Brandon Leon

Weekly outlook · recap · the week to November 6, 2026

The Market Did Not Buy the Beat

The market did not buy AMD’s beat, as the VanEck Semiconductor ETF (SMH) closed the week at 463.96, below the 546.40 line set in our Wednesday outlook, −11.2% on the week and −18.3% since the first 2027 spending guides, while the S&P 500 fell 2.7%. October payrolls rose 165,000 versus the 59,000 expected by economists polled by Dow Jones, the Kalshi price of a December rate increase stood at 77.5% late Friday after the report, and Vistra initiated 2027 adjusted EBITDA guidance with an $8.1B midpoint. Our first test resolved in the direction predicted by our financing reading, that investors are repricing the companies most dependent on outside money to fund the AI buildout, while the third was inconclusive. The December test remains pending until Kalshi’s daily close is available.

Author Brandon Leon Posted Friday, November 6, 2026, after the close Data Prices as of Friday’s close (Yahoo Finance); the December-increase price as of 4:33 p.m. ET Friday (Kalshi via DefiRate); the October employment report and Vistra’s release as published Coverage The week in five sessions · the scorecard · what Friday added · where the bill sits · what we carry forward

We Read the Week as a Repricing of the Sellers of AI Capacity

We expected the market to keep selling the suppliers of AI capacity after AMD’s report, and it did. SMH closed at 463.96, 15.1% below the 546.40 line set in our Wednesday outlook, after an 11.2% fall on Thursday that was its worst session since March 2020 and a 3.6% rebound on Friday (Chart 1).1,2 Across the week the ten-year ended at 4.86% and the high-yield fund moved −0.3%, while the Kalshi price of a December rate increase went from 80.5% to 64.5% and back to 77.5%, so the losses landed far harder on chip shares than on the broad market, the bond market or credit. We read that as a repricing inside the sector with financing exposure an important part of it, although returns alone cannot isolate how large a part, and the Democratic sweep added a question on Thursday about whether policy changes could slow approvals for the buildout.

1. The week in five sessions

The Market Asked a Different Question Each Day and Graded the Suppliers on Every One

  • Monday: Oracle said it would sell $10B of stock and $10B of bonds for its AI buildout and Marvell agreed to buy Lumentum in an all-stock deal valued at about $115B of enterprise value, so the week opened with the suppliers raising money and the chip ETF closing at 508.31.3
  • Tuesday: AMD reported a third-quarter non-GAAP gross margin of 56%, meeting the line we had registered, and guided fourth-quarter revenue to approximately $17B, above the $16.14B consensus our October 26 coverage named as the number to beat, and overnight projections showed Democrats winning control of both chambers of Congress.3
  • Wednesday: The Dow rose more than 1,500 points in its best session since April 2025 while the chip ETF fell 2.7%. AMD rose 1.0% in its first session after reporting, the first of the five prints we tested this season that was not sold on the day, and our weekly, dated that Wednesday, carried its tests into the rest of the week.2
  • Thursday: All three averages had their worst day since April 2025 and the chip ETF lost 11.2%, its worst session since March 2020, as the sweep was read as a threat to the pace of the data-center buildout and AMD gave back 17.3%, with the selling spreading well beyond technology.2
  • Friday: Payrolls rose 165,000 versus the 59,000 expected by economists polled by Dow Jones, and the chip ETF recovered 3.6% with AMD up 11.0%, while the Dow slipped 0.5%.4,5

For the week the S&P 500 fell 2.7%, the Nasdaq Composite fell 3.4%, the Dow fell 0.9% and the semiconductor ETF fell 11.2%, and the VIX ended at 25.6 after touching 28 on Thursday. The losses were disproportionately concentrated in semiconductors, although Thursday’s selloff spread across the broad market (Chart 1).1

Chart 1: The Week in Five Sessions

SMH ended the week down 11.2% while the Dow fell 0.9%, with Thursday doing most of the damage.

October 30 close = 100 84 88 92 96 100 104 Fri Oct 30 Mon Tue Wed Thu Fri Nov 6 Dow −0.9% S&P 500 −2.7% Nasdaq Composite −3.4% Semiconductors (SMH) −11.2%

The week’s losses concentrated in semiconductors, although Thursday’s selloff spread across the broad market and Friday’s rebound was led by the chips.

