Weekly outlook · two weeks to October 30, 2026
The Bill Moved Down the Supply Chain
Two weeks ago the question was whether the cloud buyers would keep spending. Three of the four said more, every proof test we set passed, and the shares were sold on each print anyway. Within three sessions the sellers were the ones raising money: Marvell paying $115 billion in its own stock for Lumentum, Oracle selling $10 billion of shares and $10 billion of bonds. The market’s question has moved from who buys to who pays, and it is now asking the suppliers.
Everyone watched the two events this column said would settle October: the Federal Reserve on the 28th and the first 2027 spending guides from the large cloud buyers. Both landed on the side the demand case needed. The committee held. Alphabet said its capital spending will “increase significantly” next year, Amazon said “meaningfully more,” Microsoft said its spending accelerates into the first half of its fiscal 2027, and the three proof tests we attached to the prints passed.1
That should have ended the argument. Instead the Nasdaq fell 2.9% on Friday, its worst session since June, and the semiconductor index fell 7% in a day on a PC-chip guide from Intel.2 The ten-year yield closed Friday 38 basis points below its October high, so the explanation we ruled out two weeks ago stays ruled out.
What the market graded was the cost of the capacity the buyers had just committed to. Three of the four quarters came with negative or shrinking free cash flow, and by Monday the suppliers were the ones issuing stock. That is the financing reading from October 20, moved one step down the supply chain.
1. The two weeks in one picture
From the October 16 close to the October 30 close the Dow fell 0.4%, the S&P 500 1.0%, the Nasdaq 2.7% and the semiconductor ETF 3.2%. The shape matters more than the end points. The chip index rose 7.7% into the Wednesday close before the first guides and lost 10.5% in the three sessions after them.3 October was the Nasdaq’s worst month since March 2025, down 5.3%, while the Dow added 1.7%.2
Chart 1: The Two Weeks in One Picture
The chip index rallied 7.7% into the first guides and lost 10.5% in the three sessions after them.
The same dispersion as two weeks ago, with one new feature: the chip line broke on the strongest demand evidence the buyers have given all year, after a hold from the Fed.
2. The scorecard
We set four tests on October 20 and a fifth in the Fed note. Tests are graded against their written wording, so the grades are the ledger’s. The one we would argue with is explained, not changed.
| Test, as written | What happened | Grade |
|---|---|---|
| 2027 spending direction: higher across the buyers; flat or lower at more than one means the cycle is turning. | Alphabet up, Amazon up, Microsoft up into early 2027, Meta no direction. Nobody lower. | Inconclusive |
| CoreWeave five-year default swap: below 800 basis points heals; at or above 1,140 deepens. | 1,041 on October 27, the last reading we can date. Shares down 37% in October. | Standing |
| The Fed: a hold changes nothing; any increase is a surprise. | Held 10 to 2, both dissents for an increase. | As written |
| Ten-year Friday close: at or above 5.31% keeps the pressure on; below, and sustained, eases one. | 4.99% on October 23, 4.93% on October 30. | Eased |
| December increase priced at or above 72% at the October 30 close (the Fed note). | 69.5% on Friday; 74.5% by Tuesday afternoon. | Dead |
The capex grade reads as a pass by this column’s wording and as inconclusive by the stricter scoring we registered on October 26, which required every buyer that spoke to point up. Meta would only say it keeps investing aggressively. We grade to the stricter text. What matters for the argument is that nobody pointed lower, so the turning branch never fired.1
Chart 2: The December Call, Graded
The December-increase price fell through the line on the reading the note named, then went back above it.
Graded against its wording: dead at 69.5% on October 30. The hike price recovered on Monday while the chip index fell another 2.7%, so the committee’s December is not what moved the chips.
The Fed test died on the reading it named: 79.5% the evening of the hold, 69.5% two days later, under a line set one point below the pre-meeting low, which left no room for a Friday wobble. It has since gone back to 74.5%. We do not re-grade a dead call on a later print; the lesson is about the line, and it is recorded.3 What the round trip shows is that the committee’s December was not what moved the chips. The hike price recovered on Monday while the chip index fell another 2.7%.
