Starr Quant Lab Desk Research

US · Recap

US Market Recap | 2026-07-29 (ET), Wednesday

Wed US Recap · 11 tables America/New_York

Machine-translated from the Chinese original. In case of any discrepancy, the Chinese version prevails.

Single-name entries

This is a reconciliation report — it carries no single-name score table. Recaps are structured around hit-rate reconciliation, theme verification and next-day outlook.

Coverage window: 2026-07-29 09:30–16:00 ET regular session + 16:00–18:05 ET after-hours Data timestamp: 2026-07-29, approx. 18:00–18:05 ET (about 2 hours into after-hours; the MSFT/META/QCOM/ARM/LRCX earnings calls were mostly in progress or already finished) Reconciliation baseline: reports/us/2026-07-29.md (this morning's pre-market list, including QC-revised items R1–R15) ⚠️ Note on the data path (must be recorded): All Yahoo Finance endpoints (query1/query2, the yfinance wrapper layer) returned 429 Too Many Requests for the entire day, so this report was unable to use yfinance. Quotes were instead sourced from the stockanalysis.com quotes API (single stocks/ETFs, close + after-hours) and the CNBC quote endpoint (indices/VIX/Treasuries/crude futures), and earnings figures were cross-checked against company press releases and media reporting. This is a different data path from the established process, and the data has not been re-verified through yfinance — please discount every price-related number in this report accordingly.

0. One-line recap

  1. Today was unambiguously risk-off, but its nature was a concentrated blow-up of "rate shock + AI-chain de-rating," not a broad panic. The Dow fell −2.19% (−1,153.18 points), its worst single day since April 2025, VIX rose +13.45% to 20.66 (the "breaks above 20" threshold set in the pre-market list was breached), and 11 sectors closed 9 down / 2 up. But the equal-weight S&P (RSP) fell only −0.90%, clearly better than the cap-weighted SPY at −1.54% — damage to the median stock was far smaller than the index level, what was killed was AI weightings and semiconductors, not all stocks (SOXL −16.02%, SMH −4.79% vs IGV +0.64%).

  2. The only direction that genuinely made money today was energy — and that is precisely the one theme the pre-market list actively downgraded. Brent September contract closed at $90.92, +8.12% (not the pre-market framing of "overnight +3.5% to $87"). XLE +1.88% was the only sector up meaningfully, with MGY +6.97%, COP +3.47%, XOM +2.42%, CVX +2.28%, USO +7.32%, BNO +7.89%. Pre-market, on the grounds that "all attacks were intercepted, no capacity loss," the theme's logical strength was cut from "medium-high" to "medium" and its persistence to "medium-low" — that downgrade was directionally wrong, and today's strongest long was right here.

  3. The FOMC was a "hawkish hold": the 3.50–3.75% range was maintained, but three regional Fed presidents (Hammack/Kashkari/Logan) voted for an immediate hike — the most dissents since September 2016. Warsh called it "a good family fight," and stressed that standing pat is not inertia and that the anti-inflation commitment is unchanged. The bond market delivered a bear steepening: 2Y −4bp to 4.236%, 10Y +5bp to 4.657%, 30Y +9bp to 5.193% (independently corroborated by this report's own ETF data: TLT −1.65%, far weaker than IEF −0.42%). Pricing for a September hike was, paradoxically, marked down by this meeting to roughly 60% on swaps.

  4. Pre-market list reconciliation: the long-leaning side delivered 2/6 (33%), the avoid side delivered 16/19 (84%), for a combined 18/25 ≈ 72%. The two most valuable entries in this report are not hits — they are two cases of "the rule won, the judgment lost": ① the avoidance rule "never chase a gapped-up stock" was validated today in extreme form — TER opened +14.74% ($367.90) and closed −0.39% ($319.41, i.e. the low of the day), so buying at the open meant −13.16%; ② BE's verification point ($188.18) missed by $2.52 (intraday high $185.66), so under the pre-set pre-market rule it was "not treated as a clean positive" — and it closed −1.85%.

  5. The after-hours script reversed: today's intraday bear thesis was partly overturned after the close by its own core counterparty. MSFT Azure growth came in at +43%, far above its own 39–40% guidance, and it traded +4.98% to $409.99 after hours; LRCX beat on both lines, +4.91% after hours; AMAT +1.96% after hours. META, on the other hand, posted a large EPS miss driven by a $2.4 billion legal accrual + $1.2 billion in severance costs, and fell −8.98% after hours — meaning the after-hours verdict on the "AI capex trust crisis" is split: cloud and equipment gave a positive answer, advertising gave a negative one, and META's problem was mainly one-time charges rather than capex.