2. The scorecard

The Market Did Not Buy the Beat

Our first test said that, with AMD’s guide at or above $16.14B, a semiconductor ETF close on Friday above 546.40 would mean the market had bought a beat and our no-bottom rule would lift. A close at or below it would mean the de-rating had moved past the numbers and the financing reading held (Chart 2).6 SMH closed at 463.96, 15.1% under the line, so under its original criteria the test left our financing reading intact, and the rule from October 23, no bottom call until the market buys a beat, stands into its final reading on November 20.

December Recovered on the Labor Data

Our second test said a Kalshi price of a December increase at or above 69.5% at Friday’s close would mean the committee’s path survived the labor data, below it that the data had started to move December, and below 50% the regime change we named after the October 28 hold.7 The price fell to 64.5% at Thursday’s close on the selloff, a day before the October employment report, and stood at 77.5% at 4:33 p.m. ET Friday after the report, so the reading is provisionally consistent with the committee’s December path remaining intact, while the test itself grades on Kalshi’s daily close in the DefiRate series, which we log in our ledger when it posts (Chart 3).

Friday Left Thursday’s Selloff Inconclusive

The third line, which we registered in our ledger on Thursday evening before Friday’s open, said a semiconductor ETF close above 470.00 would read Thursday as positioning and a close at or below 447.73, Thursday’s close, as a selloff with legs, with any close between those thresholds deemed inconclusive.6 SMH closed at 463.96, inside the inconclusive range our line specified, after taking back 29% of Thursday’s loss, so under its original criteria the line supports neither a positioning-only interpretation nor a sustained-selloff interpretation. The standing indicators give a mixed reading, as the ten-year at 4.86% sits well below the 5.31% that puts the pressure back while Lumentum ended at a 15.2% discount to the implied value of the 4.5 Marvell shares offered per share, past the 15% at which our card says the market doubts the close.1

Chart 2: The Beat Test and the Thursday Line

The chip ETF closed 15.1% under the line for a bought beat and inside the band that reads Thursday.

SMH daily close, $ 440 460 480 500 520 540 560 580 Oct 16 Oct 20 Oct 22 Oct 26 Oct 28 Oct 30 Nov 3 Nov 5 Nov 6 546.40: a bought beat (test 1) 447.73 to 470.00: inconclusive band (test 3) 463.96

Both lines are graded on Friday’s close as written. The shaded band is the inconclusive range of the line we registered on Thursday evening, before Friday’s open.

Chart 3: December, Under the Line and Back

The December-increase price fell under the 69.5% line on Thursday’s selloff and rose back to 77.5% after Friday’s jobs report.

Kalshi: the Fed raises on December 9, %; daily closes, then a Friday intraday reading 50% 60% 70% 80% 90% Oct 14 Oct 17 Oct 20 Oct 23 Oct 26 Oct 29 Nov 1 Nov 4 Nov 6 69.5%: December survives the data (test 2) 50%: the regime change 77.5% at 4:33 p.m. ET Friday (not yet a close) 64.5% Thu

Kalshi daily closing prices through November 5, followed by an intraday observation of 77.5% at 4:33 p.m. ET on November 6, so Friday’s test result remains pending until the daily close posts. Thursday’s drop came a day before the October employment report, on the same selloff that took the chips down, and Friday’s recovery followed the report.

3. What Friday added

A Stronger-Than-Expected Payroll Report Was Followed by Higher Odds of a December Increase

The Bureau of Labor Statistics reported that payrolls rose 165,000 in October versus the 59,000 expected by economists polled by Dow Jones, with August and September revised up by a combined 84,000, the unemployment rate at 4.1% and average hourly earnings up 0.4% on the month and 3.1% on the year.4 Breadth: professional and business services added 70,000, health care 42,000, transportation and warehousing 25,000 and manufacturing 19,000, gains across several industries that a committee that raised rates in September can read as room to raise again. The Kalshi price of a December increase rose from 64.5% to 77.5%, 13 percentage points, while the ten-year ended at 4.86%. The chips: the semiconductor ETF rose 3.6% on the same day, which does not fit a simple account in which higher odds of an increase drive chip shares down.7