3. Passed tests, sold shares
Set the prints side by side. Meta raised its 2026 spending floor on October 21 and fell 7% the next day. Microsoft beat the Azure line, 48% against 45, and closed flat. Alphabet’s cloud backlog cleared $560 billion at $561 billion and the stock had its worst day on record, down 12.6%. Amazon’s AWS grew 39.3% against the 37% test and fell 1.4% in a market down 2.9%. Intel, which had no test because its reaction turns on foundry news with no number attached, guided the fourth quarter to $15 to $16 billion against a $17.1 billion consensus and fell 29%.4 AMD cleared its line on Tuesday night, rose 7% in late trading and closed Wednesday up 1.0% at $488.78 while the chip ETF fell 2.7%, the first beat of the season that was not sold, which is why it is the week’s test and not this fortnight’s story.7
Chart 3: Passed Tests, Sold Shares
Five prints cleared the number set in advance. Four were sold on the day. AMD’s was the first one not sold.
Every proof test measured demand and passed. The four sold quarters shared the bill: negative free cash flow or outside funding. Intel had no test. AMD’s bar is Wednesday’s close: up 7% in late trading Tuesday, most of it given back overnight on the election result, +1.0% on a day the chip ETF fell 2.7%.
What the four sold passes had in common was cash. Alphabet’s free cash flow was negative $9 billion for the quarter. Amazon’s trailing free cash flow is negative $42 billion, and it has raised $115 billion of long-term debt this year against $0.7 billion in the same nine months of 2025. Microsoft will use outside funding when the terms are attractive. Meta’s quarter was cash-negative by $2.1 billion.1 After October 28 the question was who writes the checks, and three of the four answers involved somebody other than the company’s own operations.
Intel is the same story from the other end. Its release blames a PC slowdown it expects to last through the first half of 2027 and cites industry-wide shortages of substrates and memory, the shortage Micron reported as its best quarter ever.4 That reading is ours, and neither company made it: part of the PC’s bad quarter is the AI buildout’s bill arriving at a different customer. The market sold the PC guide as a chip-demand signal and took AMD, Qualcomm and Arm down with it.
4. Rates eased, and the companies that need money fell
Start with what did not move. The ten-year closed October 30 at 4.93%, down from 5.10% on October 16 and 5.31% at the October 5 high, and closed Tuesday at 4.95%. The investment-grade and high-yield bond funds ended the stretch within 1% of where they began.3 A rates event, or a general credit event, does not take a tenth off the chip index in three sessions while both sit still.
Chart 4: Rates Eased and the Chips Fell Anyway
Two Friday closes under the 5.31% line while the chip index lost a tenth of its value.
The October 20 test said a sustained Friday close under 5.31% would ease one pressure. It did, and the bond funds moved less than 1%. What fell was the group that has to raise money to build.
What moved was the group that has to raise money to build. CoreWeave’s shares fell 37% in October and its default swaps were last quoted, on the 27th, near 1,040 basis points, between our two lines. A Financial Times report on Friday, relayed by CNBC, said Blackstone had told investors that Anthropic may not get an investment-grade rating on its debut.5 Then Monday. Oracle said it will sell $10 billion of stock and $10 billion of bonds for its AI buildout and fell 8%. Marvell announced an all-stock purchase of Lumentum and fell 13%; Lumentum fell 18%.6 Intel’s release, for its part, lists a $23 billion public equity offering completed in the quarter.4
The Marvell terms say where the bill went. Each Lumentum share becomes 4.5 Marvell shares, which the companies value at $115 billion of enterprise value, about $1,132 a share and a 23.9% premium to the October 28 close, the last day before the talks leaked. The close is expected in the first half of 2028. Management calls the deal accretive in its first year after closing; the presentation does not say whether that counts the billion dollars of synergies it promises, and our arithmetic on calendar-aligned numbers puts it about 2% dilutive before them.6 From the unaffected close to Monday, Marvell lost 32% and Lumentum 25%. At Tuesday’s close the ratio valued Lumentum at $774 a share and Lumentum traded at $680, a 12% gap for an eighteen-month wait.3
Chart 5: Paying With Shares the Market Is Marking Down
From the unaffected close, the buyer lost 32% and the target 26%.
