the Treasury doubled its long-end buybacks, the minutes leaned hawkish, Seoul had its sidecar night, and the index closed up 0.21%. Thursday the buyback relief reversed inside a session and Walmart was sold nine percent on a double beat: S&P −0.84%, VIX 16.01. Friday the chart pick rose thirteen percent, the CAC snapped seven red sessions, and IEF failed its weekly reclaim a second time: S&P +0.41%.

Five sessions, and the S&P lost 1.37% to 765.72, the Nasdaq 100 2.41% to 713.44 — its 50-day average to the cent — the equal-weight S&P only 0.49%. The VIX rose 6% on the week and still closed at 15.13. Everything that happened this week happened underneath the index again; this time it was mostly down.

The sector read — three green, and the three are a sentence. Health care +4.33%, energy +2.79%, materials +1.90% — the only green lines of eleven, and the same three the global sector board produced. The bottom three are the buildout pair and its neighbour: technology −3.53%, utilities −3.48%, industrials −3.36%. Chips and the power to run them fell together, while the sector that sells medicine led by a point and a half. Last week the sentence was “a capex board with a geopolitical spike inside it”. This week it is a board with the capex taken out and the non-tech growth put in.

The sector rankings — relative strength, three lenses. Our sector-RS board ranks the eleven against the S&P on three horizons and tags each a quadrant. Health care is Leading on all three reads — RS 106.5, +4.4% against the index over 21 days, +15.0% over 63, z-score +1.84. Energy is Leading (RS 126.1, z +2.05); materials Improving (+2.7% / +4.1%, z +1.35); financials second on the 63-day lens (+8.1%) though still lagging on the month. And technology is tagged Weakening: RS 116.9, the second-strongest one-year line on the board, but −1.0% over 21 days, −0.6% over 63, z −0.80 — a leader on the year that lags on every shorter horizon, which is the board’s own definition of a tired leader. Utilities is last on every lens (−11.1% over 21 days). The four sectors this letter is watching — tech, health care, materials, financials — sit at ranks eight, two, four and seven on the month, and at six, one, four and two on the quarter.

Underneath the four focus sectors — participation breadth. The sector board shows the ETF; the house dispersion pages show what its members are doing. Information technology: 47% of its 72 covered constituents above their 50-day average, 39% above their 5-day, no name at a new 52-week high — 17 five-day highs against 15 five-day lows, a sector still +2.7% on the month carried by a narrowing half. Health care: 81% above the 50-day, 85% above the 100-day, ten names at new 52-week highs — 17% of the sector — 23 five-day highs against 3 lows, and the median member +9.2% on the month, a point ahead of the ETF itself: broad, not carried. Materials: 69% above the 50-day, 84% above the 200-day, 54% of the sector at a five-day high on Friday, two at 52-week highs. Financials: 64% above the 50-day, 87% above the 200-day, 26 five-day highs against 7 lows. The strong sectors are strong underneath. The largest sector is narrow underneath — which is the week’s axis in one more instrument.

The factor read — momentum cracked, and the regime gauge moved below its trend. Last week our factor-regime gauge put momentum over low volatility at the 94.6th percentile and above its 50-day trend, and this letter wrote that a spread that far into its distribution “has never mean-reverted gently”. This week the gauge reads 92.3rd percentile, below trend: indexed 156.83 against a 50-day of 158.27, regime label momentum repair attempt below trend. The pair behind it: the S&P Momentum ETF −2.99% on the week against the Low-Volatility ETF’s −1.38%; momentum’s twenty-day rate of change +1.52% against low-vol’s −2.44%, but over sixty days momentum −0.58% against low-vol +2.35% — the defensive leg has been quietly outrunning the aggressive one for a quarter, and this week the twenty-day window caught up with the sixty.

Saturday’s Global letter found the same turn built from international instruments: momentum’s breakout failed on its weekly close, global min-vol made a new high. Two continents, two constructions, one crack — which makes it a market fact rather than an American artefact, and means the turn is not contained to one geography either.

The axis turned — from software-vs-semis to tech-vs-everything-else. For most of this year the interesting line inside the tape ran through technology: software against semis, the yin and yang of the AI trade, with one half paid and the other charged. That line is still visible — software lost 0.68% this week while semis lost between 4.7% and 8.5%, and over a month software is +12.6% against semis −5.9% — but it is no longer the line that decides the index. The line that decided this week ran around technology. The Nasdaq 100 ex-technology closed at 104.37, +0.49% on the week, at a three-year high, through the 103 shelf that had capped it since February. The Nasdaq 100 technology sleeve closed 314.19, −3.82%, 6.4% below its June high and under a descending line drawn from it. Same index, two halves, opposite charts.

The windows say how far that carries. On the year tech still owns it: the Nasdaq 100 +16.1% against its ex-tech cut’s +5.0%. On the month and the quarter the order has turned: ex-tech +6.0% and +6.0%, the Nasdaq 100 +0.6% and −0.1%; the S&P ex-tech +2.7% and +3.8% against the S&P’s +2.3% and +3.1%. Health care +9.0% on the month and +17.9% on the quarter. And the 2025 stars are back alongside the industrials: gold +13.0% on the month, the bitcoin fund +16.0%, copper miners +20.9%, materials +6.9% — every one of them closed Friday at its one-month high, while the Nasdaq 100 closed 2.5% under its. It usually travels with a weaker dollar, and it did: the dollar index fund −0.75% on the week.

Growth with no AI exposure is strong; growth with it is not, this month. The year has not changed hands. The month has. Probability, not prophecy — and the month is what September trades. One standing reminder belongs here: by the commonly cited multi-decade tallies, the best-performing stock in the S&P 500 over the long run is not a technology company at all — it is Monster Beverage, an energy-drink maker.

Underneath the factor read, participation held better than the index: the equal-weight cut lost 0.49% against cap-weight’s 1.37%, the Magnificent-7 basket closed −1.44% — a fourth consecutive week in which the defaults trail the average stock — and the best line on the family’s entire board was a health-care name. The average stock is not falling apart. The index’s largest weights are.

Inside tech — the year’s winners did the most, downward. The five green lines on the tech board are a list of what is not the AI stack: crypto equities +9.68% (with the coin), ARKK +6.30%, fintech +2.95%, lithium +1.83%, uranium +1.56%. The four that led last week all reversed — WTAI +4.69% became −4.43%, cloud +4.18% became −1.30%, data-centre REITs +3.60% became −3.44%, quantum +3.40% became −5.17% — and semis were the worst of the board: XSD −8.46%, fabless −6.15%, SMH −4.66%, all still between +45% and +55% on the year. Last week’s tell was “the names that ran hardest did the least”. This week they did the most.

2 · The State

The sorting rule, one week on: nothing was paid. Last week the market paid challengers and charged champions. This week it charged both, and paid almost nothing. The challengers it paid seven days ago gave it back and more: Nebius −21.1% on the week to 219.13, CoreWeave −16.5% to 87.85 — the two payments this letter said had held through day three did not hold through day eight. The champions it charged were charged again: Coherent −11.1% from its scoring entry, after the market had repented on day three (+7.8% on Monday, then −12.8% on Tuesday — a one-day pardon); Applied Materials −2.9% from its; Cisco’s charge held at +1.1%. And two new champions arrived and were charged on clean prints: Fabrinet, ten straight beats on the line, −19.4% on the day and −27% from its pre-print close; Walmart −9.15% on a revenue and earnings beat with a raised outlook. Baidu missed and lost 12.7%. The only print paid on the week was Target, +4.3% on the day, +7.1% on the week — the discount name.

So the rule did not break; it narrowed. What the market will pay for shrank to what arrives at a discount, and what it charges expanded to include good challengers as well as perfect champions. A tape that charges nearly everything for a week is not sorting by expectation any more. It is de-risking — and the thing being de-risked is the AI complex specifically: health care, energy and materials rose while every AI sleeve fell.

The one thing that was paid was the chart pick. Friday’s Weekly Chart Pick published Robinhood at 95.10 before the open — wave five out of a year-long triangle, the record quarter with crypto revenue down 38%, a manual flow set reversing up — and it closed at 108.13, +13.70% on day one, through the 100 shelf, on the session the bitcoin complex confirmed its breakout. One day is one day, and this letter scores at one week, one month and beyond. But in a week the market paid nothing it was supposed to pay, the one setup it paid was the one published on the page the same morning. The diary records it; it does not take a lap.

The veto, scored at the close: failed twice. IEF broke 93.17 on Monday, reclaimed it on Wednesday on the buyback — 93.38 at the close, above the line — and gave it back: 92.82 on Friday. By this book’s convention the weekly close is the verdict, and the verdict is the second failed reclaim in a row. The company the veto keeps is the real story: the Treasury doubled its long-end buybacks on Wednesday and the long end backed up through them within a session — the 30-year at 5.23% by Thursday midday, 5.25% by Friday morning, 5.275% at the final — while the Fed’s minutes leaned hawkish with three dissents for a hike. Saturday’s letter carries the sovereign board and the G7 detail; the tape’s reading is simpler: the bond market has now declined to sign the equity resolution two weeks running, and this week the equity market finally noticed.

The macro print was the right shape for the wrong reason. Walmart beat on revenue and earnings, raised its sales and operating-income outlook, guided full-year earnings under the street and printed US comparable sales of 2.6% against the 3.5% expected — and fell 9.15%, the worst earnings-day reaction in ten quarters, the fourth straight earnings-day decline, on the name this letter’s print record carried as nine flats in ten. The nine-flats era is over. The consumer read is a comp half a point under where the street had it; the tape read is that the broadest consumer name on the board has joined the list of things the market will charge for being expected to win.

The count, long term — unchanged from last week. The wave count on the Nasdaq 100 reads as it did seven days ago: 1-2 complete off the April low, wave 3 ongoing, 746.16 the June high that confirms it, 694 the level that voids it. This week moved the index inside that map — from above 723.85 to its 50-day — without touching either line. The chart is the same chart; the price is closer to the lower line.

3 · The Outlook

The four indices — the lightest layer rose, the heavy layers bled. The family printed the week’s axis in one row: Agentic Winners +2.7% (the applications; still −4.6% YTD), HALO +0.0% (broad growth, no AI thesis; +8.4%), Agentic Ecosystem −6.5% (the opex layer; +58.4%), Rubin Build-Out −7.3% (the capex layer; +90.5%) — with Euro-AI −3.5% across the Atlantic, +31.7% on its year.