Vistra’s 2027 Guidance Midpoint Matched the Number Our October Note Set

Vistra reported third-quarter Ongoing Operations Adjusted EBITDA of $2,378M against $1,581M a year earlier and narrowed its 2026 guidance to $7.1B to $7.5B. It initiated 2027 adjusted EBITDA guidance of $7.7B to $8.5B, a midpoint of $8.1B that matches the number our October 3 note set, with $4.3B to $5.1B of adjusted free cash flow before growth and an additional $1.5B share-repurchase authorization.8 Our note asked for $8.1B including Cogentrix and Meta and the release does not say what the range includes, so the grade against that wording belongs to our Vistra coverage, which will read the call and the filings. The release also announced a 20-year, 207 MW power purchase agreement at its Odessa plant in Texas and said its 405 MW Pulaski solar facility in Illinois had commenced operations under a power purchase agreement with Microsoft, and the shares rose 10.6% on the day.

4. Where the bill sits

Most Hyperscale Buyers Held Up While the Suppliers and Financing-Exposed Names Fell

Measured from the October 28 close, the night the first 2027 guides arrived, Microsoft, Amazon and Meta are between −0.4% and +1.8%, with Alphabet the exception at −12.3%. The suppliers are down 10% to 30%, the five companies in our sample that raised money or paid in shares 20% to 44%, the two listed renters of capacity 25% to 29% and the two data-center landlords in our sample −3.1% to +0.5% (Chart 4).1 That ordering is what the financing reading predicts, with the companies that need outside money to build down the most and the companies whose cash pays for the buildout down the least, although a ranking of returns cannot by itself separate financing from valuation. Bank of America’s Wamsi Mohan made the valuation side of that point on Friday, writing that the sell-off in AI-exposed hardware “has come from lower valuation multiples, not lower earnings estimates.”5

Chart 4: Since the First 2027 Guides

Three of the four buyers are roughly flat, the suppliers are down 10% to 30%, and the companies that raised money are down 20% to 44%.

From the October 28 close, % −50% −40% −30% −20% −10% 0% Buyers Microsoft −0.1% Alphabet −12.3% Amazon +1.8% Meta −0.4% Suppliers Nvidia −13.3% Broadcom −16.5% TSMC −14.6% Qualcomm −9.6% AMD −22.4% Micron −29.8% Arm −26.0% Raised money Oracle −19.7% Intel −41.1% CoreWeave −27.4% Marvell −43.9% Lumentum −41.1% Renters Nebius −29.0% Iren −25.0% Landlords Digital Realty −3.1% Equinix +0.5% Power Vistra −12.6% Constellation −3.2%

The two landlords in our sample held up substantially better than the renters and the companies that raised money, which is where a financing reading expects the bill to land.

Friday’s Rebound Partly Reversed Thursday

Friday’s rebound sorted by Thursday’s damage, with a correlation of −0.62 between the two days’ moves across 22 names, which is consistent with a partial reversal of Thursday’s selling but cannot tell positioning apart from a change in fundamentals or in appetite for risk (Chart 5). The chip ETF took back 29% of Thursday’s loss, short of the 470.00 our third line set, so on our own framework Thursday reads neither as positioning nor as a selloff with legs. The renters’ side of the bill also arrived in writing, as Iren reported fiscal first-quarter revenue of $228M against $137M in the prior quarter and said more than $4B of annual recurring revenue “has been delayed until at least January 2027,” which moves part of a renter’s revenue into next year.5

Chart 5: Did the Worst-Hit Bounce the Most?

The names Thursday hit hardest bounced the most on Friday, a correlation of −0.62.

Each dot is one company: Thursday's move (across) against Friday's (up) −20% −16% −12% −8% −4% 0% 0% +4% +8% +12% +16% +20% Thursday, November 5, % fitted line, correlation −0.62 Microsoft Alphabet Amazon Meta Nvidia Broadcom TSMC Qualcomm AMD Micron Arm Oracle Intel CoreWeave Marvell Lumentum Nebius Iren Digital Realty Equinix Vistra Constellation Buyers Suppliers Raised money Renters Landlords Power

The relationship is consistent with a partial reversal of Thursday’s selling; it does not show who traded or why.