At Tuesday’s close 4.5 Marvell shares were worth $774 and Lumentum traded at $680, a 12% gap for a close expected in the first half of 2028.
A supplier paying for capacity with its own shares, while the market marks those shares down, is the financing question arriving at the sellers’ end of the chain. On October 20 the stressed balance sheet belonged to a customer. On November 2 it belonged to Oracle, Intel and Marvell, each of which chose equity. That is the strongest evidence so far that the drawdown is about the cost of capacity and not the demand for it, with one caveat we cannot size: Friday had month-end and election positioning in it too.2
5. What could clarify the outlook
AMD reported Tuesday after the close: revenue $13.4 billion, up 45%, data center a record $8.5 billion, up 97%, a non-GAAP gross margin of 56% with the first Helios racks running at Anthropic, Meta, Microsoft, OpenAI and Oracle, and a fourth-quarter guide of about $17 billion. Our lines were a guide at or above $16.14 billion and a margin at or above 56% with racks shipping. Both were met. The shares rose about 7% in late trading, gave most of it back overnight as the election result came in, and closed Wednesday up 1.0% at $488.78 on a day the chip ETF fell 2.7%.7 Two beats this year were sold on the day; this one was not, and whether that holds to Friday is the week’s test. The same release shows the supplier side of the pattern: free cash flow of $554 million, down from $1.6 billion, and total debt of $7.1 billion against $3.2 billion in June.
The midterms were Tuesday and the count came in a Democratic sweep. NBC News projected the House for Democrats at 11:04 p.m. Eastern and the Senate at 11:23, with six seats flipped (Iowa, Kansas, North Carolina, Ohio, South Carolina and Texas) for 51 called against 39 and ten races open; Republicans had gone in holding the chamber 53 to 47, and prediction markets had made Democrats favorites in both chambers. The sell-side framing CNBC carried on Monday was that a sweep was not priced and would weigh on AI shares, and the first answer agreed: at 11:20 p.m. S&P 500 futures were down 1.2% and Nasdaq-100 futures 2.1%, the ten-year yield had slipped to 4.90%, and AMD’s late-trading gain had shrunk to about 3%, with CNBC tying the move to the sweep and to the Texas governor-elect’s pledge of a data-center moratorium. By 9:25 p.m. Wednesday NBC had the Senate at 54 to 44 with seven seats flipped (Maine added) and two races uncalled, and the House at 242 to 151. The cash session reversed the futures: the S&P 500 closed up 1.6%, the Dow 3.0% and the Nasdaq 1.4%, while the chip ETF fell 2.7% and the ten-year ended at 4.89%. We have no edge on the count and set no test on it; the week’s tests stand as written.8
Also this week: Treasury’s refunding details Wednesday at 8:30 a.m. Eastern, with $618 billion of borrowing estimated for the quarter; Coherent and Arm after the close Wednesday; the October jobs report Friday at 8:30, after September’s 29,000; and Vistra’s first 2027 guide on Friday, graded against the $8.1 billion line in our October note.9
- AMD and the reaction (Friday’s close): no bottom call until the market buys a beat. AMD cleared its line, so if the semiconductor ETF closes Friday above 546.40, the October 16 weekly close, the market has bought a beat and the rule lifts. At or below 546.40, the de-rating has moved past the numbers and the financing reading holds.
- December pricing after the jobs report (Friday’s close): at or above 69.5%, the post-meeting low, the committee’s path survives the labor data. Below it, the data has started to move December. Below 50% is the regime change we named on October 28.
- The ten-year (standing): a Friday close at or above 5.31% puts the pressure back. A close back above 5.10%, the October 16 level, after Wednesday’s refunding says the supply is biting without a new high.
- The financing names (standing): CoreWeave’s five-year swap below 800 basis points heals, at or above 1,140 deepens. The gap between Lumentum’s price and 4.5 Marvell shares wider than 15% says the market doubts the close; under 8% says it believes it.