Two things to hold from that row. The control group did not fall: HALO flat against the buildout −7.3% and the operating layer −6.5% says this was an AI unwind, not a growth unwind — the mirror of last week’s “an AI bid, not a growth bid” — and the index carrying no AI thesis is now the house expression of the week’s axis. And the lightest layer rose for a second week while the two heavy ones gave back more than a third of their recovery off the July lows: capex +14.7% off its 29 July low from +23.6% a week ago, opex +15.2% from +23.1%, applications +26.4% off its 23 July low from +31.2%.

Inside the indices — where the week actually happened. Inside Rubin one line was green of twenty-four: HBM Memory +3.78%, +178% on the year, the week Korea crashed and SK hynix announced the largest buyback in the country’s history. Storage gave back 4.16% of last week’s +21%. The bottom is the fab floor: wafer processing −11.25%, testing and metrology −10.66%, foundry −10.35%, fab subsystems −9.88% — and no region hid: US constituents −8.61%, Japan −6.05%, Europe −7.14%.

Inside the Agentic Ecosystem the names that made 2026 led the way down — Foundation Models −11.02% against +753% on the year, substrate −9.29%, edge −9.22%, compute operators −9.19% — with operations and observability the best line at −3.70%. Inside Agentic Winners the inversion ran the other way for a second week: Application Leaders +6.05%, Enterprise +4.11%, Control Plane +2.81% — and Megacap Gateway −2.47%, the only sub-index positive for the year at +6.9%, red again. The applications layer spent a second week buying what it sold all year and selling the one thing it owned.

Best and worst week, by name — the whole story in ten tickers. Across the family’s constituents with a full week of prints, the five best: Tempus AI +39.5%, Freeport-McMoRan +15.3%, Illumina +14.9%, Celsius +14.7%, Uranium Energy +13.8% — then e.l.f. +11.5%, Chipotle +10.1%, Duolingo +10.0%, Ivanhoe +8.4%, Vertex +8.4%. Nine of the ten are HALO names or application names; none is a chip, a tool or a cloud. The five worst: Aehr Test Systems −24.1%, Fabrinet −23.4%, Nebius −21.1%, AT&S −18.8%, Applied Optoelectronics −16.9% — then AeroVironment, Fastly −16.6%, CoreWeave −16.5%, Allegro −16.3%, Tower −16.2%.

Put the two lists side by side and the week’s axis is legible without an index number: last week’s three best names — Nebius +47.7%, Fastly +30.4%, Aehr +30.1% — are this week’s three worst among the AI indices. Inside the capex index four names of ninety closed green; inside the operating layer three of thirty-one; inside the applications twenty-two of thirty-six; inside HALO forty-eight of eighty-nine. The median AI-infrastructure name lost nearly nine percent in five sessions. The median growth name without an AI thesis gained.

The AI Handoff Board — the stack jumped two rungs. The handoff ratios did in a week what they had been doing in months: use-against-operate +11.3% to 1.2076, use-against-build +9.0% to 1.5293, beyond-gateways +5.2% to 1.2892 — the applications pulling away from both the operators and the builders at once, which is the Application-Leaders-up, Foundation-Models-down week expressed as a ratio. Two went the other way and both deserve naming: operate-against-build −2.1% — the operating layer fell faster than the buildout it runs on — and verification-against-design −4.1%. Last week the stack shifted one rung up, leaving the picks behind. This week it left the operators behind too.

The hyperscaler cohort — the leaders sat out a fourth time, and this time fell. The Mag Pulse board reads the basket at 67.28, −1.44% on the week and 3.2% below the 69.5 shelf that has capped it since spring — further from it than last week’s 1.8%. Underneath: the hyperscalers (Microsoft −2.27%, Amazon −1.53%, Alphabet −0.31%) −1.18% as a cohort, consumer AI −2.23% with Meta −6.77% and Apple +1.12% pulling opposite ways, the supplier (Nvidia) −4.64% into its own print, the torque name (Oracle) −2.69% while still +22.0% on the month. Fourth consecutive reading in which the defaults trail a tape whose average stock is holding up. The December pivot at 62.56 sits seven percent below; the question this letter asked last week — rotation or distribution — got a week of evidence for the second reading and is still open.

Compute tightness — the demand side was sold, not refuted. The two neocloud payments that held through day three did not hold through day eight: Nebius −21.1%, CoreWeave −16.5%. Fabrinet beat for a tenth straight time and was sold 19.4%. Inside the family the Data & Memory sleeve fell 3.9%, Compute Operators 9.2%. Against that, SK hynix announced a 40-trillion-won buyback and cancellation — the largest by a Korean listed company — and HBM Memory was the only green sub-index in the capex table. Physical shortage is still a slow fact; positioning was a fast one this week, and it ran the other way. Last week this letter wrote that price and flow could not disagree for long. They stopped disagreeing — price came down to flow.

Structural inflation — the gauge cooled as the bond market heated. The house composite in the macro lab reads 49, “Contained”, direction flat, the persistence core at −0.14 and the supply-shock bucket at +0.36 not spreading to breadth or expectations. Last week the gauge’s own named risk — expectations — printed hot and this letter called that “the instrument doing its job in an uncomfortable direction”. This week the instrument says the structural side is contained — while the 30-year sat at 5.275% and the Fed’s minutes counted three votes for a hike. That is the tension of the moment: the long end is not pricing inflation. It is pricing supply.

Breadth and temperature — participation held, thermometer warmed. The equal-weight cut beat cap-weight by nine tenths of a point, the average stock fell half a percent in a week the index fell one and a half, and the best names on the board were outside the AI complex entirely. The Money Temperature composite reads 53, two points warmer than a week ago, still the middle of its range and still labelled transition. Equities fell and the gauge warmed, because the hard-asset bid is risk appetite by another name and the board read it as such. Participation good, thermometer neutral-warm, and neither reading front-runs anything.

The count — retreated, not voided. Last week this letter filed the count as advanced: QQQ through the 723.85 high-water mark, 746 the confirmation, 694 the kill-switch. This week it retreated: 723.85 was lost on Tuesday and never regained, and Friday closed at 713.44 — the 50-day average to the cent. The chip index, which never challenged its 505 floor last week, closed 2.9% above it. What has not happened is the kill-switch: 694 is 2.7% below, and a count that loses its first marker while holding its last is a count on the wrong side of the map, not off it. The difference prints at 694 — or at 746, and the index is now 4.4% from the second and 2.7% from the first. Nvidia prints on Wednesday with both lines in range.

Read more

Edition of 2026-08-16 · from the C+ archive

1 · This Week’s Action

The tape, day by day. The week arrived carrying a binary and disposed of it early: CPI cool on Tuesday, PPI cool on Thursday — and the market did almost nothing with either. The S&P added 0.40% across five sessions to 776.34, the Nasdaq 100 1.11% to 731.07, and the VIX fell to 14.32.

Friday was the exception that defined it. Retail sales printed −0.6% against +0.1% expected, Michigan sentiment 51.0 against 54.5, and the index shrugged — the S&P −0.22%, the Nasdaq 100 −0.14%. Two cool inflation prints and a demand miss inside five sessions, and the index moved four tenths of a percent. Everything that happened this week happened underneath it.

The sector read — nine green, two red, and the red is a sentence. Energy took the week outright at +7.67%, and the honest reason is the barrel rather than a thesis: crude round-tripped +7.31%, taking back nearly all of the prior week’s 8.7% fall on the Iran file. Utilities +1.61% and communications +1.53% followed.

The two red lines are the sentence: materials −0.61% and discretionary −1.38% — the demand-facing ends of the board, in the week the Michigan number said demand was softening. A sector table where energy leads by seven points on a war headline while discretionary is the worst line is not a risk-on sweep. It is a capex board with a geopolitical spike inside it.

The factor read — momentum at the 95th percentile, participation underneath. Our factor-regime gauge reads momentum-leading, risk appetite building: momentum against low volatility at 2.0077, the 94.6th percentile of its own history and above its 50-day trend, with momentum’s twenty-day rate of change at +6.55% against low-vol’s −0.27%.

The pair behind that number is tradable, which is why this letter tracks it rather than an abstract factor series: the S&P Momentum ETF closed at 153.31 against the Low-Volatility ETF’s 76.36, and the ratio between them is the gauge. Over sixty days momentum has compounded +9.02% against low-vol’s +3.99%; over twenty, +6.55% against −0.27%. The spread is not drifting apart — the defensive leg has stopped moving while the aggressive one keeps going.

Two things follow. The ratio sits above its own 50-day trend (159.4 indexed against 158.5), so the regime is confirmed rather than merely elevated. And at the 94.6th percentile, roughly nineteen of every twenty readings in this pair’s history have been less extreme than today’s — which is a statement about crowding, not about direction. Momentum can lead for quarters from here. It has also never mean-reverted gently from this part of its distribution.

Saturday’s Global letter reports the same pair built from international instruments, where momentum leads defensives by 1.9 points on the week and 4.5 on the year. Two continents, two independent constructions, one regime. That makes it a market fact rather than an artefact of American index concentration — and it means a turn, when it comes, will not be contained to one geography.

Underneath the factor read, participation held: the equal-weight cut beat cap-weight roughly three to one — RSP +1.22% against SPY’s +0.40% — while the Magnificent-7 basket closed red at −1.27%. Momentum leading, the average stock outrunning the index, and the largest names sitting it out: three readings that rarely print together, and all three did.

Inside tech — the infrastructure sleeves led, the year’s winners lagged. The AI-infrastructure cut took the week: WTAI +4.69%, cloud +4.18%, data-centre REITs +3.60%, quantum +3.40%, cybersecurity +1.79%. Software managed +1.35% and remains −1.5% on the year — one good fortnight does not repair a de-rating.

And the bottom of the shelf is the tell: fabless semis −1.44%, small-cap semis −0.07% — both red inside a green group, both up more than 55% for 2026. The names that ran hardest into this week did the least in it. That is the sorting rule showing up in ETFs before we reach a single ticker.

