Chip Stocks Fall With Yields as AI Credit Spreads Widen

• Macro: Treasuries rallied as AI credit jitters and wider French-German spreads drew buyers, and the 10-year yield fell 19.7 bps to 4.949% at 3:01 PM ET. Jobless claims at a three-month low of 194,000 failed to halt the rally.
• Market Moves: The S&P 500 fell 1.4%, the Nasdaq-100 3.1%, small caps (IWM) slipped 0.2%. Lower yields lifted homebuilders (ITB +5.2%) and staples (XLP +3.0%). AI chip and power funds sank on credit and supply worries (SMH -6.0%, AIPO -7.2%). Still, 53% of our universe rose.
• Factors: Style factors explained 61% of stock variance, versus a 23% one-year norm. Beta fell 3.89% (z=-4.1) as high-beta AI names sank. Treasury (Duration) rose 1.05% (z=+3.3) with bonds. Short Interest rose 0.52% (z=+3.1) as the most-shorted stocks outperformed, and our HF de-risking score is in the top 1% of sessions since 2020.
• Micro: Meta Platforms fell 8.1% after AI spending pushed free cash flow to a deficit and its credit default swaps to a record. Lam Research fell 11.5% on component-shortage shipment constraints. Procter & Gamble rose 6.0% on a core EPS beat.
• Watch: Do Western Digital and Bloom Energy, both of which slid on their Beta tilts today, give AI hardware and power names a company-specific anchor when they report after today's close?

October 22, 2026 15:11 ET · Updated 16:43 ET

De-Risking Score Reaches 99.4th Percentile as Beta Erases Its One-Year Gain

Today's factor pattern is consistent with a hedge-fund de-gross, pushing our HF de-risking score to the 99.4th percentile since 2020 so far today (10th highest of 1,671 sessions, compared against full-day history) and the 98.6th percentile over three days. Three of the score's four legs fired today: last week's relative moves continued instead of reverting, 12-month momentum winners fell 1.69%, and the most-shorted stocks rose 0.52%. The most-shorted stocks' gains came as Goldman Sachs Prime Brokerage data reported by Bloomberg on October 22 showing hedge funds sold shorts at a 5-to-1 ratio over long buys this month, driving short leverage to the 99th five-year percentile and net leverage near a three-year low as prime brokers including Goldman Sachs and Barclays encouraged clients to add shorts to conserve scarce balance-sheet financingBloomberg.

The Beta and 12-month momentum factors have now given back 26.5 percentage points since their shared June 22 peak, surpassing the 25.0-point slide from February 2021 to May 2022 as the deepest drawdown since 2020. Beta alone has borne the brunt of the selling: after climbing from -4.1% on March 30 to +18.1% on June 22 and standing at +3.3% at yesterday's close, today's 3.89% drop pushed its trailing-year cumulative return negative (-0.61%) for the first time since April 6. Beta's 10-day return of -10.1% is deeper than any since 2020 except March 2020 (-11.8%) and two earlier windows of this same slide. By contrast, 12-month momentum still holds a +6.42% cumulative gain over the trailing year.

The pressure converged on the high-beta momentum cohort, where the top 50 names dropped 10.9% (0.2nd percentile since 2020) and the broader ~190-name group lost 8.6% (sixth-worst day since 2020), while stocks outside both screens rose 1.2%. Because the cohort's exposures correlate negatively with Short-Term Momentum (-0.31) and positively with Beta (+0.43) and 12-month momentum (+0.45), today's +1.05% jump in 5-day trend (z=+3.5; 20-day +2.39%, z=+1.8) while yesterday's winners reverted (-0.28%, z=-1.0) reflects the same multi-week losers continuing to slide. Across our 2,112-stock universe, the median stock rose 0.26% even as the equal-weighted average fell 0.57% and the Nasdaq-100 sank 3.1% — the same split, in one session, that Michael Burry described on October 17 for the months after the March 2000 peak, Benzinga reported.