Three equity raises in a week is a pattern to us and a coincidence to a skeptic, and the skeptic has a point: Intel’s offering closed before any of this, Oracle’s need was known, and Marvell’s ratio was set at a price three weeks old. Friday had month-end and election positioning we cannot separate out. The capex grade is inconclusive on our own scoring, so the demand evidence is three statements of direction without a 2027 number among them. And the author holds several of the names that fell, Marvell among them, which is a reason to check this reading harder, not a reason to trust it.
6. Something to think about
On January 10, 2000, America Online agreed to buy Time Warner for about $165 billion in AOL shares, the largest merger ever at the time. The acquirer had the expensive stock; the target had the cable systems, the magazines and the film studio; and the structure was all paper, with AOL holders keeping 55% of the result. The Nasdaq peaked two months later.10
It is remembered as the worst deal in corporate history, which is near the lesson without being it. The lesson is what an all-stock offer says about the buyer’s view of its own currency. A board that pays in shares would sooner hand you its stock than its cash, and the market reads that message whether or not the board means to send it. Marvell fixed its ratio at a price from October 28, and the market has since marked that currency down by a third.
Which leaves the question for every all-stock deal struck inside a de-rating: is the buyer’s stock the cheapest thing it can pay with, or the most expensive thing it owns?
Sources
- The 2027 directions, the proof tests and the cash figures are in our capex note, A Mixed Answer From the Buyers (posted November 3), with its primary sources: Alphabet Q3 2026 8-K exhibit and call (October 28), Microsoft FY27 Q1 release and call (October 28), Amazon Q3 2026 8-K Ex. 99.1 (October 29: AWS +39.3%, trailing free cash flow −$42.4 billion, $115.1 billion of long-term debt proceeds in nine months against $0.7 billion a year earlier, 2026 cash capex about $220 billion), and Meta’s Q3 2026 release (October 21). The capex call is graded in our ledger as inconclusive on the scoring registered October 26. ↩
- Friday, October 30, 2026 session: Nasdaq Composite −2.86% to 25,437.00, its worst day since June; S&P 500 −1.82% to 7,580.07; Dow −1.09% to 51,770.31; the SMH semiconductor ETF down 7% intraday; Intel “tanked almost 29%” after “a weak outlook for the current quarter, pointing to a slowdown in PCs”; AMD “close to 11%”, Nvidia 4%, Broadcom 5%, Qualcomm 11% at midday; October: Nasdaq −5.3%, its steepest month since March 2025, S&P 500 −0.9%, Dow +1.7%; “investors took money off the table into month-end and ahead of next week’s contentious midterm elections”; the Financial Times report on Blackstone and Anthropic; CoreWeave “cratered 37%” in October (CNBC live blog, updated October 30, 4:20 p.m. ET). ↩
- Daily closes from Yahoo Finance: SMH 539.79 (October 16), 527.20 (October 23), 567.67 (October 28), 522.49 (October 30), 508.31 (November 2), 518.52 (November 3); Dow, S&P 500 and Nasdaq Composite closes for the same dates; ten-year yield 5.10% (October 16), 4.99% (October 23), 4.93% (October 30), 4.95% (November 3), with the 5.31% intraday high on October 5 as cited in the October 20 outlook; HYG −0.8% and LQD −0.3% from October 28 to November 3; Marvell 251.65 (October 28), 171.41 (November 2), 171.97 (November 3); Lumentum 913.95, 684.40, 680.16. December-increase pricing: Kalshi daily closes via the DefiRate aggregator, 79.5% (October 28), 70.5% (October 29), 69.5% (October 30), 72.5% (November 2), 74.5% (November 3, 2:00 p.m. ET observation); the October 30 grade in our ledger used the Kalshi trade feed directly (last trade 0.69). ↩