2 · The State

The sorting rule, named. Strip the week to its load-bearing facts and one rule explains all of them: the market paid challengers and charged champions.

The challengers were paid, and the payment held. CoreWeave rose 19.3% on its print and Nebius 34.1%. Payment retention is precisely where paid prints have died all season — the verdict lands and the market takes it back inside two sessions. Not this time: both retained through day two at −1.3% and −1.6%, and by Friday Nebius had extended another 8.8% to 277.58. Three sessions on, the payment is larger than the verdict was.

And the champions were charged, on clean beats. Cisco fell 8.4% on its twenty-first consecutive double beat — the deepest single-day charge in the only twenty-for-twenty record on the docket. Coherent fell 8.0% after running 8.2% into its print. Applied Materials beat on both lines and closed Friday −5.1% at 507.18. None of the three missed. All three were charged for being expected to win.

The veto is back — scored at the close, per the contract. Last week this letter scored the bond veto dead: IEF reclaimed its two-year trendline and closed at 93.17, above it, and by this book’s convention the close is the verdict. This week the same convention reverses it. IEF broke 93.17 on Monday, reclaimed it Thursday on the cool PPI print, and then closed the week at 93.04 — below. TLT finished at 82.04, within five cents of its 52-week low.

The rule does not get to be applied only when it flatters the position. The veto was dead on a close; it is live again on a close. What makes it more serious than a level is the company it keeps: the equity tape priced two cool inflation prints, the bond tape priced something else, and volatility got cheaper while they disagreed.

Then a buyer walked in under the discount. Silver Lake is reported in talks to take Workday private. The stock rose 17.8% on Thursday, halted intraday on 3.6 times normal volume — a name that traded near 250 last September and bottomed at 110 in April — then gave back a third of the spike on Friday to 198.68. Read as market structure rather than as a deal: operators are selling the software they think agents will eat, and financial buyers are underwriting what they think survives.

And then the supplier turned out to be a shareholder. Nvidia’s second-quarter filing, disclosed Friday, carries a stake in SpaceX worth about $21 billion — 122.8 million Class A shares as of 30 June, readable at all only because SpaceX listed in June. That makes it Nvidia’s second-largest public equity position, behind roughly $30 billion of Intel.

It did not begin as a SpaceX position. It began as Nvidia’s $10 billion investment in xAI, which SpaceX absorbed. Alongside the disclosure, Musk said SpaceX will build its AI infrastructure and data centres exclusively on Nvidia hardware — Vera Rubin included — and pointed at orbit-based compute satellites running on Nvidia systems.

Read structurally, that is a supplier holding a large equity position in a customer that has just committed to buying exclusively from it, with the commitment itself among the things that marks the position. None of this is new in kind: the buildout has been part-financed by its own vendors for two years and this letter has tracked it as the capex-anchor question throughout. What is new is the scale, and that it is visible in a filing rather than inferred. A $10 billion cheque marked at $21 billion is a good outcome. It is also a reminder that some of the demand underwriting the capex layer sits on the supplier’s own balance sheet.

The macro print was the wrong shape. Retail sales −0.6% against +0.1% expected, ex-autos −0.3% against +0.2%; Michigan 51.0 against 54.5, with one-year inflation expectations rising to 4.3% and the five-year at 3.3%. Weak demand with firming expectations is the first stagflation-shaped print of this run, and it landed after the two cool inflation reads had already been banked — which is why the bond market’s answer carries more weight this week than the equity market’s silence.

3 · The Outlook

The four indices — the operating layer took it. The family printed the week’s sorting in one row: Agentic Ecosystem +7.5% (the opex layer; +69.4% YTD), Rubin Build-Out +4.9% (the capex layer; +107.4%), Agentic Winners +3.3% (the applications; still −7.0% on the year), HALO +1.0% (broad growth, no AI thesis; +7.2%) — with Euro-AI +3.5% across the Atlantic, +39.4% on its own year.

Two things worth holding from that row. The operating layer led while the buildout still owns the year by a distance — the money goes into building, but the fastest-moving claim this week was on running. And HALO, the control group, gained under one percent: this was an AI bid, not a growth bid, and anyone reading the broadening as general risk-on is reading the wrong instrument.

Inside the indices — where the week actually happened. The sub-index tables are the week’s best microscope. Inside Rubin, the top of the board is one thesis three times over: Storage +21.1% — the strongest line in the entire twenty-four-sector table and now +357% on the year — with HBM Memory +15.3% (+168%), AI Factory Systems +10.0% and Testing & Metrology +9.6%. And the bottom of the same board: EDA & Chip IP −1.3%, the weakest line, with Power Semiconductors −0.5% and Fab Subsystems −0.1% the only other red. Physical capacity paid; the design layer that sells into everyone equally was charged.

Inside the Agentic Ecosystem, every sector was green — Edge & Distribution +12.8%, Data & Memory +12.3%, Substrate +11.4%, Compute Operators +11.1% — while its own defining sleeve stood still: Foundation Models +1.9% against +858% on the year. Inside Agentic Winners, the inversion runs the other way: Application Leaders +4.9%, Enterprise +3.6%, Control Plane +3.5%, Consumer +2.7%, every one still negative on 2026 — and the only sector up on the year, Megacap Gateway at +9.7%, was the only one red this week at −1.6%.

Best and worst week, by name — the whole story in ten tickers. Across the family’s 291 constituents with a full week of prints, the five best: Nebius +47.7%, SanDisk +35.4%, Everpure +31.2%, Fastly +30.4%, Aehr Test Systems +30.1% — a paid challenger, the memory complex, a test house. The five worst: Trend Micro −16.0%, On Holding −14.2%, Coherent −14.1%, Lasertec −10.4%, DroneShield −10.1%.

Put the two lists side by side and the sorting rule is legible without a single index number: the paid challenger tops the board and the charged champion sits fourth from the bottom. Coherent beat on both lines and finished the week down fourteen percent. Nebius beat and finished up forty-eight. Same market, same five sessions, opposite verdicts — and the variable is not the quarter, it is what was expected of it.

The AI Handoff Board — the stack rotated one layer up. The handoff ratios moved exactly where the sorting rule pointed: operate-against-build +2.5% to 1.2723, operators-against-suppliers +5.9% to 1.2704, beyond-gateways +5.9% to 1.2158 — the applications pulling away from their distribution chokepoints, which is the Megacap-Gateway-only-red week expressed as a ratio.

Two ratios went the other way and both deserve naming: use-against-operate −3.8% and trust-against-execution −5.4%. Last week this board went quiet because every layer rose together. This week it is loud again — and it is not rotation between winners and losers, it is the stack shifting one rung upward, leaving behind the rung that sells the picks.

The hyperscaler cohort — the leaders sat out again. The Mag Pulse board reads the basket at 68.26, −1.27% on the week and 1.8% below the 69.5 shelf that has capped it since spring — further from that shelf than last week’s 69.14, which was half a percent under it. Underneath: the hyperscalers (Microsoft, Amazon, Alphabet) −2.58%, consumer AI (Meta, Apple) −1.46%, the supplier (Nvidia) +0.54% — barely green in the week its own customers’ capex was the market’s leading question. The torque name, Oracle, added 2.38% and is +21.2% on the month while still −37.9% on the year.

Third consecutive reading in which the default holdings underperform a broadening tape. At some point that stops being rotation and starts being distribution; the December pivot at 62.56, eight percent below, is where that question gets answered.

Compute tightness — the demand side kept printing. The week’s quietest strong signal is the one the price boards buried. The two neocloud prints did not merely get paid, they kept the payment through a macro-miss Friday — CoreWeave at 105.26 and Nebius at 277.58, the latter still extending on day three. Inside the family, the Compute Operators sleeve rose 11.1% and Data & Memory 12.3%.

Against that, storage and memory led the entire capex table on price while their largest names sit at the bottom of our flow ledger — SanDisk and Kioxia both decelerating-up: price rising with less behind it every session. Both readings can be true briefly, because physical shortage is a slow fact and positioning is a fast one. They cannot both be true for long, and next week’s prints are where the argument gets its next data.

Structural inflation — our gauge against the tape’s calm. The house composite in the macro lab is the instrument that tells us in advance which buckets will matter, and this week it earned that: the two cool prints landed as it expected, and the third datapoint — one-year inflation expectations at 4.3% — came in on the bucket the gauge has flagged as its elevated one all month. When the composite’s own named risk is the thing that prints hot, that is not a surprise. It is the instrument doing its job in an uncomfortable direction.

Breadth and temperature — participation good, thermometer neutral. Participation stayed strong on the two readings settled at the close: the equal-weight cut beat cap-weight three to one, and the momentum factor pulled away from low volatility without the largest names contributing at all. That is broad participation with a hole exactly where the defaults sit.

The Money Temperature composite reads 51 — the middle of its range, and cooler than a week carrying two cool prints, a record in the world index and a 94.6th-percentile momentum spread has any right to feel. Read together: participation is good, the tape is not stretched on its own thermometer, and neither reading front-runs anything. It is the state where rotations get funded — and where first mistakes get funded too.

The count — advanced, not confirmed. Two weeks ago this letter filed the contrarian count: the pullback complete, the market inside the third wave of the third of the third. Last week it went live when 694 converted to support. This week it advanced again, quietly: QQQ closed 731.07, through the 723.85 high-water mark that was the map’s upper marker, with Thursday’s 732.07 the week’s best close.

What has not happened is confirmation. 746 remains the line that turns the count from live to confirmed, 2.0% above Friday’s close, and the chip index never challenged its 505 floor. A count that advances a notch in a week the index moves four tenths of a percent is a count carried by internals rather than by price — which is either the healthiest way to advance or the quietest way to stall. The difference prints at 746.

Read more

Closelooknet is a reader-supported publication — to receive new posts and support the work, become a free or paid subscriber.

This week’s edition of Closelook@US Stock Markets, dated August 9, 2026.

Last week’s letter closed with three contracts: the shelf had to hold without the calendar, the flush low had to stay a low, and Monday’s perfect-record print had to survive its own regime. All three paid — and then the week went somewhere none of the contracts anticipated. Palantir was bought 29% on its beat, breaking the sold-beat regime as a class. The Nasdaq took out 694 on Monday and never looked back.