  6. Tone for the next day: cautious, but do not chase shorts. VIX above 20, the 10Y closing in on 4.70%, and Brent above $90 stack up into a valuation-suppressing combination; but the positive after-hours answers from MSFT/LRCX are real fuel for a semiconductor bounce. On 7/30, AAPL + AMZN report the same evening after the close, and pre-market brings the Q2 GDP advance estimate + June core PCE + initial jobless claims — an event density no lower than today's.

1. Market overview

Index Close Change Change % Prior close
Dow Jones Industrial 51,594.14 −1,153.18 −2.19% 52,747.32
S&P 500 7,316.15 −112.63 −1.52% 7,428.78
Nasdaq Composite 24,442.94 −433.97 −1.74% 24,876.91
Russell 2000 2,906.31 −47.49 −1.61% 2,953.80
Equal-weight S&P (RSP) 215.73 — −0.90% 217.69
Mag7 (MAGS) 62.56 — −1.39% 63.44

The Dow fell more than the Nasdaq, which is counterintuitive within an "AI de-rating" narrative. The reason is that industrials and financials were hit twice over — by rates and by risk appetite: XLI −3.19%, JPM −3.53%, HON −2.40%. Today was not simply "sell tech," it was "sell duration + sell high valuation + buy crude."

Sentiment and cross-asset

Metric Close Change Read
VIX 20.66 +2.45 (+13.45%) Prior close 18.21. The pre-market list pre-set "breaks above 20 = shifts from non-event to risk event" — the threshold was breached
VIXY 22.53 +6.07% Confirms in the same direction as VIX
US 10Y 4.657% +5bp ⚠️ CNBC's 17:05 after-hours quote separately showed 4.687%, a post-close drift; see §7 open item 1
US 2Y 4.236% −4bp Short end down + long end up = bear steepening
US 30Y 5.193% +9bp The long end took the most damage
TLT (20Y+) 82.85 −1.65% This report's own data, independently corroborating long-end underperformance
IEF (7–10Y) 93.17 −0.42% The gap versus TLT = direct evidence of steepening
Dollar index DXY 100.80 −0.61% Prior close 101.417. Rates up while the dollar falls — that is an "inflation/stagflation premium" combination, not a "growth premium" one
Brent September $90.92 +$6.83 (+8.12%) Prior close $84.09. ✅ Also closes pre-market open item 6 (see §7)
GLD (gold) 371.08 +0.46% Mild haven demand

The most informative cross-asset combination: rates↑ + dollar↓ + crude↑ + gold↑. All four holding at once points not to "a strong economy" but to a rising inflation premium. Three Fed presidents demanding an immediate hike, Warsh stressing the anti-inflation commitment, and Brent +8% in a single day are two sides of the same thing.

Market breadth

Metric Reading Note
11 SPDR sectors 9 down / 2 up Only XLE +1.88% and XLP +0.34% rose
RSP vs SPY −0.90% vs −1.54%, equal-weight outperformed by 0.64pp Median-stock damage smaller than the index → a concentrated decline, not a breadth collapse
NYSE advance/decline ❌ Not obtained for the fourth day running See §7 open item 2. This report does not use any substitute data as a stand-in for that metric; the two rows above are independent breadth proxies and are not equivalent to A/D

2. Pre-market list reconciliation

Judgment basis: "watch closely" (long-leaning) counts as delivered if the stock closed higher or clearly outperformed the −1.52% market; "avoid / watch only" counts as delivered if it fell or underperformed the market. ⚠️ After-hours moves are flagged separately and are not mixed into the same-day delivery judgment (iron rule 3).