- Intel Q3 2026 results, 8-K Ex. 99.1, October 29, 2026: revenue $16.5 billion, +21%; Client Computing and Physical AI Group $8.1 billion, −5%; Data Center and AI $7.3 billion, +78%; Intel Foundry $6.7 billion, +58%, operating loss $1.85 billion; non-GAAP gross margin 44.1%; Q4 outlook revenue $15.0 to $16.0 billion, non-GAAP gross margin 38.0%, non-GAAP EPS $0.25; “we see the PC slowdown continuing at least through the first half of next year” (CFO David Zinsner); “unlock additional supply to meet extraordinary demand for our Xeon family” (CEO Lip-Bu Tan); risk language citing “current industry-wide substrate and memory shortages”; “Completed a $23 billion public equity offering.” Consensus for Q4 of $17.1 billion revenue and $0.43 EPS per LSEG (CNBC, October 30 premarket). Micron: our October 19 update. Next-session moves from Yahoo closes: Meta 697.52 to 648.85 (−7.0%); Microsoft 531.01 to 530.14 (−0.2%); Alphabet 324.89 to 283.82 (−12.6%); Amazon 243.79 to 240.35 (−1.4%); Intel 92.07 to 65.76 (−28.6%). ↩
- CoreWeave five-year credit default swap: 1,041.4 basis points, “the lowest in more than a week,” Markit data as reported October 27 (roic.ai); the October 15 record above 1,140 (CNBC) or 1,194.5 intraday (ICE, as reported). We found no reading we could date after October 27; the test stays open until we can. ↩
- Marvell and Lumentum: joint release, 8-K Ex. 99.1, November 2, 2026 (4.500 Marvell shares per Lumentum share; $115 billion implied enterprise value, about $1,132 per share; 23.9% premium to the October 28 unaffected close; about 67%/33% pro forma ownership; close expected in the first half of calendar 2028; more than $1 billion of annual run-rate cost synergies; accretive to non-GAAP EPS in year one post-close); 8-K Item 1.01 (termination fees $5.8 billion payable by Marvell, $2.9 billion by Lumentum, $3.2 billion regulatory); investor presentation, Ex. 99.2. Monday moves and Oracle: “Marvell Technology shares dropped almost 13%, while Lumentum shed 18%”; “Oracle shares also declined 8% after the company said it plans to sell $10 billion in common stock as well as $10 billion in bonds to finance its AI buildout” (CNBC live blog, November 2). The dilution arithmetic is ours: Lumentum’s earnings re-timed to Marvell’s fiscal year on diluted shares, before synergies. ↩
- AMD: the $16.14 billion and 56% lines were written in our coverage work and registered in our ledger before the print (FT-AMD-20261103-REUW, PG-AMD-20261103). Results: AMD Q3 2026 8-K Ex. 99.1, November 3, 2026: revenue $13,377 million, +45%; Data Center $8.5 billion, +97%; non-GAAP gross margin 56%, non-GAAP EPS $2.09; free cash flow $554 million against $1,558 million in Q2; total debt $7,079 million against $3,226 million at the end of Q2; Q4 outlook “approximately $17 billion, plus or minus 3%,” non-GAAP gross margin “approximately 55%”; Helios customers as named in the release. Late trading: about $517 at 4:29 p.m. ET against the $483.90 close (Yahoo Finance), a move of about 7%. Wednesday, November 4 closes (Yahoo Finance): AMD $488.78 against $483.90, +1.0%; the SMH semiconductor ETF 504.27 against 518.52, −2.7%. Chart 3 shows the close. ↩
- Midterms: “Prediction market traders see a nearly 75% likelihood of Democrats taking the Senate”; “The stock market is not yet pricing in a Democratic sweep of Congress”; “If Republicans keep control of the Senate or both chambers, the market may see short-term gains, JPMorgan said this week” (CNBC live blog, November 3). Going-in pricing and seat math: Kalshi, “Final 2026 midterm odds,” November 2, 2026 (prices as of noon ET: Democrats 74% for the Senate, 94% for the House; “Republicans currently hold 53 Senate seats, compared to 47 for Democrats and the two independents who caucus with them”; “Democrats need a net gain of four”; first polls close 6:00 p.m. ET). The result: NBC News live blog, “Democrats have won control of the House of Representatives, NBC News projects” (11:04 p.m. EST) and “Democrats have captured full control of Congress, NBC News projects” (11:23 p.m. EST); NBC News Senate results, updated 