And on Friday morning the payroll report printed minus 23,000 — the first negative jobs number of the entire cycle — and the market’s response was to rally, because within the hour the bond fund that had spent four sessions breaking the trendline this publication calls the veto reclaimed it, and the discount-rate threat that has shadowed every chart since July died at the close.

In between, something happened that has not happened since this letter split the AI trade into layers in June: all three of them were paid in the same week. The applications showed tremendous life — Atlassian up 47% in five sessions, Palantir up 40%. The operating layer rose sharply — Twilio through its 52-week high, Cloudflare and Snowflake bid — with a downturn only where the threat is direct and named. And the chip complex recovered broadly, led by the equal-weight cut, with exactly two exceptions: memory-and-storage, and Tokyo’s capex corner on Friday.

Two weeks ago this letter filed a wave count that requires exactly this configuration — broad, simultaneous, discount-rate-fed — and filed it as contrarian, against the seasonals, adopted by nobody. This week produced the evidence. Next week gets to confirm it or take it away.

1 · This Week’s Action

The tape, day by day. Monday opened the week by answering last week’s biggest question in one session: QQQ closed at 700.07 — through the 694 line that outranked everything on the board — while the semiconductor index defended a three-percent morning break of its 505 floor and closed flat: a failed breakdown at the exact level the bears needed, on day one. Monday night Palantir smashed (revenue +93%, US commercial +149%, a Rule-of-40 score of 155) and Tuesday the market did what it had refused to do for three weeks of sold beats: it paid — +29.4% to 162.66, through the descending line off the November high, and the whole complex ran with it: QQQ +3.4% to 723.85, the chip index +6.8%, Micron +7.6%.

Wednesday the index went nowhere and the market changed horses underneath — the Nasdaq red, gold +4%, the gold miners +7% — while AMD beat every line and round-tripped its week after hours: the first warning that the bid was selective. Thursday was the punishment-phase encore: Datadog beat, raised, and lost 19.0% on the OpenAI in-house-observability overhang — the season’s textbook direct-threat repricing — while the equal-weight chip cut quietly broke the downtrend line it has carried since the June top.

And Friday the macro arrived: payrolls −23k against +80k expected, government jobs −53k, private hiring +30k against 78k expected, wages +0.1% on the month — and the market bought the bad news, because bad news for the labor market is death for the rate-rise scenario. The bond fund reclaimed its line within the hour (close 93.17, above the two-year trendline near 93.1 — the veto’s official death certificate, at Friday’s close). Twilio, which had printed Thursday night, was paid +24.9% through its 52-week high and held it into the close. Atlassian +35.3%. The Nasdaq closed at 723.03 — a five-percent week, parked eighty cents under Tuesday’s high-water close.

The sector read — eight green, three red, and the red is a sentence. Tech +7.2% on top, and for once the label is accurate: the software complex and the chip complex both live there and both ran. Materials +4.8% behind it — the metals bid at sector degree (the full metals map belongs to Saturday’s Global letter). Then the long middle: discretionary +3.3%, industrials +3.0%, communications +2.8%. The red column: energy −3.4% (oil’s −8.7% week wearing its sector costume), utilities −1.7%, real estate −0.2% — the two bond-proxies and the barrel, and nothing else.

Compare it with last week’s board, where utilities were the worst sector because rates were rising. This week utilities were red while rates fell — because in a week when everything offensive was bid, the defensive yield-proxies were the source of funds. Same sector, bottom of the board two weeks running, two opposite reasons. That is what a regime handoff looks like at sector degree.

The factor read — participation without exuberance. The S&P +3.5%, the equal-weight +2.4%, small caps +3.6% — everything marched, and the ordering matters more than the levels: small caps beat the cap-weighted S&P in the year’s strongest tech week, which does not happen when a rally is running on six names. And the biggest names ran anyway — NVIDIA +11.6%, Broadcom +9.9% — the traditional megacap chip complex joining the Mag re-rating rather than funding it. Our breadth engine (full readings in §3) confirms what the factor board sketches: the percentage of S&P members above their long-term trend jumped from 68.5% to 74.0% in five sessions, the median member is now up 11.2% on the year — a new high for that reading — and Friday printed eighteen fresh 52-week highs against exactly one new low. Eighteen is not euphoria (the manic markets of memory printed sixty), but it is triple last week’s four. Broad, strengthening, not yet stretched: that is the internals’ verdict, and it is the single most important input to the count in §3.

Inside tech — the three-layer bid, printed in the sleeves. For a month this table has been a rotation ledger — one sleeve paid, another charged. This week it is a participation ledger: equal-weight chips +11.7%, broad software +8.6%, cloud software +8.4%, the chip majors +7.8%, the chip index +7.6% — every sleeve up seven percent or more, with the equal-weight cut leading the complex for the first full week of the cycle.

Two structural events hide inside those numbers, both on the grid below. First: IGV at 102.69 has cleared the 96 resistance shelf this letter left it “stuck inside” seven days ago — the operating layer’s waiting room emptied. Second: CLOU at 27.28 is through its June ceiling of 26.38 — a four-year high for the cloud basket (still, for honesty’s sake, well below its 2021 mania peak).

And the exception that proves the bid is discriminating rather than indiscriminate: memory and storage. Western Digital −20.3% on the week, Seagate −5.1%, SanDisk flat with a −3.7% Friday — red inside an eleven-percent equal-weight chip week. The market is not buying “semis.” It is buying the parts of the complex whose margins were never the argument, and refusing the corner where the whole world’s position already lives. Hold that thought for the house desk below — our own index internals make it precise.

2 · The State

The confluence, named. Strip the week to its load-bearing facts and there are three, one per layer. The applications came alive — not drifted, ignited: Atlassian +47.6% in five sessions, Palantir +39.8%, Shopify +29.4%, and the buying was earnings-attached (Palantir’s +93% revenue quarter; Atlassian’s Friday +35% print reaction).

The operating layer rose sharply — Twilio +22.3% on the week and through its 52-week high, Snowflake +12.7% to a 52-week high of its own, Cloudflare +7.6% — with exactly one casualty: Datadog, −12.7%, beaten not by its quarter (it beat and raised) but by the disclosure that its biggest AI customer is building the same product in-house. The market drew the line with a scalpel: platform software was re-rated up while the one name with a named, direct competitive threat was re-rated down nineteen percent in a session.

And the chip complex recovered broadly — +7.6 to +11.7% across the sleeves, led by the equal-weight cut, except the memory-and-storage corner and, on Friday, Tokyo’s capex complex (the laser-annealing and test names sold two to fourteen percent into the weekend while American chips rallied — the one cloud on the week’s hardware sky, and the Global letter’s Asia docket owns its follow-up).

Capex bid, opex bid, apps bid, all at once: this letter has been writing “the layers trade apart” since June. This week they traded together, and that is rarer and more informative than any rotation — because it is what the start of a broad leg looks like, if it holds.

The veto is dead — scored at the close, per the contract. Wednesday’s edition of this publication put it plainly: everyone is watching the software wreck; the line that matters is breaking in bonds. Thursday the intermediate-Treasury fund closed at 92.95, below the two-year trendline it had been fighting for four sessions — the veto, provisionally in force. Friday’s payroll print reversed it inside an hour: IEF reclaimed the line and closed the week at 93.17, above it — and by this book’s own scoring convention, the close is the verdict.

The bond market looked at a negative payroll print, a falling oil price and a 3.2% wage number, and withdrew its objection to the four bullish charts it had been vetoing all week. What remains is the conditional: no rate rise this year is now the market’s working assumption, and next week’s inflation prints are the confirmation hearing. Our own structural-inflation gauge (the desk, §3) sides with the market — but it is the CPI’s courtroom, not ours.

The earnings floor — 47% with an asterisk the size of Alphabet. The reporting season that just passed its peak is the strongest in years: blended S&P 500 earnings growth near 47% year over year, against roughly 29% expected when the season began, with analysts raising estimates into the prints — the reverse of the usual drift. Both halves of the house caveat apply on every citation: the aggregate is heavily skewed by mega-cap outliers whose results carry exceptional non-operating investment gains, and ex-those-outliers the growth rate compresses back toward the high twenties — extraordinary rather than impossible.

The distinction matters now because the composition of the season changes next week: the June-quarter giants are done, and the July-quarter cohort begins — monday.com before Monday’s open, CoreWeave Tuesday, Nebius Wednesday, with the enterprise-software names behind them. The second leg reports into a tape that has already re-rated the entire software complex eight percent in a week. That is the setup — generous and demanding at once — and the name-level map of it is Sunday evening’s Hypergrowth letter.

The natural experiment returned its verdict — entry premium is the variable. Two perfect print records, twenty-four hours apart, opposite outcomes — that was last week’s setup. This week scored it. Datadog: ten-for-ten on earnings and revenue, entered its print 0.6% off its 52-week high, lost 16% then, and this week lost 12.7% more when the competitive overhang surfaced. Twilio: the second perfect ten-for-ten record in the series, entered its print 19.6% below its high — and was paid 24.9%, through the high, holding into Friday’s close. Cloudflare sits between them: a near-perfect record, entered elevated, initially sold, recovered to +7.6% on the week.

The record tells you the company executes; the entry price tells you what the market has already paid itself for that execution. Perfection at the high is fully insured against nothing; perfection at a discount is the cheapest asymmetry the print calendar offers.

Monday delivers the fourth data point, and it is almost too clean: monday.com carries the third perfect EPS record in the series and enters 66.8% below its high. The knowledge corner (§7) formalizes the rule; the record does the predicting, not us.

3 · The Outlook

New this week, and standing from now on: the house desk — our indices, boards and gauges, read in one block. Everything below is Closelook instrumentation; every number is a Monday-to-Friday week.

The four indices — all green, first time, and by a lot. The index family printed the confluence in one row: Rubin Build-Out +7.1% (the capex layer; +105% YTD), Agentic Ecosystem +8.9% (the opex layer; +57% YTD), Agentic Winners +8.0% (the application layer; still −11% YTD, and closing that gap at eight points a week), HALO +5.3% (broad growth; +9% YTD) — with Euro-AI +4.7% across the Atlantic.