2.1 §9① Today's top 5 to watch + theme anchors

Ticker Pre-market conclusion Today's change % Open/High/Low Delivered? Comment
TER Watch closely, explicitly "do not chase"; verification point = gap-and-go or gap-fill −0.39% $319.41 Open 367.90 (+14.74%) / High 372.23 / Low 319.19 ✗ Direction not delivered / ✅ Avoidance rule fully delivered Verification point answer: a complete gap-fill. The open was the high zone of the day and the close was the low of the day ($319.19 ≈ the closing price). Buying at the open: −13.16%. After hours −0.85%. Listing "chasing TER" as avoidance item 2 pre-market was the single most valuable judgment of the day
BE Watch closely; verification point = can it reclaim $188.18 −1.85% $163.75 Open 183.50 / High 185.66 / Low 157.33 ✗ Not delivered (but the pre-set rule blocked it) The verification point failed by $2.52 (1.34%) — the intraday high of 185.66 never touched 188.18, after which it was cut all the way down to 157.33. An S-grade beat-and-raise closing lower on the day proves the four pre-market downgrades ("−52.5% from the high, +37.5% dilution, net debt, backlog questioned by short sellers") were necessary. Volume 44.04 million shares (elevated). After hours −1.23%
MSFT Event anchor, watch only ahead of the print; verification point = FY27 capex guidance + cloud gross margin −0.71% $390.54 Open 393.40 / High 401.25 / Low 388.74 Intraday ✅ (outperformed the market by 0.81pp) / After hours ✗ After hours +4.98% $409.99. "Watch only ahead of the print" was correct as process (avoiding a binary event + IV crush), but as an outcome it gave up a +5% after-hours move — recorded as such. See §4
F Watch closely; verification point = hold the after-hours $15.71 +2.14% $15.28 Open 15.57 / High 16.29 (+8.9%) / Low 15.16 ✅ Delivered (the only one of the five to close higher) The cleanest single delivery on the long-leaning side today. ⚠️ But the verification point was strictly not met: the close of $15.28 < $15.71, and it too formed a "gap up, run up, fade" pattern (giving back 6.2% from the high). Volume 98.23 million shares
ISRG Watch closely (the only name with dual tailwinds: stock-specific positive + sector at an all-time high) −2.40% $353.10 Open 361.46 / High 361.80 / Low 351.94 ✗ Not delivered, and underperformed its sector XLV −0.61%, so ISRG underperformed its sector by 1.79pp — the risk flagged pre-market, "rating-driven moves usually persist 1–2 days," played out the same day. And the verification point "can XLV's all-time high extend past the FOMC" was answered: no. A dual-tailwind name falls harder when both tailwinds disappear
XLE / MGY (off-rule theme anchor) Watch closely; verification point = can XLE turn positive XLE +1.88% / MGY +6.97% XLE open 59.08 / High 59.30 ✅✅ Strong delivery Verification point answer: yes, and it was the only sector up meaningfully. But the pre-market list cut both the theme's logical strength and its persistence, badly underestimating it — actual Brent +8.12% (pre-market framing +3.5%). ⚠️ Internal divergence must be recorded: E&P strong (COP +3.47%), oil services actually weak (OIH −0.92%, SLB −2.04%)

Long-leaning / watch-closely hit rate: 2 / 6 = 33%.