5. What we carry forward

Nvidia’s Report Is the Next Beat the Market Can Buy or Sell

We carry the financing reading into the second half of the month, with the no-bottom rule standing until the market buys a beat and its final reading on November 20. We see Nvidia’s report later this month as the next earnings release that could lift the no-bottom rule, because its quarter speaks to demand directly and, through its receivables and its investments in customers, to who is paying for the buildout. The lines below are standing reading guides carried from earlier notes and not new tests, and the next weekly sets the next tests before the events they grade.

What would change our view — standing lines, not new tests
  1. The no-bottom rule (final reading November 20): it lifts only when the market buys a beat, and a chip ETF weekly close above 546.40 remains the line that says it has.
  2. December into the December 9 meeting: the Kalshi price at or above 69.5% keeps the committee’s path intact; below 50% is the regime change we named on October 28.
  3. The ten-year: a Friday close at or above 5.31% puts the pressure back, and above 5.10% says the supply is biting.
  4. The financing names: Lumentum’s discount to the implied value of 4.5 Marvell shares wider than 15% says the market doubts the close and under 8% says it believes it; CoreWeave’s five-year credit default swap spread below 800 basis points heals and at or above 1,140 deepens.
What makes this note wrong

A week that held a midterm sweep, a supplier equity wave, a stronger-than-expected payroll report and the largest single-day moves in chip shares since 2020 cannot be assigned to one cause, and the skeptic will say we read our October thesis into all of it. The financing reading rests on five companies that each had their own reason to raise money and on a ranking of returns that fits the reading without proving it, while the policy reading rests on television framing and a governor-elect’s comments and not on a bill or a rule. The author holds several of the names discussed, which is a reason to check this reading harder and not a reason to trust it.

What it means: AMD’s good report did not stop the selling in chip shares, Thursday’s drop in the chip ETF was the largest since 2020 and spread across the market, and a strong jobs report on Friday was followed by a partial rebound even as the odds of a December rate increase rose. Three of the four big cloud buyers are roughly where they were before their spending guides, while most of the companies that sell them chips and capacity are down by double digits.

6. Something to think about

The Odds of an Increase and the Chips Moved Together

On Thursday the Kalshi price of a December increase fell 16 percentage points while the chip ETF fell 11.2%. On Friday the price rose 13 percentage points after the payroll report while the chip ETF rose 3.6%, so over these two sessions the odds of tighter money and the chip shares moved in the same direction.

That pattern does not fit a simple account in which higher odds of an increase drive chip shares down. The more plausible reading is that the December price itself responded to Thursday’s selloff and to Friday’s data, although two sessions cannot isolate financing or policy as the cause that moved the chips instead. It is also the reason we grade on lines written before the events and not on the explanation of the day, because the explanation changed four times in five sessions while the closes kept their own record.

Which leaves the question for the rest of the month: if the odds of tighter money and the price of the chips keep moving together, what is the market telling us it fears more than the Fed?