Chip Stocks Fall 6% as Bonds Rally, a Growth Link That Has Held All Year

Style factors accounted for 60.6% of cross-sectional variance against a 23.1% one-year average (100th percentile of the past year) as falling Treasury yields lifted traditional duration beneficiaries while AI infrastructure stocks tumbled. The 10-year Treasury yield fell 19.7 bps to 4.949%CNBC and the 7-year yield dropped 20.9 bps to 4.844%CNBC as investors sought safety amid AI credit jitters and a widening in French-German 10-year sovereign spreads to a 15-year highBloomberg. That bond rally lifted Treasury (Duration) 1.05% (z=+3.3), boosting long-term Treasuries (TLT +2.0%), homebuilders (ITB +5.2%), and regional banks (KRE +2.5%). Over the 10 sessions through October 21, however, the 10-year yield fell 13.1 bps while the VanEck Semiconductor ETF (SMH, -6.0% today) lost 10.9%, Falling yields and falling chip stocks both read as worries about growth, the link that has held over the past year; the prior 20 sessions, when chips rose as yields fell, were the exception.

Hyperscalers and semiconductor suppliers are trading on their own balance-sheet demands and physical bottlenecks rather than the risk-free discount rate. Meta Platforms dropped 8.1% (stock-specific -5.9%, erasing $143.9B of market value) as its five-year credit default swaps climbed as much as 8.9 bps to 119.3 bps — the highest since active trading began a year ago — after its $135B–$145B 2026 capital expenditure plan pushed third-quarter free cash flow to a $2.12B deficitBloomberg. Amazon fell 5.0% (stock-specific -2.9%) , having raised its 2026 spending forecast to about $220 billion in July. In hardware, Lam Research tumbled 11.5% (carried -8.2% by style factors and -2.2% stock-specific) after flagging $300M–$400M in December-quarter shipment constraints from component shortages alongside an expected sequential drop in China revenueSEC. Factor pressure pulled Nvidia down 4.5% (style -3.2%) and Advanced Micro Devices down 8.2% (style -9.6%), while Western Digital (-7.2%) and Bloom Energy (-11.0%) slid on their Beta tilts ahead of their after-close earnings reports.

Earnings reactions so far todayReporters worth over $100 billion, or moving more than 10% with a value above $5 billion

Company	Move	Stock-specific	Market cap	Revenue YoY	Margin YoY	What drove it
Carpenter Technology CRS	-20.1%	-17.0%	$18B	—	—	Q1 revenue missed estimates
Graco GGG	-15.8%	-16.4%	$13B	—	—	Q3 EPS and revenue miss; weak Q4 guide Financial Times
Nokia NOK	-14.9%	-6.8%	$52B	—	—	Lowered 2026 free cash flow conversion guidance
FirstCash FCFS	-14.2%	-12.3%	$9B	—	—	Fell despite a record Q3 beat
MSC Industrial Direct Co. Inc. Class A MSM	-13.7%	-12.0%	$7B	—	—	$42M impairment; Q1 margin guide below Q4
Lam Research LRCX	-11.5%	-2.2%	$326B	+54%	+5.6 pts operating	Supply bottlenecks and projected China revenue decline SEC
Meta Platforms META	-8.1%	-5.9%	$1.8T	+26%	-20.0 pts operating	Negative FCF, $135B+ AI capex, $9.9B legal reserve Bloomberg
Valero Energy VLO	-2.9% light volume	+0.8%	$126B	+43%	+14.8 pts adj. operating	—
Tesla TSLA	-2.7%	+0.1%	$1.5T	-0.2%	-3.2 pts operating	Operating income fell 56%; cut 2026 capex The Wall Street Journal
Texas Instruments TXN	-0.7%	+3.6%	$260B	+28%	+12.4 pts operating	—
IBM IBM	-0.4%	-2.5%	$237B	+5%	+0.7 pts adj. pre-tax	—
Union Pacific UNP	+2.9%	+1.8%	$171B	+15%	-0.3 pts adj. operating	—
SAP SAP	+3.3%	+2.3%	$269B	—	—	Q3 beat and raised full-year cloud outlook
Procter & Gamble PG	+6.0%	+3.9%	$364B	+4%	-0.7 pts adj. operating	Core EPS beat on 4% organic sales growth
Lockheed Martin LMT	+6.5%	+3.8%	$127B	+11%	-0.1 pts adj. operating	—
Spotify Technology SPOT	+10.4%	+8.8%	$108B	—	—	Subscriber growth and operating income beat guidance
Waste Connections WCN	+10.5%	+7.1%	$42B	+7%	+2.7 pts operating	—
Pool POOL	+10.9%	+7.1%	$6B	—	—	—