Since the family launched, every strong week belonged to one layer at the expense of another; the ladder inverted twice in July alone. This is the first week all four rose five percent or more together. When the layers stop competing for the same money, either the money got bigger or the week was borrowed — that is exactly what the follow-through test in §4 exists to decide.

And the levels shout what the weekly changes only whisper — the change in focus, printed on the grid above. The Agentic Ecosystem closed Friday at 1555 — a tenth of a percent from its June record (1557), the first layer back at its high. The Agentic Winners are already through theirs: 881 against a June high of 862 — the cohort 2026 punished hardest is the first to surpass its June mark, even with its January peak still far overhead. And the build side? Rubin sits 17.5% below its June top, and every semiconductor sleeve is 13–18% under its own — SOXX −17%, equal-weight chips −16%, the chip majors −13%. Same trade, same June, three altitudes of recovery: the market has moved its high-water line from the layer that builds the machines to the layers that run and use them. That is the focus change this letter has been mapping since the layers split — no longer as relative performance, but as who is back at record prices and who is not.

Inside the indices — where the week actually happened. The sub-index tables are the week’s best microscope, because they were built before the week needed them. Inside Rubin, the top of the board is one thesis three times over: High-Speed Interconnects +21.6% — the optical-networking complex this publication named name-for-name on Thursday as the new hardware bid — with the whole Connectivity & Materials layer +16.5%, Substrates +12.2%, the chip Architects +11.2%. And the bottom of the same board: Storage −7.9% and HBM Memory −7.2%, the only meaningfully red sub-indices in the entire family — the memory-and-storage exception, isolated by construction.

Inside the Agentic Ecosystem, the ranking reads like a demand report: Models +26.2%, Runtime +12.3%, Compute +11.2% on top — the sleeves closest to raw AI consumption — and Ops +3.0% at the bottom, the observability sleeve where Datadog’s direct threat lives. The index knew where the knife fell.

Inside the Agentic Winners, the Control plane led at +15.6% — the layer the house strategy paper argued owns the choke point (The Control Point) — with Enterprise +9.5% behind it and the consumer sleeve at the rear. And inside HALO, the week’s leader was Defense +15.7% — a bid this letter did not discuss all week and its index caught anyway.

Best and worst week, by name — the whole story in ten tickers. Across the family’s roughly three hundred constituents, the week’s five best: Coherent +44.2%, Applied Optoelectronics +43.8%, Fabrinet +29.2%, Aehr +28.9%, Entegris +27.7% — four optical-interconnect names and a test house, the Rubin S14 complex sweeping the podium. The week’s five worst: Western Digital −20.3%, SK Hynix −17.2%, Samsung −12.0%, Hanmi Semiconductor −10.4%, AppLovin −12.4% — the memory-and-storage corner plus Korea’s flagship large-caps plus the punishment phase’s remaining defendant. Our own index internals thus reproduce, name for name, the global signature Saturday’s letter mapped across countries: the crowded flagships of the last leg sold, everything adjacent to fresh demand bought.

The AI Handoff Board — quiet, and the quiet is the message. After July’s violent re-ratings, the handoff ratios barely moved: use-against-build at 1.41, +0.8% on the week; use-against-operate actually ticked down 0.8%. A handoff gauge measures relative motion between layers — it goes silent precisely when all layers rise together. One ratio did move: beyond-gateways +6.8%, the applications pulling away from their distribution chokepoints — the AW40’s control-plane week expressed as a ratio. The board’s July story was rotation; its August story, one week old, is expansion.

The hyperscaler cohort — the leaders sat out the broadening. The Mag Pulse board’s cleanest reading yet: the hyperscaler cohort (Microsoft, Amazon, Alphabet) gained just +1.9% in the week the market gained 3.5 — trailing the Nasdaq by more than three points — after two weeks in which the same three names were the market. The mega-cap basket overall rose 4.8% to 69.14 and now sits half a percent below the 69.5 shelf that has capped it since the spring; the December pivot at 62.56 is nine points below.

Two readings coexist: the discount-closing thesis paused while the rest of the market caught up (healthy — this is what broadening means from the top down), and the basket faces its trigger level with the tape already warm. A weekly close through 69.5 would put the market’s most-owned names back in new-high territory for the first time in twenty months — the Mag Pulse scores it daily.

Compute tightness — the demand side kept printing. The week’s quietest strong signal: CoreWeave +26.3% into its Tuesday print — buyers front-running a neocloud quarter — while the Agentic Ecosystem’s Compute sleeve rose 11.2% and SpaceX-adjacent compute names ran with it. Add the season’s supply-side testimony (SanDisk guiding bits on allocation beyond 2027; flat-at-record margins across the memory guides) and the reading is unchanged but louder: compute remains the binding constraint, the market pays whoever owns capacity, and the price weakness in memory stocks is a positioning phenomenon arguing with a physical shortage. Both can be true for a while. They cannot be true forever.

Structural inflation — our gauge sides with the bond market. The house inflation composite (the macro lab) reads 49 — “Contained” — with the labor-tightness bucket the loosest on the board (43) and shelter right beside it: the payroll print did not surprise this gauge. The two watch-items it flags: long-run inflation expectations at 62, the one elevated bucket, and supply/energy momentum at 63 — a shock reading its own methodology currently files as “not yet spreading; treat as noise.” Translation for next week: our instrument expects the CPI to cooperate, and tells us exactly which two buckets to blame if it does not.

Breadth and temperature — strong, warming, not stretched. The breadth engine: 74.0% of S&P members above the 200-day (68.5% a week ago), 66.9% above the 50-day, the median member +11.2% on the year — a new high-water for that median — and 18 new 52-week highs against one low. The Money Temperature board warmed from neutral to the low sixties, with the S&P instrument at 87 and the Nasdaq at 79 — individually hot. Read together: participation is the best of the cycle and the tape is no longer cheap on its own thermometer. That combination front-runs nothing; it just tells you the margin for a hot CPI is thinner than the charts feel.

The count — from filed to live. Two weeks ago this letter filed a contrarian count: the pullback complete, the market entering the third wave of the third wave of the third — the segment where everything rises at once and the sold corners get bought hardest — filed, explicitly, not adopted, with its tells listed.

Audit the tells against this week: QQQ through 694 on Monday and never back (tell one). The sold corner — the equal-weight chip cut — leading the whole complex at +11.7% (tell two: the strongest bid where the selling was). All four house indices up together (tell three: no more rotation-funding). Breadth expanding into the move, 74% and climbing (tell four). Even the world index agrees — Saturday’s letter carries the same count on VEU, one cent from its record close, and the total-world index at a new all-time high.

The count is now live — upgraded from filed, still not proven. What it requires next is written in §4, with the level that would kill it. And it proposes to do all this through August and September, the calendar’s two worst months — which is exactly why the follow-through rule does the deciding, not the narrative. Probability, not prophecy.

🔒 Be Informed — the rest is for subscribers. Free: the LinkedIn newsletter + closelook.net. C+ (paid): real-time push on every portfolio move → closelook.net/subscribe/

Read more

Closelooknet is a reader-supported publication — to receive new posts and support the work, become a free or paid subscriber.

Read more

Closelooknet is a reader-supported publication — to receive new posts and support the work, become a free or paid subscriber.

This week’s edition of Closelook@US Stock Markets, dated July 26, 2026.

Last week ended with a question — was Friday’s round trip a spring or the last reclaim? — and this week answered it twice. First the bulls’ way: the semiconductor index rallied six percent off the washout, reclaimed its June zone, and on Tuesday closed within two dollars of the broken 554 line. Then the bears’ way: a rate repricing that took September hike odds from one-in-five to four-in-five, an oil print through $100, and a Friday that closed the chip complex below its capitulation band — on the day Intel reported its fastest revenue growth in fifteen years and was sold eight percent for it. In between, six of the biggest prints of the season landed, and the market applied one consistent rule to all of them: it did not score the earnings. It scored the layer of the stack the company lives in. That rule — who gets sold on a beat, who gets paid on one — is the whole letter.

1 · This Week’s Action

The tape, day by day. Monday drifted; the week was always going to start Wednesday night. Tuesday gave the bulls their evidence: SOXX closed at 552.69, a full reclaim of the June zone that stopped almost exactly at the old 554 floor — tested from below, not regained. Wednesday was verdict night: Alphabet held the beat, raised its capex path again and showed Cloud accelerating; ServiceNow beat on essentially every line; IBM missed with a cut, its mainframe segment down 42% against distributed compute up 37% — the product cycle turning inside one income statement; Tesla missed. Thursday the macro arrived: spot crude traded through $100, September-hike odds spiked toward 80%, the tape fell 2% — Tesla −13.7%, Alphabet −6.5% through its 200-day on a raised guide, gold failing its haven test — while SAP printed positively and rose 6% off its 52-week low and Intel rallied 10% into its own report. Friday sorted the survivors: Intel’s double-beat was sold 7.9%, SOXX fell 4.4% to 527.01 — below the 530–532 band, QQQ broke 694 on its third test — and on the same tape SAP rose another 9.3%, ServiceNow 7.4%, and the S&P closed green with ten of eleven sectors up.

The sector read. Energy led for the third straight week (XLE +3.6%, +33% YTD) with utilities (+2.1%), real estate (+1.4%), materials (+1.2%) and industrials (+1.1%) behind it — the defensive-and-real shelf, again. The bottom two sectors are really two stocks: consumer discretionary −5.9% is Tesla’s −19% week wearing a sector costume, and communications −4.1% is Alphabet’s. Tech itself closed just −0.7% — because inside it, one complex cracked and another rallied, and they nearly cancelled.

The factor read — the average stock skipped this week. The equal-weight S&P closed at −0.1% against the cap-weighted −1.1% and the Nasdaq’s −2.6%; value gained 1.8% while growth lost 2.7%; low-vol and min-vol were green. For the second stress episode running, this is a concentration event, not a market event — the damage lives where the index weight lives.

Inside tech. The weekly numbers hide a violent round trip: SOXX’s −1.3% week contains a six-percent reclaim and a six-percent rejection. The month-scale ledger is the honest one — SMH −11.9% over a month against cloud software +9.2% and the broad software basket +3.8%, a twenty-point spread inside “tech.” And July’s semiconductor breadth is three green names out of twenty-two: NVIDIA, AMD, ASML — everything else red, Intel −30% for the month with the season’s best print. The complex is being repriced from its speculative edge inward: on Friday the chip-designer sleeve (XSD) fell hardest, the broad complex least.