Get the next note by email →   All writing

Sources

  1. Prices from Yahoo Finance, as of Friday’s close: S&P 500 7,376.86 (+0.3% on Friday, −2.7% for the week), Nasdaq Composite 24,579.42 (+1.1%, −3.4%), Dow 51,285.73 (−0.5%, −0.9%), SMH 463.96 (+3.6%, −11.2%; −18.3% since October 28), VIX 25.64, ten−year 4.86%, HYG −0.3% for the week. SMH daily closes: Oct 28 567.67; Oct 29 561.67; Oct 30 522.49; Nov 2 508.31; Nov 3 518.52; Nov 4 504.27; Nov 5 447.73; Nov 6 463.96. Company prices and weekly moves: Microsoft 530.61 (+1.7% for the week); Alphabet 284.78 (+1.8% for the week); Amazon 259.30 (+7.9% for the week); Meta 610.15 (−0.8% for the week); Nvidia 186.64 (−8.8% for the week); Broadcom 293.69 (−9.9% for the week); TSMC 375.44 (−7.6% for the week); Qualcomm 165.41 (+0.4% for the week); AMD 449.04 (−11.0% for the week); Micron 690.00 (−22.0% for the week); Arm 174.62 (−15.2% for the week); Oracle 116.10 (−11.2% for the week); Intel 56.28 (−14.4% for the week); CoreWeave 46.40 (−15.4% for the week); Marvell 141.20 (−28.1% for the week); Lumentum 538.59 (−35.9% for the week); Nebius 129.49 (−16.0% for the week); Iren 21.64 (−9.6% for the week); Digital Realty 174.20 (−3.6% for the week); Equinix 1,012.71 (−0.1% for the week); Vistra 124.80 (−9.8% for the week); Constellation 249.95 (−4.7% for the week). From the October 28 close: Microsoft −0.1%; Alphabet −12.3%; Amazon +1.8%; Meta −0.4%; Nvidia −13.3%; Broadcom −16.5%; TSMC −14.6%; Qualcomm −9.6%; AMD −22.4%; Micron −29.8%; Arm −26.0%; Oracle −19.7%; Intel −41.1%; CoreWeave −27.4%; Marvell −43.9%; Lumentum −41.1%; Nebius −29.0%; Iren −25.0%; Digital Realty −3.1%; Equinix +0.5%; Vistra −12.6%; Constellation −3.2%. Thursday and Friday moves: Microsoft −3.6% / −1.0%; Alphabet −2.2% / +0.8%; Amazon −2.2% / −0.3%; Meta −4.0% / +0.8%; Nvidia −9.3% / +2.8%; Broadcom −10.4% / +0.7%; TSMC −10.8% / +6.3%; Qualcomm −9.5% / +5.5%; AMD −17.3% / +11.0%; Micron −17.5% / +5.1%; Arm −13.1% / +4.7%; Oracle −6.9% / +0.8%; Intel −10.3% / +7.7%; CoreWeave −9.8% / +6.3%; Marvell −16.4% / +5.7%; Lumentum −18.7% / +5.7%; Nebius −14.9% / +7.6%; Iren −17.0% / +20.0%; Digital Realty −5.5% / −0.3%; Equinix −5.4% / +1.5%; Vistra −6.2% / +10.6%; Constellation −2.5% / +4.3%. Lumentum’s discount to the implied deal value: 4.5 Marvell shares × 141.20 = 635.40 against Lumentum at 538.59, a 15.2% discount. Correlation of Thursday and Friday moves across these 22 names: −0.62. Yahoo daily history: SMH’s −11.21% on November 5 was its largest one-day decline since March 16, 2020 (−14.41%); SOXX’s was −12.15% (−15.23% on March 16, 2020). ↩
  2. CNBC live blog, updated November 5, 4:18 p.m. EST: “All three major averages posted their worst day since April 2025”; the SOXX semiconductor ETF had its “worst day since the Covid-19 breakout of March 2020”; the Dow “more than wiped out its advance from Wednesday’s massive postelection rally,” a gain of more than 1,500 points and its best day since April 2025; the VIX “briefly climbed to 28.09 Thursday”; “investors accelerated their move away from technology stocks on fears of a data-center buildout slowdown.” ↩
  3. Monday and Tuesday are from our November 4 weekly and its sources: Oracle’s $10 billion stock and $10 billion bond offerings, the Marvell-Lumentum terms (4.5 Marvell shares per Lumentum share, about $115 billion of enterprise value), AMD’s Q3 2026 8-K Ex. 99.1 (third-quarter non-GAAP gross margin 56% as reported, 55.94% unrounded from $7,483 million of non-GAAP gross profit on $13,377 million of revenue; fourth-quarter outlook of revenue “approximately $17 billion, plus or minus 3%,” and non-GAAP gross margin “approximately 55%”), and NBC News’ projections of Democratic control of both chambers. The $16.14 billion figure is the fourth-quarter revenue consensus our October 26 AMD coverage named as the number to beat, with a soft-guide line below $15.6 billion; the line registered in our ledger as FT-AMD-20261103-REUW was a third-quarter non-GAAP gross margin at or above 56%, set in the company’s whole-percent reporting unit and graded on the reported 56%. ↩