Stock-specific: the part of a stock's move not explained by the market, style or thematic (industry and theme) factors in the FactorPulse risk model.

Revenue and margin from the company's earnings release, year over year; margin change in percentage points; — where the release is not parsed or the figures failed checks.

Light volume: the company's earnings filing confirms the report, but trading volume has not caught up yet, so the move is an early read.

Of the 500 largest companies, 92 have reported (23% of market cap), on day 16 of the season.

How earnings reactions split, day 16 of the seasonThe 500 largest US companies with a confirmed earnings reaction so far (88), vs the same day in the past eight seasons and in four-season stretches back to mid-2020

Earnings reactions this season vs the past eight seasons, day 16

Season	≤ −5σ	−5 to −2σ	−2 to 0σ	0 to 2σ	2 to 5σ	≥ 5σ
This season	3%	23%	26%	27%	17%	4%
Past 8 seasons	14%	22%	15%	16%	18%	15%
Q2 2026	14%	26%	15%	14%	15%	16%
Q1 2026	15%	20%	15%	21%	15%	14%
Q4 2025	14%	26%	21%	12%	17%	11%
Q3 2025	6%	23%	14%	18%	16%	23%
Q2 2025	18%	26%	16%	14%	14%	13%
Q1 2025	19%	8%	21%	18%	30%	5%
Q4 2024	6%	20%	14%	15%	30%	15%
Q3 2024	17%	22%	10%	16%	15%	19%
Q3 '23–Q2 '24	13%	21%	17%	13%	21%	15%
Q3 '22–Q2 '23	11%	21%	18%	17%	22%	11%
Q3 '21–Q2 '22	16%	18%	12%	21%	25%	8%
Q3 '20–Q2 '21	6%	27%	21%	24%	17%	5%

σ: a stock's earnings-day stock-specific move divided by its typical daily stock-specific move over the past year. Each bar is cut at the same business day of the season; companies weighted by the square root of market cap.

Earnings reactions across the broader universe remain muted even as big-cap defensives find support. Weighted by the square root of market cap, 21% of third-quarter reporters among the 500 largest US companies have risen 2σ or more on a stock-specific basis so far, below the 33% average at this point across the past eight seasons, while 26% have fallen 2σ or more against a 35% eight-season norm. SpaceX (SPCX) dropped 8.7% (style -5.3%, stock-specific -3.3%) as investors weighed capital-structure risk after Tesla (-2.7%) trimmed its 2026 capex forecast to $20B and floated a potential combination with the space companyNYTThe Wall Street Journal. On the upside, Ryder System gained 8.3% (stock-specific +9.5%) after beating third-quarter earnings estimates and raising full-year guidance, and Blackstone rose 7.3% (stock-specific +6.2%) following its morning report.