2 · The State

The rule the reaction function is running. Line up the week’s six verdicts and the pattern is not subtle. Beats sold: Alphabet — held guide, raised capex, Cloud accelerating — closed the week −8.9%, below its 200-day. Intel — earnings roughly double the estimate, revenue +25%, its best growth since 2011 — jumped double digits after hours and gave back the entire move by Friday’s open, closing −7.9%; the reversal tracked the one number in the release that wasn’t about the quarter: Intel raising its own capex plans. Its record now reads three beats, zero relief rallies, eight declines. Beats paid: SAP +9.3% on Friday alone, ServiceNow +7.4% — the two names that used their calls to claim the same seat: orchestrator of the agentic enterprise. Misses executed: Tesla −18.9% on the week, unrescued; IBM sold on the cut. The market is not grading earnings anymore. It is grading position in the emerging stack — and this week it told us which transition it is pricing. We published the orchestration half this week — the Pulse (Agentic AI: The Orchestration Knock Out) and the long read (The Agent Is Not the Product. The Orchestrator Is.). The other half has a name too.

The handoff, named: capex → opex. Every infrastructure cycle runs the same sequence: first the market rewards building the capacity, then operating it, then using it. The switch between the first two does not happen on a date — it happens when the constraint moves from building AI to operating and monetizing it, and the market’s tell is precise: capex announcements stop producing positive reactions. That tell printed twice this week, on both sides of the same trade, within 24 hours. Wednesday night the demand side: Alphabet’s quarterly capex ($44.9B) exceeded its operating cash flow — quarterly free cash flow printed negative for the first time on record — while the full-year guide rose toward a ~$200B midpoint and disclosed purchase obligations swelled dramatically (a number to handle with care: those commitments span far more than chip orders, so they support the build’s visibility without being a semiconductor backlog). A marginal-return warning, not a financing crisis — trailing-year cash flow remains solidly positive — but the question changed from how much are you building to what does it return. Thursday night the supply side answered with the same sin: Intel, spending well ahead of its own cash generation, raised capex again — and the market sold the season’s best print for it. When both sides of a trade are spending beyond their cash flow, the sustainability of the orders in between becomes the risk. That is what Friday repriced.

The repricing, read at the sector degree. Sort Friday’s tape by dependence on future hyperscaler orders and the market’s new rule is legible in the sleeves before any single name: the chip-designer cut fell hardest, the broad complex less, the diversified complex least — the same speculative-end-first hierarchy the whole July de-rate has followed. The equipment sleeve repriced before the chip sleeve, as it always does — capex changes reach order books before they reach shipments — and behind it the physical layer of the build (the power, cooling and grid-connection complex) took three-to-five-percent hits as a group: fewer data centers under construction means fewer of everything those buildings contain. Meanwhile the software shelf next door was green. One number quantifies the whole boundary: the spread between the complex’s most order-dependent name and its most usage-priced one was seven points in a single session — the market drawing the line between what is priced on orders and what is priced on workloads. The full name-by-name ladder is Sunday’s Hypergrowth material; what belongs here is the architecture it reveals.

The four layers, named — the thesis this letter now runs on. This week forced the AI trade apart into four layers, each with its own index expression and its own verdict:

1. AI Capex — the build trade (the semiconductor complex, the broad build-out index): repriced. Priced on order flow whose sustainability is now the open question; the band crack is this layer’s chart.

2. The Rubin constraints — scarcity inside the build (test & verification, metrology, consumables, advanced memory and packaging — the layers that get more scarce, not less, when designs and deployments multiply): held. The Handoff Board’s inside-Rubin ratios — verification-versus-design and consumables-versus-tools — were green through the crack week. When the broad capex trade de-rates, the constraint layers are where the build’s value concentrates.

3. AI Opex — the operating layer (software’s consumption side, security, observability, orchestration; our Agentic Infrastructure index, +38% YTD): paid. The orchestrator Friday was this layer’s coming-out session.

4. AI Beneficiaries — the using layer (the Agentic Winners cohort): selective, print by print. No basket treatment yet — this layer must show attributable economics name by name, and mostly hasn’t.

The market stopped trading “AI” this week. It started trading the layers — and every index needed to score that transition is already on the site.

The house instrument already tracks this — and it agrees. We built the AI Handoff Board precisely for this transition: nine ratios measuring value migration across the three-index chain — Rubin builds, Agentic Infrastructure operates, Agentic Winners use — each normalized to 1.00 at June 30. The readings as of Friday: the master ratio (operate ÷ build) stands at 1.12 — the operating layer has re-rated 12% against the building layer in four weeks; the scoreboard (use ÷ build) stands at 1.24. This week both ratios dipped — Rubin stabilizing while the operators gave back — which is exactly what a broadening looks like as opposed to a clean replacement. The board is the standing referee for everything this section claims, updated daily, in public.

The macro engine underneath. None of this happened in a vacuum: Thursday’s flush repriced the September meeting from roughly 20% hike odds to above 80%, real yields pressed toward 3%, and — the week’s signature — Treasuries fell with stocks (IEF −0.8%, TLT −1.3%). A tape where bonds can’t catch the safety bid is a tape repricing its discount rate, and the most extended complex on the board absorbed the hit. Oil is the input behind the odds — spot through $100 this week — but the full term-structure read (and why the December contract near $80 says temporary) is Saturday’s Global letter; for the US tape the transmission is one line: oil → inflation odds → the discount rate → the concentrates.

The levels, and what broke. Tuesday’s 552.69 close is now the defining rejection — the old 554 floor confirmed as ceiling. Friday’s 527.01 close sits below the 530–532 June capitulation band (the first close-basis crack in that band since the spring advance) but above the 522.24 June washout line. One degree up, the escalation line we flagged last week — QQQ closing below the 686.37/686.76 double-washout floor — fired: Friday’s third test of 694 resolved as a break, and the 684.23 close sits two points beneath the floor. And yet the degree above that refused to confirm: the S&P closed green on Friday, the equal-weight was flat on the week, and ten of eleven sectors rose. Last week’s containment question has a split answer — the Nasdaq degree cracked, the market degree did not. Both facts are in ink; neither cancels the other.

The quiet strength, read correctly. NVIDIA finished the week green (+0.5%, above both moving averages) and was the only major AI-hardware name to sit Friday out — not an accident, per the boundary above: it is the one name in the complex priced on workload execution rather than on the next order. A sector-degree break that its usage-priced leader declines to confirm is not yet a bear market in the AI trade — it is the market splitting the AI trade along the order/usage line. Whether the leader eventually pulls the complex up or the complex pulls the leader down remains the cleanest single tell in the US book.

3 · The Outlook

The four indices — the reaction rule, visible in our own instruments. The sold-the-beat/paid-the-plane rule from §2 printed straight through the Closelook index family. On the week: Rubin +0.4% (+92% YTD) — the build-out stabilizing after its washout, even as its ETF proxies cracked. AW40 −3.7% (−23% YTD) — but +2.2% on Friday alone, the orchestrator day, the index’s Control-Plane cohort doing exactly what SAP and ServiceNow’s tape did. The Agentic Ecosystem −2.9% (+38%), HALO −1.9% (−1%). One tape, four different verdicts — the AI trade is no longer one trade, and the full four-bucket flow system behind these numbers is the Hypergrowth letter’s home turf (also out this Sunday). Here the Agentic Winners index earns one sentence: it is where the orchestration thesis gets scored print by print, in an index, in public.

The regime gauge. Money Temperature at 50, up from 44 — back to dead-neutral. The forced-selling phase cooled; nothing euphoric replaced it. At 50 the market funds rotations, not chases — which is precisely what the sector board shows.

What would confirm the handoff — the diary’s checklist. One violent Friday names a transition; it does not confirm one. We will call the capex→opex handoff confirmed when, and only when, several things print together: capex raises keep getting sold across the hyperscalers (Wednesday and Thursday are live tests); analysts stop raising out-year estimates for equipment and memory while orders are still strong; consumption-linked revenue — inference, observability, security, orchestration — visibly accelerates off the installed base; and the opex names show attributable AI revenue, not “AI is a tailwind.” By that standard, July 23–24 was the opening shot, not the verdict: confirmation needs roughly two earnings cycles of relative estimate migration, which points at the second half of 2026 as the transition window. And the both-hands clause, stated plainly: this is a broadening, not a replacement — inference growth itself demands compute, so the usage-priced end of the hardware complex can stay a winner while the order-priced end de-rates. The vulnerable cohort is narrow and specific: marginal, long-duration capacity whose valuation requires hyperscaler capex to accelerate forever.

The week ahead is the season’s center of gravity. Cadence prints Monday — one session after Kimi K3’s promised weights release, making it the one name carrying both the design-shock verification and the sold-beat regime at once. Then the block that funds the entire AI trade: Microsoft and Meta on Wednesday, Amazon and Apple on Thursday — the four largest capex budgets reporting within 24 hours, into a cracked semiconductor band, an 80% hike probability, and month-end flows. Our Microsoft print-record card carries the season’s sharpest stat into Wednesday: twenty beats in twenty quarters — and a stock that has been sold anyway in the recent ones. If the sold-beat rule survives that print block, it is a regime. If it breaks, Friday’s crack loses its sponsor.

Share Closelooknet

4 · What May Lie Ahead

The levels, updated. SOXX: the map now reads 530–532 as overhead (the band, lost on Friday’s close), then 522.24 (the June washout line — Friday closed 527.01, between the two), then the primary rising channel floor near 520 — the line that decides whether the down-channel from the June highs stays corrective — and below it Friday-of-last-week’s 498.54 print. The structural read is unchanged from our chart work: the decline from June still counts as a flag-shaped correction while the channel floor holds. QQQ: 694 and the 686.37/686.76 double-washout floor are both overhead now — Friday closed 684.23, the first close beneath the floor; a fast reclaim of 686–694 would mark Friday as a false break (SPY’s green close argues that branch is live), while acceptance below it points the map toward the June congestion zone. SPY: green Friday, ~745 (the 50-day) overhead, structure intact. Alphabet: closed 319.74 against the 200-day at ~324 — the raise-and-sold print leaves it below the line institutions defend; a reclaim would be the first evidence the collector-selling is exhausting. Intel: the 3-0-8 record needs no level — it needs a reaction regime change.