  4. Bureau of Labor Statistics, The Employment Situation, October 2026 (USDL-26-1632, released November 6, 8:30 a.m. ET): “Total nonfarm payroll employment rose by 165,000 in October, and the unemployment rate changed little at 4.1 percent”; professional and business services +70,000, health care +42,000, transportation and warehousing +25,000, manufacturing +19,000; average hourly earnings +14 cents, or 0.4 percent, to $38.00, and +3.1 percent over 12 months; August revised from +133,000 to +186,000 and September from +29,000 to +60,000, “84,000 higher than previously reported.” ↩
  5. CNBC live blog, updated November 6, 2:54 p.m. EST: payrolls “well above the 59,000 that economists polled by Dow Jones had anticipated”; Bank of America’s Wamsi Mohan, “The recent sell-off in AI-exposed IT hardware has come from lower valuation multiples, not lower earnings estimates”; Iren’s fiscal first-quarter revenue of $228 million against $137 million in the prior quarter and more than $4 billion of annual recurring revenue that “has been delayed until at least January 2027.” ↩
  6. Ledger: the beat test is CALL-WEEKLY-20261104-BUYTHEBEAT (seq 264) and the December test CALL-WEEKLY-20261104-DECX2 (seq 265), both preregistered November 3; the Thursday line is CALL-THU-20261105-LEGS (seq 272), preregistered November 5 after the close and before Friday’s open; the no-bottom rule is CALL-SEMIS-BOTTOM-20261023 (seq 221/256, final reading November 20). ↩
  7. December-increase pricing: Kalshi via the DefiRate aggregator, daily closes Oct 28 79.5%; Oct 29 70.5%; Oct 30 69.5%; Oct 31 69.5%; Nov 1 69.5%; Nov 2 72.5%; Nov 3 74.5%; Nov 4 80.5%; Nov 5 64.5%; the Friday reading is an intraday observation of 77.5% at 4:33 p.m. ET Friday, and the daily close had not posted when this note was published. The 69.5% and 50% lines are from our October 28 Fed note and our November 4 weekly. ↩
  8. Vistra, “Vistra Reports Third Quarter 2026 Results,” November 6, 2026: Ongoing Operations Adjusted EBITDA $2,378 million against $1,581 million; 2026 ranges narrowed to $7.1 to $7.5 billion (EBITDA) and $4.3 to $4.7 billion (FCFbG); 2027 initiated at $7.7 to $8.5 billion and $4.3 to $5.1 billion (the $8.1 billion midpoint is our arithmetic; the release does not say whether the 2027 range includes Cogentrix); “an additional $1.5 billion of share repurchases” authorized; “a 20-year 207 MW power purchase agreement with New Era Energy and Digital at our Odessa combined cycle gas plant in Texas”; the 405 MW Pulaski Solar Facility “under a power purchase agreement with Microsoft.” The $8.1 billion line is in our October 3 note: “At or above $8.1 billion including Cogentrix and Meta confirms our base case.” ↩

Nothing on this page is investment advice. We work in scenarios and trip-wires, not predictions. See disclaimer.

Disclosure: at the close on November 6, 2026, the author holds long positions in AMD, Marvell, Micron, Nvidia and Alphabet (class C shares), per the author’s brokerage positions page at 12:45 p.m. ET on November 4 and the author’s confirmation that the only trade since was an addition to the Nvidia position on November 5. AMD, Marvell and Micron closed through the author’s approved exit levels on November 5, and Nvidia through its approved partial-exit level; exits are judged on closing prices, and none had been executed or placed by Friday’s close. The author holds no position in Microsoft, Amazon, Meta, Broadcom, TSMC, Qualcomm, Arm, Oracle, Intel, CoreWeave, Lumentum, Nebius, Iren, Digital Realty, Equinix, Vistra or Constellation. This note is research and opinion, published to all readers alike, and is not investment advice for any person; the tests above are the author’s own and are graded in public against their written wording. bpleon research received no compensation from any company named.