2.2 Avoid / watch-only list reconciliation

Ticker Pre-market conclusion Today's change % After hours Delivered? Comment
VRT Watch only −17.26% $223.04 −1.36% ✅✅ Strong delivery The single biggest correct avoidance among today's main names. ⚠️ But the reason does not match the pre-market explanatory framework, see §3 theme 3 — VRT raised both full-year revenue and EPS guidance, both above consensus, and was still killed 17%
META Watch only (ahead of the print) −1.31% −8.98% $533.00 ✅✅ Strong delivery Outperformed the market intraday; after hours −8.98% delivered on the "binary event + IV crush" judgment. See §4
FVRR Avoid / short watch −20.53% $9.21 −1.30% ✅✅ Strong delivery Made a 52-week low ($8.75). ⚠️ But the item flagged pre-market as open — "did it also maintain/raise EBITDA guidance" — remains unverified (§7 open item 4) — delivering is not the same as having had sufficient evidence at the time
KLAC Avoid (cause identified and valid) −10.80% $170.19 −0.42% ✅✅ Strong delivery Down a third consecutive day, about −23% over three days. Revision R1 (changing "cause unknown" to "cause known") was supported by the outcome
ARM Avoid (ahead of the print) −8.11% $224.89 −7.11% $208.90 ✅✅ Double delivery Down 8.11% intraday, then another 7.11% after hours, roughly −15% across both legs. The pre-market self-instruction to be "especially restrained" on this name was correct. See §4
QCOM Avoid (ahead of the print) −4.42% $155.68 −6.86% $145.00 ✅✅ Double delivery Same as above. See §4
NXPI Watch only −7.00% $240.98 −0.04% ✅ Delivered Auto semis were killed alongside SMH −4.79%
MU / SNDK / WDC Watch only (pre-market softened the bearish stance) −9.94% / −7.32% / −0.32% −1.97% / −2.15% / −1.25% ✅ Delivered (but the softening was wrong) ⚠️ Pre-market softened the call because "SK 海力士 (SK Hynix), on review, was a mixed rather than a one-sided negative" — and MU fell 9.94%. The softening was wrong on the outcome. ⚠️ But see §6 — tonight's QCOM earnings delivered a piece of evidence pointing the exact opposite way
ASML Avoid −2.04% $1,550.69 −0.24% ✅ Delivered (mildly) The decline was smaller than SMH −4.79%, so the counter-argument recorded pre-market (order book full through 2027) got relative-strength support
JETS / DAL / UAL Watch only (softened down from avoid) −2.79% / −3.49% / −3.51% — ✅ Delivered, but the softening was wrong Brent +8.12% hit airlines directly. Pre-market softened "avoid / short watch" to "watch only" on the reasoning that "a single-day +3.5% does not constitute a fundamental reversal" — crude actually rose 8.12%, so the very number underpinning that softening was itself an underestimate
HON Watch only −2.40% $241.12 +0.04% ✅ Delivered The judgment that "an upgrade to neutral is a weak signal" holds
NVDA Watch only −3.55% $190.01 −0.02% ✅ Delivered Closed at the low of the day, $190.01. Its status as the sole gainer of the prior day's "only one of 20 semis up" lapsed the very next session
RCL Watch only (disputed item, corrected in R10) +0.33% $323.58 −0.41% ✅ Delivered, R10 correction validated Closed higher in a −1.52% market and in the worst possible fuel environment of Brent +8.12%, and clearly outperformed CCL −1.49% / NCLH −2.21%. R10's "do not classify a company that just raised guidance as a loser" was correct
CSGP Watch only (raised from "avoid" by R3) −1.65% $29.83 +0.54% ✅✅ R3 correction strongly validated Opened at $26.03 (−14.2%), low of $25.89, closed at $29.83 — a +15.2% bounce off the low, recovering almost the entire decline. Verification point "will the decline converge" answered: it converged, and the market did start pricing "protect the profit guidance." This is the most successful reversal call of this QC revision round
CAPR Avoid (pure binary event) +1.52% $6.67 — ✅ Process delivered Volume only 27,500 shares; the decision to avoid a binary event needs no outcome validation
LRCX Avoid (binary event ahead of the print) −6.40% $252.35 +4.91% $264.74 🟡 Intraday delivered / after hours inverted The intraday drop of 6.40% delivered, but the print was a double beat and it rose +4.91% after hours. "Avoid a binary event ahead of the print" protected the intraday session but missed the after-hours move. Same type as MSFT
STX Watch only (demoted to #12 because of a source blacklist) +2.29% $764.43 −0.58% ✗ Not delivered Opened 786.85, high 814.97 (+9.1%), then faded, but closed up 2.29% for the day, an extremely strong relative showing in an SMH −4.79% environment. Source discipline (the ts2.tech blacklist) was methodologically right, but the cost was missing a stock that rose against its sector — recording that cost as it is
Micro-cap movers group Avoid DFNS +109.25%, AMIX +66.18%, STFS +31.68%, VIVK +23.71%, GMM +18.62%; STKH −24.72%, CSAI −22.67% DFNS another +59.80% ✅ Process delivered ⚠️ Recorded as it is: they were the largest gainers in the entire market today, with DFNS up +109% on the day and another +59.8% after hours. "Avoid" was because they are unanalyzable and non-reproducible, not because they were predicted to fall — that distinction must be written down clearly

Avoid/watch-only hit rate: 16 / 19 = 84% (✗: MSFT after hours, STX; 🟡: LRCX counted as a half).

2.3 Hit rate and self-critique

Combined: 18 / 25 ≈ 72%. Long-leaning side 33%, avoid side 84%.