The calendar just testified — against itself. July is, by a wide margin, QQQ’s most reliable month: +3.45% average, positive 88% of the time — 15 of the last 17 years. July 2026 printed −7.08% — not merely the first red July since 2024, but the worst July in the entire 17-year sample, more than four points beyond the previous worst. When the strongest seasonal tailwind of the year produces the worst print in its own history, the message is not seasonal noise — it is that something in the tape currently overpowers the calendar’s most dependable pattern. And the calendar now stops helping: August is mediocre on QQQ’s history (+0.65% average, 56% positive) and September is the only negative-average month of the year (−0.31%, positive less than half the time). The next reliably green columns are October (+2.50%) and November (+2.92%). The band cracked at precisely the moment the seasonal floor was scheduled to leave the building — which cuts both ways: the bears lose their “it’s just seasonality” objection for July, and the bulls lose the calendar as a crutch until October.

The three bellwethers. The architect diverged from its complex: NVIDIA +0.5% on the week, green, above both averages — the only major AI-hardware name that ignored Friday. Micron +4.0% — the first green week after a −24% month, the memory unwind pausing at scale. TSMC −1.1%, quiet, still below its 50-day (~426) but far above its July low: the paradox holder is consolidating, not breaking. A band crack with all three bellwethers flat-to-green is a very different animal from June 9, when they led the washout. Watch whether that divergence resolves upward (leaders pull the complex) or downward (the complex pulls the leaders) — it is the cleanest single tell in the US book right now.

Read more

On Friday the market ran the entire distance to the June capitulation zone, touched it to the dollar, reclaimed everything by lunch — and gave most of it back by the close. This is what a verdict week looks like when both sides get their day.

Read more

Last week we drew two lines and said the market would have to choose: SOXX at 554 and NVIDIA at 191. This week it chose — twice. Tuesday broke the first line and looked like the start of a correction. By Thursday the break had reversed, extended, and been paid for with the largest foreign US listing on record. And underneath the round trip sits the real story: where the buying went. Not to the architect — down the stack, to the layers that had not yet repriced. Which happens to be exactly where the AI dollar itself is headed. That is this week’s letter.

1 · This Week’s Action Free

The tape, day by day. Monday was setup. Tuesday was the software day — SOXX broke 554 and our build-out index had its worst session of the series (−5.6%). Wednesday reversed it: SOXX reclaimed the line at 562, NVIDIA rose 3.7% back above 200, Broadcom added 4.8% on the ~$30B Apple deal. Thursday extended: SOXX +3.5% to 581.7 — the second close above 554 — Micron +4.5% to a $1.12 trillion market cap, software green on a semi day for the first all-green session of the four indices since Monday. Friday resolved the week in character: the memory complex took profits into the SK Hynix debut, while the architect had its comeback — NVIDIA +4.0% to 210.96, back above its 50-day, and SOXX closed at 581.3, flat on the day and twenty-seven points above the line it broke on Tuesday. The week’s net: SOXX −0.3%. The round trip was violent; the destination was where it started. That is what a shakeout looks like on a weekly print.

The sector read. Energy led the week (+3.8%) as oil caught a real bid — the “oil isn’t buying the war” read of recent weeks weakened this week. Tech (+1.7%) and communications (+1.4%) carried the rest; every other sector closed red, with materials (−2.0%) and industrials (−1.5%) at the bottom. Narrow, commodity-accented risk-on.

The macro frame. The FOMC minutes put rate hikes back on the distribution as a contingency — the hawkish tail is alive, the 10-year sits near 4.5%, and the long end sold off again into the weekend. Gold eased but held the 4,000 belief line; bitcoin continued its repair (+4% on the week). The rates question is not resolved; it simply was not this week’s headline. Tuesday it becomes the headline.

2 · The State Free

The count, updated. The five-year view puts the week in its structural place. The advance from the late-2022 low counts as a completed five — that was Wave 1, topping in early 2025. The spring-2025 flush was Wave 2. Everything since is Wave 3, ongoing: its own first sub-wave completed into the winter, its second bottomed at the spring-2026 low, and price now rides the upper half of the red channel in the third-of-third — the strongest part of the sequence, if the count is right. Tuesday’s break never seriously threatened the channel midline, let alone the lower rail; the weekly stochastic has reset to mid-range rather than overbought. The bearish alternative — a completed five at the high — needed the break to stick, and it lasted one session. The channel rails are the count’s own invalidation levels: the midline is where the bulls get their first real question, the lower rail is where the count dies.

The semi complex — the group that decided the week. The four semiconductor wrappers tell the shakeout story in one frame: Tuesday’s break, Wednesday’s reclaim, Thursday’s extension — and a Friday that held the ground. The fabless sleeve (SMHX, +3.2% on the week) led the complex out while the broad wrappers ended flat: break → reclaim → extend is the signature of forced selling absorbed by size, not the start of distribution.

Inside the buying — down the stack. The detail that separates this reversal from a simple bounce: on the strongest day of the week the architect sat out. NVIDIA closed red on Thursday while the layers beneath it re-rated — Storage +6.9%, Deposition & Etch +6.1%, Advanced Packaging +5.7% inside our build-out index, and Micron closed the day worth $1.12 trillion, roughly ten times its 52-week low, still near 6.6x forward earnings. Capital did not chase the top of the stack; it bought the layers that had not yet repriced. That broadens the trade rather than narrowing it.

Where the dollar itself is going. The down-the-stack buying is not a one-week quirk — it mirrors the migration of the AI dollar itself. Last week’s Global edition traced the geography: of every $100 the hyperscalers spend, roughly $45–55 stays in the US — but what stays is the design and IP margin, the cloud economics and the concrete, while the physical system bill runs through the supply chain.

The forward path matters more than the snapshot. Generation by generation — Blackwell today, the Vera Rubin platform entering the market this half, Rubin Ultra scaling toward 576-GPU domains, Feynman behind it — our estimate of the rack dollar shifts the same way the week’s tape did: compute’s share falls from roughly 43% toward ~31%, total memory rises from ~29% toward ~39%, and networking stops being an accessory and becomes part of the processor. The buyers who re-rated storage, etch and packaging on Thursday were, knowingly or not, front-running that table.

3 · The Outlook Free

The four indices — the week in one line each. Rubin Build-Out: −5.6% Tuesday, +3.9% Thursday, −0.1% on the week — a round trip that proved the holder base. The Agentic Ecosystem Index led into Thursday and gave back 3.3% on Friday into the debut, closing the week +1.4%; AW40 added +1.7%; HALO lost −2.6%. Year-to-date the ladder reads Rubin +119%, AEI +47%, HALO +5%, AW40 −21% (equal-weight variants). The barbell swapped sides twice and ended the week near balance — a tug-of-war, not a regime change. Friday’s split inside the pair — the architect +4% while the agentic complex gave back — is the same rotation §2 describes, running in real time.

The road ahead — our working map, and the number that will decide it. Here is the path we are working with, stated plainly enough to be wrong. The tape keeps rising into earnings season — mid-season, perhaps to its end, at the outside until the July-quarter reports land in August. Then the calendar turns hostile: the seasonally weakest stretch of the year meets midterm-election uncertainty, and by then the peak in earnings expectations may already be behind us. Consensus has S&P 500 earnings growing roughly 23–24% this year, and with normal beats the second quarter alone could print near 29% — growth rates the index normally only achieves coming out of a recession. If that is the expectations peak, the autumn setup is a correction — and after the election, the year-end rally.

But the 24% is not what it appears to be. The aggregate is profit-dollar-weighted: a small cohort of AI, energy and capital-markets names growing 25–40% pulls the headline while the median S&P 500 company grows perhaps 7–10% — a respectable expansion, not a boom. The index’s earnings line is growing roughly three times faster than the representative constituent’s. And part of the aggregate is not operating earnings at all: when Amazon books gains on its Anthropic stake — Alphabet likely the same — valuation marks become profits, and inside the AI build-out itself one side of every dollar records 100% as revenue today while the other side spreads its cost across years of depreciation. Earnings booms built on that arithmetic can run for a long time — and they can also unwind through the income statement rather than the multiple.

We will track this through the season with one number: the Earnings Dispersion Ratio — aggregate index earnings growth divided by median constituent growth. Around 3.0 today, by our working estimate. What that ratio does from Tuesday’s bank prints onward tells us which regime this actually is: an earnings boom — or a concentration wearing one as a costume. The paid half of this letter maps the levels, the bellwethers and the book; the EDR gets its own treatment in the coming editions.

The road ahead — our working map, and the number that will decide it. Here is the path we are working with, stated plainly enough to be wrong. The tape keeps rising into earnings season — mid-season, perhaps to its end, at the outside until the July-quarter reports land in August. Then the calendar turns hostile: the seasonally weakest stretch of the year meets midterm-election uncertainty, and by then the peak in earnings expectations may already be behind us. Consensus has S&P 500 earnings growing roughly 23–24% this year, and with normal beats the second quarter alone could print near 29% — growth rates the index normally only achieves coming out of a recession. If that is the expectations peak, the autumn setup is a correction — and after the election, the year-end rally.

But the 24% is not what it appears to be. The aggregate is profit-dollar-weighted: a small cohort of AI, energy and capital-markets names growing 25–40% pulls the headline while the median S&P 500 company grows perhaps 7–10% — a respectable expansion, not a boom. The index’s earnings line is growing roughly three times faster than the representative constituent’s. And part of the aggregate is not operating earnings at all: when Amazon books gains on its Anthropic stake — Alphabet likely the same — valuation marks become profits, and inside the AI build-out itself one side of every dollar records 100% as revenue today while the other side spreads its cost across years of depreciation. Earnings booms built on that arithmetic can run for a long time — and they can also unwind through the income statement rather than the multiple.

We will track this through the season with one number: the Earnings Dispersion Ratio — aggregate index earnings growth divided by median constituent growth. Around 3.0 today, by our working estimate. What that ratio does from Tuesday’s bank prints onward tells us which regime this actually is: an earnings boom — or a concentration wearing one as a costume. The paid half of this letter maps the levels, the bellwethers and the book; the EDR gets its own treatment in the coming editions.

