The cut came from the supply chain

The most important sentence of earnings week came from Microsoft's finance chief, and the market heard something else entirely. CNBC reported that Amy Hood said on the earnings call that third-party providers of AI chips had delays that will push out some available capacity until the first half of 2027 (source), which is why Microsoft reduced its capex forecast for the year to $170 billion from $175 billion. Our call: the week's AI selloff priced a demand problem, the transcripts describe a delivery problem, and the spending Wall Street wrote off on Friday has just moved into next year's order book.

Alphabet said the same thing in its own way. CFO Anat Ashkenazi told analysts, per CNBC, that outside computing power "came online only at the very end of Q3." (source) The Motley Fool reported that Alphabet said its capex guidance for 2026 holds, $195 billion to $205 billion, but said it expects capex in 2027 to "increase significantly." (source) So two of the biggest buyers of AI hardware in the world spent less than analysts modeled because they couldn't get the gear installed fast enough. That's a backlog.

CNBC's headline said the AI build-out hits snag as Google and Microsoft put off raising capex forecasts (source), which is accurate as far as it goes. But a snag in delivery and a snag in appetite lead to opposite places. On October 21 we argued the binding constraint had moved to the supply side, and this week two CFOs said so on the record and put a dollar figure on it.

Friday's tape ignored the difference. Intel's guidance, which is a story about personal computers, set off a rout in the companies that sell accelerators and memory into data centers, and Yahoo Finance noted that October marked the worst month for the Nasdaq since March 2025 (source). The money went into duration instead: Yahoo also reported that energy shares gained and Treasury yields fell amid geopolitical tension and defensive demand (source).

Microsoft trimmed its spending plan because the chips showed up late. Late chips are next year's revenue for the same suppliers the market dumped on Friday.

Three stories the market sold as one

Start with the scoreboard. CappNotes reported that Intel's collapse handed the S&P 500 its worst day since June (source), with the index down 1.82% to 7,580.07. Trading Economics had the week at a 0.9% loss for the S&P 500, 2.4% for the Nasdaq and 1.2% for the Dow (source), and Yahoo Finance noted that the drop extended the index's losing streak to three consecutive weeks (source).

The more useful number sits underneath. On Friday the equal-weight RSP fell 0.79%, as the average stock fared far better than the index (source), while the Nasdaq-100 fell 3.46%, per CappNotes. The average stock had a bad afternoon. The index had a bad day because a few dozen of its biggest names did.

Intel is a PC story. The quarter itself was fine. CNBC, using LSEG estimates, had adjusted earnings of 52 cents vs. 40 cents expected, and revenue of $16.51 billion vs. $16.5 billion expected (source). The guidance did the damage. Per LSEG data, Intel guided fourth-quarter revenue to between $15 billion and $16 billion, against the $17.07 billion analysts expected (source). CFO David Zinsner added that "We see the PC slowdown continuing at least through the first half of next year." (source)

Look at where the miss came from. CNBC's Investing Club noted that sales outpaced expectations, driven by strength in the Data Center AI and Foundry segments, partially offset by a miss in the Client Computing and Physical AI Group, which has exposure to the PC market (source). The stock got no credit for that: CappNotes reported that INTC plunged 28.58% on fourth-quarter guidance of 25 cents in EPS versus 43 cents expected, plus a cash-burn warning (source), which TheStreet called its worst day since 1974 (source).

The market had already started on Intel's second problem. 24/7 Wall St. reported that on Thursday Elon Musk's Terafab chip initiative confirmed it's in active discussions with Taiwan Semiconductor Manufacturing about a dedicated fabrication plant (source), which undermines Intel's earlier positioning as a central foundry partner for the project. A money-losing foundry watching its marquee prospect walk to the competitor is a company-specific problem. It tells you very little about how many GPUs Microsoft wants.

The selling didn't make that distinction. Trading Economics tallied the damage: Nvidia fell 4.2%, AMD slumped 11%, Broadcom shed 5.5%, and Micron lost 5.3% (source). For the accelerator and memory names, this is a price disconnect, not a thesis break. Their biggest customers just complained on the record that suppliers can't deliver fast enough, and in our experience that's about the best complaint a vendor ever hears.

The hyperscalers are now graded on cash. Alphabet had the best cloud quarter in the group and still logged its worst day as a public company. Per the Motley Fool, Alphabet's cloud unit grew 103%, earning $11.1 billion in operating income on $30.7 billion in sales (source). But Alphabet spent $51.4 billion on capital expenditures, up from $24.0 billion a year ago. CNBC reported a free cash outflow of $9.07 billion for the third quarter, compared to the $2.7 billion analysts expected (source), and on Thursday the shares fell 13%, wiping out roughly $500 billion in market cap.

Microsoft spent heavily too, and the reward was a flat stock, which in this tape counts as a standing ovation. The Motley Fool had sales up 19%, reaching $92.3 billion (source) and Azure grew 48%, while CNBC noted that free cash flow fell 27% to $18.63 billion but remained positive (source). Meta reported a week earlier, and Vantage Markets read its release as showing capital expenditure of $34.91bn and free cash flow of negative $2.12bn (source). The rule the market wrote this week is simple. Spend what you like, as long as the business pays for it.

That rule changes how you read the delivery delay. If capacity slips into the first half of 2027 and Alphabet's 2027 budget rises "significantly" on top of it, the cash drain doesn't end. It just moves. Suppliers book the same revenue a few months later, while the buyers' wait for positive free cash flow gets longer. Friday traded as though it were the other way around.