One-line self-critique: today's score came almost entirely from "what was not done," not from "what was done right." Of the 6 long-leaning names, only 2 delivered, and even the one clean delivery, F, failed to meet its own verification point; while the 84% hit rate on the avoid side is essentially "being bearish is easy" on a day that was −1.52% with 9 of 11 sectors down. The three places that genuinely expose capability are: ① the energy theme was self-downgraded (direction right, magnitude judgment wrong, and it was the only money-making direction of the day); ② the hawkish-beneficiary logic for XLF was falsified; ③ the AI power theme was called backwards overall. None of these three are luck problems, they are framework problems — see §3 and §6⑤.

3. Today's theme verification

# Theme Pre-market strength/direction Actual today Leaders/laggards Stage Conclusion
1 FOMC (energy inflation + employment) S / two-way (the draft's "bearish" was changed to "two-way" by R2) Hawkish hold: 3.50–3.75% maintained, 3 votes for an immediate hike (the most since September 2016). Indices fell across the board, bear steepening Beneficiaries: none; losers: duration and high valuation Already detonated, single day ⚠️ The event was called right, the direction wrong — and it was led astray by a "correct process." R2 withdrew the stale pricing data (correct process) but inferred "two-way" from "hike pricing retreating from 46.5% to about 30% ahead of the meeting." In reality the Fed was more hawkish than the retreating pricing. The withdrawn draft conclusion of "bearish" was actually closer to the outcome. Lesson in §6⑤ item 1
2 AI capex trust crisis A+ / bearish Fully delivered intraday: SMH −4.79%, SOXL −16.02%, KLAC −10.80%, AMAT −8.40%, MU −9.94%, GEV −4.57%, PWR −4.63%, TSM −4.50%, AMD −5.51% Laggards: the entire semiconductor chain Intraday climax, partly rebutted after hours ✅ The strongest bear thesis of the day, called right. ⚠️ But its "direct counterparty" gave a positive answer after hours: MSFT Azure +43% (far above its own 39–40% guidance), LRCX a double beat and +4.91% after hours, AMAT +1.96% after hours. This theme is no longer one-sided after hours
3 AI power / data-center electricity A+ / bullish (internal divergence) ❌ Called backwards overall: BE −1.85%, VRT −17.26%, GEV −4.57%, PWR −4.63%, XLU −1.34%. Not a single name in the theme rose Laggards: all of them Ebbing ❌ The biggest theme misjudgment of the day. Pre-market it was listed as the third-strongest theme with a "bullish" direction. More importantly, the explanatory framework was overturned too: pre-market it claimed "the market rewards visibility on new orders (BE) and punishes margin delivery on existing orders (VRT)" — in reality VRT raised both revenue and EPS full-year guidance, both above consensus, and was still killed 17.26%; BE had an S-grade beat-and-raise and still closed lower. Today the market drew no distinction between "orders" and "margins" for AI power — it de-rated all of it indiscriminately. Mitigating factor: the stock-level labels were "watch closely / watch only," not buy
4 Energy and geopolitics A / bullish, but logical strength and persistence were both cut ✅✅ Strongest delivery and badly underestimated: Brent +8.12% $90.92, XLE +1.88% (the only sector up meaningfully), MGY +6.97%, COP +3.47%, XOM +2.42%, CVX +2.28%, USO +7.32%, BNO +7.89% Leaders: E&P; laggards: oil services OIH −0.92%, SLB −2.04% Main advance, persistence medium ✅ Direction right, magnitude judgment wrong. Pre-market, on the grounds of "all attacks intercepted, no capacity loss, diplomatic channels active," strength was cut to "medium" and persistence to "medium-low" — that reasoning underestimated the fact that "no capacity loss can still push up the risk premium." ⚠️ Newly discovered internal structure: E&P up, oil services down = the market is buying "the oil price," not "the capex cycle," and this was entirely absent pre-market
5 Back-end test ≠ front-end inspection (hypothesis under test) B+ / to be verified, only TER left as a single data point 🟡 Failed to establish itself as a trading basis. TER −0.39% did outperform SMH −4.79% by 4.4pp (relative-strength support) but gave back the entire +14.74% gap; meanwhile the "front-end" names LRCX (+4.91%) and AMAT (+1.96%) were up after hours — Hypothesis not passed ❌ The axis does not hold. If "front-end capex exposure is impaired" were true, LRCX's and AMAT's after-hours moves should not have been positive. Recommendation: stop using this dichotomy, see §6⑤ item 3
6 Momentum breakdown / value reallocation A / bullish (value side) ❌ Called wrong: XLV −0.61% (lost its all-time high), XLF −1.60%, JPM −3.53%, XLI −3.19%. Only XLP +0.34% Laggards: financials, industrials, healthcare Interrupted ❌ Called wrong, and this is the most diagnostically valuable case. Pre-market explicitly wrote "in a hawkish scenario XLF is the only beneficiary sector." The Fed was indeed hawkish, yet XLF fell 1.60% and JPM fell 3.53%. Reason: a bear steepening (2Y −4bp / 30Y +9bp) erodes duration and credit rather than helping net interest margin; and in risk-off, the beta character of financials overwhelms its rate character. The "hawkish → long financials" mapping was falsified this round
7 Victims of AI disruption (hypothesis under test) B / bearish FVRR −20.53%, a 52-week low Laggards: FVRR Continuing ✅ Delivered, but still a single sample, and the open item it listed for itself pre-market (the EBITDA guidance basis) was never verified. Do not upgrade a hypothesis to a conclusion just because it fell
8 Beneficiaries of falling oil C / already cut to "watch" ✅ Correctly zeroed out: crude went the other way at +8.12%, JETS −2.79% — Extinct ✅ The downgrade action was correct