11:28 p.m. ET: Democrats 51, Republicans 39, “Flipped 6 seats (+6 net gain),” “10 uncalled races,” projected winners Turek (Iowa), Hamilton (Kansas), Cooper (North Carolina), Brown (Ohio), Andrews (South Carolina), Talarico (Texas); House results page, same update: 161 to 93 called, +13 net. Futures from Yahoo Finance at 11:20 p.m. ET: S&P 500 (ES) 7,504.75 against the 7,596.75 prior settlement (−1.2%), Nasdaq-100 (NQ) 27,712.75 against 28,305.75 (−2.1%), Dow (YM) −0.7%. CNBC live blog, updated November 3, 11:26 p.m. EST: “Stock futures fall on speculation Democrats to win big in midterms, slow data center buildout”; S&P 500 futures −1.3%, Nasdaq-100 futures −2.2%, Dow futures −410 points; the ten-year “dropping more than 4 basis points to 4.904%”; AMD “initially surged double digits” and was “last up about 3%”; the AP projection of Gina Hinojosa over Gov. Greg Abbott in Texas and her data-center moratorium pledge. Kalshi’s Senate-control contract traded at 99.7 cents at 11:30 p.m. ET and had not yet settled. Count refreshed from the NBC Senate and House results pages, “Updated: 9:25 PM ET” November 4: Senate 54 to 44, “Flipped 7 seats (+7 net gain)” (Maine, Jackson over Collins, added to the six above), “2 uncalled races” (Alaska, Mississippi); House 242 to 151, “Flipped 35 seats (+34 net gain),” 42 seats uncalled by subtraction from 435. Wednesday closes (Yahoo Finance): S&P 500 7,688.62 against 7,568.92 (+1.6%), Dow 53,356.12 against 51,786.72 (+3.0%), Nasdaq Composite 25,736.51 against 25,372.01 (+1.4%), SMH 504.27 against 518.52 (−2.7%), ten-year 4.89%. ↩
- Treasury: “Additional financing details relating to Treasury’s Quarterly Refunding will be released at 8:30 a.m. on Wednesday, November 4, 2026”; October–December privately held net marketable borrowing estimated at $618 billion assuming an $850 billion end-December cash balance (Treasury, November 2, 2026). Employment Situation for October 2026: Friday, November 6, 8:30 a.m. ET (BLS release schedule); September payrolls +29,000 as reported. Vistra: the $8.1 billion 2027 line is in our October 3 note. ISM manufacturing for October 55.5 against a 54.5 consensus, prices paid 75.2 (CNBC, November 2). ↩
- AOL and Time Warner, January 10, 2000: a stock-for-stock merger valued at about $165 billion at announcement, AOL holders to own about 55% of the combined company (CNN Money, January 10, 2000; AOL Time Warner Form 10-K405 for 2000, merger agreement dated January 10, 2000, completed January 11, 2001). The Nasdaq Composite’s closing high was March 10, 2000. ↩
Methodology: tests are graded against their written wording, not their intent, and the grades are logged in an append-only ledger before they appear here. Index and yield figures are daily closes from Yahoo Finance unless a dated report is cited. Where our arithmetic differs from a company’s own (the Marvell accretion), both are shown. Nothing here is a price target.
Nothing on this page is investment advice. We work in scenarios and trip-wires, not predictions. See disclaimer.
Disclosure: per the author’s brokerage positions page at 12:45 p.m. ET on November 4, 2026, the author holds long positions in AMD, Marvell, Micron and Nvidia, and in Alphabet (15 class C shares bought between October 29 and November 3; our November 3 Alphabet note disclosed no position at publication, and that disclosure is being reconciled against the fill date). The author holds no position in Intel, Microsoft, Amazon, Meta, Oracle, Lumentum, CoreWeave, Coherent, Arm or Vistra; the Vistra position disclosed in our October 3 note has been sold. Marvell traded through the author’s approved exit level during the session of the read; exits are judged on closing prices and none had executed at the time of the read. This note is research and opinion, published to all readers alike, and is not investment advice for any person. bpleon research received no compensation from any company named. See disclaimer.