Read more

Almost everything rose — but the two ends of the barbell we live on went opposite ways, and the divide ran straight through technology. The read below holds that this is rotation, not risk-off, with a rising earnings bar behind it. The schedule of our other publications is at the end.

1 · This Week’s Action

Five boards, top to bottom: the macro backdrop first, then the US sectors, technology, the regions, and the world. Each is sorted by Weighted Alpha, leaders at the top.

The cross-asset backdrop. The bellwethers set the tone before equities open. This week, long bonds softened as yields firmed, the dollar drifted lower, gold held its ground, and Bitcoin pushed higher through the long weekend — a backdrop that leans risk-friendly, with the one caveat that rising yields sit under the rate-sensitive corners of the equity market.

Sorted by Weighted Alpha (leaders → laggards) · Barchart, close of week.

The tell here is what is absent: no haven bid, no credit stress, no scramble for protection. Volatility fell on the week even as one corner of the market was repriced hard — the fingerprint of repositioning, not de-risking.

The eleven US sectors. The broad market had a good week — the S&P and its equal-weight version rose together, so participation was genuinely wide. Underneath, financials and communication services did the leading; the defensives and the rate-proxies — utilities, energy, real estate — sat at the bottom. Technology finished near flat, which sounds unremarkable until you look inside it.

Sorted by Weighted Alpha (leaders → laggards) · Barchart, close of week.

Read top-to-bottom, this was a mixed board that leaned decidedly green: most of the eleven sectors finished higher, a textbook risk-on rotation with banks and communications on top, and only the bond proxies — utilities, energy, real estate — plus a near-flat technology in the red. The anomaly is not that tech led or lagged; it is that the sector’s flat headline hides the widest internal split we have seen since the spring.

Inside technology. Here is the week’s real story. The tech shelf broke cleanly in half: software and cloud ran hard, while the entire semiconductor complex fell — and the heavier the equipment and memory tilt, the deeper the damage. A double-digit gap opened between the software baskets and the semis in five sessions.

Sorted by Weighted Alpha (leaders → laggards) · Barchart, close of week.

The split inside tech — software firm, silicon soft — is the cleanest tell on the board, and the one we lean on most below. Note the path, not just the close: the semis were up sharply at midweek before giving it all back and more into the holiday, and the memory and equipment names were the epicenter.

Across the regions. Regionally, the US kept a relative bid while Europe stayed firm, and the emerging complex was mixed. Nothing here argues for a leadership change away from US large-cap yet, though the softer dollar quietly helps the non-US books.

Sorted by Weighted Alpha (leaders → laggards) · Barchart, close of week.

The global sector picture. Globally, the same signature repeats — growth-and-cyclicals lead, defensives lag — which tells us the week’s move is a factor story, not a US-only one.

Sorted by Weighted Alpha (leaders → laggards) · Barchart, close of week.

Whether the software leg is durable broadening or a violent one-week rotation against a still-unresolved semi selloff is the next section’s question.

2 · The State

The US market map. Read through the relative-rotation lens, the eleven sectors cluster the way the boards imply: financials and communications sit in the Leading quadrant, the defensives are Lagging, and technology sits astride the divide — because the sector is no longer one thing.

The one-line reduction: broad participation on the surface, a deep fault line underneath.

The thesis. The tape is risk-on and broadening at the index level, but the broadening is uneven — and the single most important relationship on our screen, chips-versus-software, just inverted hard. For a year the AI trade meant buy the build-out; this week the market paid for what uses intelligence and sold what builds it.

• How wide — the software-over-semis spread is the widest in months, and it opened in a single week. Spreads this wide, this fast, tend to be either the start of a genuine leadership change or a repositioning that snaps back — rarely nothing.

• Where the damage concentrated — not “big tech,” and not broad semis evenly. It was the most capex-sensitive nodes: memory, storage, and the equipment names. The application and software layer was not just spared — it led.

• The earnings backing — the reason the market can rotate rather than de-risk is the profit picture behind it. Consensus still carries an unusually high 2026 earnings bar, and it is concentrated: tech and infrastructure carry the heaviest projected growth, the low-expectation half of the market much less. This week the low-expectation half rallied while the highest-expectation corner repriced — a rehearsal of what a bar this high does to a small miss.

• Seasonality — July is historically one of the Nasdaq’s better months, which argues for continuation; but 2026 is a midterm year, and the run into a midterm has historically been the weakest stretch of the four-year cycle. The tilt argues both ways — we hold both.

Leadership across the themes. The more useful cut this week is thematic, and it maps directly to our own indices — the deployment side of the trade and the build-out side went opposite directions by a wide margin in five sessions.

• Leading — the agentic and deployment cohort: applications, security, identity, the names that run on AI. This is the AW40 and HALO side of the barbell.

• Lagging — the build-out core: memory, storage, foundry and fab subsystems — the Rubin side, where the capex sensitivity is highest.

But the damage inside the build-out was selective, not uniform — and the internal tell cuts the other way. Inside Rubin the periphery held while the crowded core cracked: the previously-lagging peripheral sectors (materials, vision, grid) actually led the index on the week, while the consensus core (memory, storage, foundry) took the hit. That relative-laggard outperformance is a mean-reversion signal worth flagging — the money leaving the crowded core did not leave the thesis, it rotated within it.

That chips-over-software split — deployment over build-out — is the single relationship we would watch break first if the regime is turning, in either direction.

The structural read — two counts, one fork. The Nasdaq-100’s weekly Elliott picture is genuinely two-handed this week, and the distance between the two readings is unusually wide — so we hold both.

• The bullish count — an extended third wave, no end in sight. On this reading the advance off the 2025 low is a large, extending third wave that is nowhere near finished. Count the sub-structure and it is still early: wave 1-of-3 ran into the end of 2025, wave 2-of-3 corrected into April 2026, and we are now inside wave 3-of-3 — the most powerful stretch of any impulse, and the one that tends to extend rather than terminate. If this is the right label, the recent chop is a pause inside a third wave, not a top, and the trend has room well beyond the prior high.

• The bearish count — a completed five. On the mirror reading, the same rally is a finished five-wave sequence from the 2022 low: the recent high in the 31–32k zone is wave (5), not wave (3). If that is the label, the impulse is complete and the correction now due is of a larger degree than anything since 2022 — a months-long affair, not a dip.

The two agree on the near term — a pause here — and disagree on everything after it. We do not need to pick today; we need the level that tells us which, and the semis are the cleanest early tell.

Inside the semis — a broken steep trend, a double bottom, one clean test. The semiconductor complex (SOXX) ran in a very steep uptrend from early April; that steep line is now broken and over. A broken steep trend is not a broken bull — steep trends almost always give way to shallower ones or a pause, not necessarily a reversal. Price now sits at double-bottom support, and the normal, constructive path from here is another bounce attempt off that base.

The read stays bullish as long as the double bottom holds. It turns only on one of two failures: if the double-bottom low gives way, or if the bounce that follows cannot carry price back to a new all-time high. Either failure argues for a more prolonged correction — and would be the semi-side confirmation of the completed-five count above.

The internals behind the map. The same picture, three ways down: the sector indices show where the absolute return is being made, the rotation map shows what is shifting beneath the surface, and the single-stock movers name which stocks did the leading and the wrecking.

Share Closelooknet

3 · The Outlook

The forward read we hold lightly and reset each week — the price will tell us before the narrative does. This week the question is a single one, and it is a level: does the semi dip get bought.

The four indices — the clearest read we have. We track this cycle through four functional indices, and last week they separated in a way that tells the story better than any single number.

The build-out side — Rubin — was the only one red, dragged by the semis. The three that lean to the deployment of AI were all green: the agentic-opex index (AEI), the agentic winners (AW40) and functional growth (HALO) — with AW40 and HALO out in front of AEI. Read alongside software’s leadership, it suggests the bottom in the software and application complex (IGV) is probably in.

The distinction between the two agentic indices is the point. AW40 is the bet that you know which applications win — a concentrated winners’ list. AEI is the humbler bet: that there will be winners, whoever they turn out to be, and that all of them get taxed by the same toll-booth infrastructure underneath. Which of the two leads over the coming months is genuinely unclear — and we hold it open.

Two scenarios into the seasonally softer late-summer and autumn window:

• The bullish one — all three AI indices (build-out, opex and winners) move up together, HALO with them, and the four form a longer-, medium-term bottom. That is the shape we lean to for July.

• The less-bullish one — the strength narrows, and only one of the three AI indices carries while the others lag.

Either way we expect positive results into the earnings season. The risk we watch is not price but expectations: earnings-growth estimates may be peaking this summer, and if they are, the downward revisions arrive by autumn — the same conclusion we reach, from the other direction, in this week’s Weekly Signal.

If the equipment and memory complex is reclaimed quickly, this week was positioning ahead of the July prints and the dispersion closes from below — the build-out rejoins the deployment side. If it is not, the market is telling us something about AI-capex expectations before a single number is reported, and the build-out core stays the live read on that conversation.

The macro frame sits underneath all of it — whether the inflation impulse is cresting or entrenching is what decides how much multiple the build-out gets to keep, and firmer yields this week are a small vote for caution on the long-duration names.

The point is not to chase the software leader but to watch the semi laggard: the asymmetry into next quarter lies in whether the build-out turns up to rejoin the deployment side, or whether the split widens into the earnings season and the summer.

🔒 Be Informed — the rest is for subscribers. Free: the LinkedIn newsletter + closelook.net. C+ (paid): real-time push on every portfolio move → closelook.net/subscribe.

Read more

In This Edition

(1) This Week’s Action

(2) Overview: The US Market Map

(3) The State

(4) The Outlook

(5) What May Lie Ahea…

Read more

Thank you for reading this week’s edition of Closelook@US Stock Markets. The next edition will be published next Sunday, 31 May. For the schedule of our other publications, please see the end of this document.

Read more

Thank you for reading this week’s edition of Closelook@US Stock Markets, dated April 26, 2026 👋. The next edition will be published on May 3, 2026.

For the schedule of our other publications, please check the publishing schedule at the end.

Read more

Read more

Share

Read more