ETF Divergence: Rate-Sensitive Funds Climb While AI Power and Gold Miners Retreat

Cross-asset ETFs split between domestic rate beneficiaries and AI-linked hardware, power, and commodity funds. The First Trust Smart Grid Infrastructure ETF (GRID -4.7%) traded at 4.2× its typical session pace as Beta dragged down grid holdings Eaton (-6.9%, style -4.5%) and Quanta Services (-9.1%, style -5.7%) despite Wood Mackenzie reporting the global data-center pipeline has passed 250 GW, according to Business Insider, and Carrier Global announcing a $600M Texas liquid-cooling plantThe Wall Street Journal. Bitcoin miners hosting AI data centers led the downside, sending WGMI down 8.8% (-34.3% month to date, its second-worst month since 2022) as Bitcoin fell 3.8% to $81,240.37 at 3:02 PM ETCNBC. In commodities, WTI crude rose 2.4% to $99.47/bbl at 2:51 PM ETCNBC and Brent climbed 3.3% to $112.40/bblEconomic Times after the U.S. State Department ordered non-emergency personnel out of Saudi Arabia following Houthi strikes, according to AP News, while energy stocks, among the year's biggest winners (XLE up 61% over the past year), fell with the rest of the year's leaders and the Oil factor fell 0.75% (z=-1.6) and Gold dropped 1.02% (z=-3.2) as spot gold slid 2.7% to $4,030.24/oz at 3:01 PM ETCNBC, with Treasuries rather than gold serving as the safe haven.

Reader Questions

How Could AI Cloud Financing Stress Spill Over to the Broader Market?

Financing stress at AI cloud providers reaches the broader market through private-credit lenders holding GPU-backed debt, high-yield data-center bonds, Nvidia's off-balance-sheet backstops, and hyperscaler credit spreads. CoreWeave's five-year credit default swaps hit an intraday record of 1,194.5 bps on October 15 — pricing roughly 60% five-year default odds — and its 2032 notes yielded about 14.9% on October 19, according to Bloomberg and CNBC. Private Equity Wire, citing Reuters, reported Thursday that CoreWeave's roughly $50B debt load was arranged largely by private-credit firms including Blackstone (+7.3% today; BXSL +2.4%), Magnetar, Carlyle, and BlackRock against GPUs and contracts with Microsoft (-2.1%), OpenAI, and Meta, while Carlyle estimated private credit may need to supply $1T for the AI buildout. In public debt, Forbes reported that neoclouds and data-center developers issued $37B of junk bonds in 2026 (with bonds backing CoreWeave-leased facilities yielding an average of 11%), while high-yield spreads overall have widened roughly 60 bps over the past 20 sessions judging by HYG (+0.1% today).

On the supplier and customer channels, Forbes reported Wednesday that Nvidia has backed $164.5B of AI data-center commitments, and Reuters reported Monday that S&P Global's Mark Habib warned hyperscalers likely carry more undisclosed AI liabilities through interconnected project backstops, though S&P sees no current rating changes. Secured bank lending remains open at steeper terms — Nebius closed a $3.5B, 364-day facility from Citigroup on Thursday at Term SOFR plus 2.75%, stepping up 25 bps every 90 days to 3.50%, according to its press release — and there is no reporting of contract renegotiations by Microsoft or OpenAI or of neocloud receivable write-downs at Dell Technologies (-7.4%) or Super Micro Computer, leaving the equity fallout concentrated in Beta (-3.89%) rather than Leverage (+0.12%, z=+0.7).

Economic Context

Initial jobless claims, released at 8:30 AM ET, fell by 6,000 to a three-month low of 194,000 for the week ended October 17, beating the 200,000 consensus estimate, while continuing claims for the week ended October 10 rose to a six-week high of 1.73 million Labor Department. Despite the firmer labor print, the US Dollar Index edged down 0.2% to 101.604 at 3:02 PM ETCNBC and Treasury yields fell across the curve ahead of next Wednesday's October 28 FOMC decision Federal Reserve.

Data compiled by FactorPulse AI; edited and verified by Jeff Klein. For informational purposes only. Does not constitute financial advice, an investment recommendation, or an offer to buy or sell any securities. Always consult a qualified financial professional before making investment decisions.

For more on factor construction methodology, see www.factorpulse.com/glossary.

← FactorPulse  ·  All Digests