Unexpected themes (completely missed pre-market)

  1. 🔴 The extreme divergence between software and semiconductors — the biggest omission of the day. IGV (software) +0.64% vs SMH (semiconductors) −4.79%, a gap of 5.43pp. The pre-market framework had it exactly backwards: "capex above expectations → positive for semis, negative for software margins." What actually happened today was software up, semiconductors collapsing. Treating XLK as a single bucket pre-market (XLK −2.64%) completely masked this 5.4pp fissure inside it.
  2. Silent bid for consumer defensives: XLP +0.34% was the second sector to rise after XLE, and the pre-market sector table gave it no place at all.
  3. After-hours surprise in restaurant consumption: SBUX +6.10% after hours (EPS $0.85 vs $0.66 expected), CMG +5.99% after hours. Pre-market §1 #4 listed "SBUX / CMG" only as a footnote and never analyzed them.
  4. Micro-cap speculation going to extremes: DFNS +109.25% on the day, then another +59.80% after hours. It was correctly avoided, but the label "noise" underestimated its information value as a risk-appetite indicator — this kind of speculation on a day when VIX was +13% says speculative capital has not withdrawn, it has only moved the battlefield.

4. After-hours earnings moves (direct catalysts for the next day)

⚠️ Important observation: after-hours prices drifted substantially during the earnings calls, and the first printed reaction is a very poor guide. The table below shows two timestamps side by side: A = immediately after the release (investinglive aggregation basis), B = 18:00–18:03 ET, this report's data pull.