It also explains why earnings season feels worse than it is. CNBC, citing FactSet, said the S&P 500 is still projected to post earnings growth of about 30% for the third quarter (source), but companies on average are only beating earnings expectations by 1.4%, far lower than both the five- and 10-year averages of 7.5% and 8%. FactSet's John Butters puts that drop almost entirely on Meta and Alphabet. Two cash flow statements are setting the mood for 500 companies.

Amazon closed out the week with what CappNotes described as a miss on revenue and profit with negative free cash flow of $42.4 billion over 12 months (source). The stock fell 1.41% on Friday.

The bond market bet against the Fed. Fox Business reported that Fed policymakers voted 10-2 to leave the federal funds rate at its target range of 3.75% to 4% (source) on Wednesday. The Fed's statement named the dissenters as Lisa D. Cook and Beth M. Hammack, who preferred to raise the target range for the federal funds rate by 1/4 percentage point (source). Thursday gave the hawks more to work with: CNBC reported third-quarter GDP increased at an annual rate of 3.4%, better than the 3.1% expected (source), and Trading Economics had headline PCE inflation accelerating to 3.6% from 3.4%, while core inflation picked up to 3.1% from 3% (source). Per CME FedWatch, the probability of a 25-basis-point hike in December rose to 82.6% (source).

Treasuries rallied anyway. CNBC reported that on Thursday the 10-year Treasury yield was more than 8 basis points lower at 4.964% (source), and the 2-year lost more than 10 basis points to trade at 4.629%. When bonds rally into a hotter inflation print and a likely hike, the market is betting that growth cracks before prices do. The income data gave it some support: Investing.com flagged that US real disposable income falls 0.3% as PCE inflation hits 3.6% (source). On October 24 we argued the AI buildout runs more and more on borrowed money, so a 10-year back under 5% was the week's one piece of good news for it, and almost nobody noticed.

Apple, AMD, a ballot and a payroll print

Next week packs more catalysts into five sessions than October managed in four weeks. CNBC counts 138 S&P 500 companies set to report (source), and the macro calendar finishes with the October jobs report.

AMD on Tuesday is the test of our read. The stock lost 11% on Friday because of another company's guidance. Watch the data center outlook and, more than that, the wording. If management talks about customers waiting on supply, packaging or power, the delivery read holds. If it talks about customers pushing out orders or "digesting" capacity, that's demand, and we'd be wrong.

Apple goes Monday. CMC Markets says the company is expected to report fiscal fourth-quarter 2026 earnings growth of 7.4% year on year to $1.99 a share, with revenue forecast to rise 10.7% to $113.4bn (source), and CNBC argues that poor results from the hyperscalers raise the stakes for Apple (source). It's the one megacap with no data center bill to defend, and that's carrying more weight than it used to.

On macro, CMC Markets points out that the ISM prices-paid gauges have been rising and are near recent highs, suggesting inflation remains elevated (source). For Friday's payrolls, Newsquawk's consensus looks for the US economy to add 55k nonfarm payrolls in October (vs 29k in September), with the unemployment rate expected to remain unchanged at 4.2% (source). Going into Tuesday's vote, Trading Economics notes polls suggesting Republicans could lose their majority in the House and potentially control of the Senate (source). Two levels matter. On the S&P it's Friday's intraday low of 7,527.67 (source), and on rates it's whether the 10-year stays below 5%.

Where the evidence leaves each theme

AI semiconductors: conviction intact. Hood's comment is the clearest demand signal of the quarter, even though it came in the form of a capex cut. Capacity that slides from late 2026 into the first half of 2027 lands in supplier revenue a few months later, and nothing this week suggested any buyer wants less of it. Nvidia, Broadcom and Micron fell on an Intel PC warning. Our view changes if AMD or Qualcomm describes customers deferring orders, or if a hyperscaler cuts its 2027 plan instead of reshuffling 2026.

Hyperscalers: cash flow decides it now. Microsoft passes the test and has earned its premium. Alphabet's cloud economics are the best in the group, roughly 36% operating margins on that $30.7 billion, but the stock now trades on a cash flow line that its own 2027 guidance says won't turn soon. Until profit growth from capex outruns the growth in capex itself, we don't see the catalyst that gets Alphabet re-rated.

Intel: a separate problem with a long runway. Management has said the PC slump runs through the middle of 2027, and the foundry still hasn't landed the outside anchor customer it needs (Terafab's TSMC talks went the other way). The AI read-through from Intel is close to zero, and its own recovery is a 2027 story with a capped upside until then.

Rates and the Fed: the hike is priced, the growth scare isn't. CNBC carried Warsh's line that "I don't believe in forward guidance absent a crisis" (source), so December depends on the data, and with hike odds at 82.6% there isn't much room left for a hawkish surprise. The more telling signal is the bond rally into a 3.6% PCE print. A hot payroll number with firm ISM prices would test it. A soft one would confirm that the long end is already easing the buildout's cost of capital while the Fed tightens around it.

This document is for informational purposes only and does not constitute investment advice. Nova One Capital makes no representation as to the accuracy or completeness of the information herein, and past performance is not indicative of future results. Nova One Capital and its affiliates may hold positions in securities discussed. Disclosure: Nova One Capital may hold positions in companies we discuss. See novaonesignal.com/disclosures. October 31, 2026.