Ticker Result vs expectations Guidance After hours A (immediate) After hours B (18:00 ET) Read
MSFT Revenue $90 billion (+18%, +17% at constant currency) vs $87.62 billion expected (beat by 2.7%); non-GAAP EPS $4.74 vs $4.24 (beat by 11.8%); GAAP EPS $4.81 (+32%); operating income $40.6 billion (+18%); Microsoft Cloud $59.3 billion (+27%); Intelligent Cloud $39.3 billion (+32%); Azure +43% y/y ⚠️ FY27 capex guidance was not in the press release (the company said it would be given on the call). Q4 actual capex $35.8 billion; FY26 full-year capex $115.9 billion +1% +4.98% $409.99 ✅ The core pre-market question, "can Azure hold up," got its strongest possible answer: +43%, far above its own 39–40% guidance. This is today's most powerful counter-evidence to the "AI capex trust crisis." ⚠️ The other verification point set pre-market (FY27 capex vs a market expectation of roughly $220 billion) could not be confirmed by this report — the widely circulated "$255–260 billion" comes from a preview article published before the print, not from disclosed guidance; see §7 open item 3
META Revenue $60.80 billion (+28%) vs $60.18–60.22 billion expected (beat by about 1%); EPS $6.18 vs $7.10–7.19 expected (miss of about 14%, −13% y/y) Q3 revenue guidance $61.0–64.0 billion (midpoint below consensus); 2026 capex range narrowed to $130–145 billion (from $125–145 billion, lower bound raised by $5 billion) −5% −8.98% $533.00 The cause of the EPS miss was one-time: a $2.4 billion legal contingency accrual + $1.2 billion of severance costs — excluding them it should have beaten. But the market simultaneously saw "capex floor lifted + Q3 revenue guidance midpoint on the low side," i.e. rising cost rigidity with a flattening revenue slope. ⚠️ This is directionally consistent with the pre-market judgment that "META's capex controversy is sharper than MSFT's," and it was confirmed today
QCOM Revenue $9.95 billion vs $9.68 billion expected (beat by 2.8%, near the top of guidance); non-GAAP EPS $2.21 (−20% y/y), broadly in line; net income $2.0 billion (−25%); QCT pre-tax margin contracted 4pp to 26% FQ4 revenue $9.7–10.5 billion, adjusted EPS $2.05–2.25 — the midpoint implies almost no improvement in earnings −10.4% −6.86% $145.00 ✅ The pre-market diagnosis of "gross margin down four quarters in a row, operating leverage running in reverse" was confirmed, and the company gave the reason: wafer, memory, advanced packaging, and test-and-assembly costs are all rising; the company is raising prices but pass-through takes time. ⚠️ This attribution creates an important cross-name contradiction, see §6 focus item ①
ARM Revenue $1.289 billion (+22%) vs $1.27 billion expected (beat by about 1%); adjusted EPS $0.45 vs $0.40 (beat by 13%, +29% y/y); licensing revenue +29% to $819 million; royalty revenue only +11% to $671 million Next-quarter revenue $1.26 billion ± $50 million (about +20%), adjusted EPS $0.40 ± $0.04 — below the quarter just reported, a sequential decline; said FY27–FY28 customer demand exceeds $2 billion (twice last quarter's $1 billion) −9.3% −7.11% $208.90 A double beat that still fell hard, the same type as KLAC. Two crux points: ① guidance declines sequentially (the same type as TER yesterday); ② royalties only +11% — for a company priced on an AI narrative (pre-market recorded a current-fiscal-year P/E of 112.8x), decelerating royalties matter far more than an EPS beat. −8.11% on the day + −7.11% after hours, roughly −15% across both legs
LRCX Revenue $6.722 billion vs $6.665 billion expected; EPS $1.82 vs $1.68 (beat by 8.3%) Not obtained +2.5% +4.91% $264.74 ✅ A double beat. Listing it pre-market as "avoid (binary event ahead of the print)" protected against the intraday −6.40% but missed the after-hours move. This is direct counter-evidence to the "front-end equipment = capex loser" dichotomy
SBUX Revenue $9.3 billion vs $9.17 billion expected; EPS $0.85 vs $0.66 (beat by 28.8%) Not obtained +7.6% +6.10% $110.49 A name never analyzed pre-market, yet the second-biggest after-hours winner. Consumer recovery is a thread that needs to be added tomorrow
CMG Figures not obtained Not obtained — +5.99% $36.29 Same direction as SBUX, an after-hours resonance across restaurants. ⚠️ This report did not obtain the specific financial figures
AMAT No earnings (dragged along by LRCX) — — +1.96% $445.02 −8.40% on the day, pulled up after hours by LRCX

Net after-hours effect: MSFT (on the order of +$500 billion of market cap) and META (−9%) went in opposite directions, with LRCX/AMAT positive and QCOM/ARM negative. This is not one unified "AI earnings-season verdict" but a stratification: cloud and equipment passed, advertising and handsets failed.

5. Flows and sentiment

Sector rotation (11 SPDR sectors + theme ETFs)

Sector/ETF Change % Sector/ETF Change %
XLE Energy +1.88% 🟢 XLB Materials −1.15%
XLP Consumer Staples +0.34% 🟢 XLU Utilities −1.34%
XLRE Real Estate −0.11% XLF Financials −1.60%
XLC Communication −0.15% XLK Technology −2.64%
XLV Healthcare −0.61% XLI Industrials −3.19% 🔴
XLY Consumer Discretionary −0.77% — —
Theme: IGV Software +0.64% 🟢 Theme: SMH Semiconductors −4.79% 🔴
GLD Gold +0.46% SOXL (3x semis) −16.02%
KRE Regional Banks −0.79% XBI Biotech −1.25%
JETS Airlines −2.79% OIH Oil Services −0.92%

Four qualitative judgments:

  1. Risk appetite: clearly risk-off, but "targeted" rather than "across the board." VIX +13.45% breaking 20, 9 of 11 sectors down, the Dow's worst single day since April 2025 — all evidence of risk-off. But the equal-weight S&P fell only 0.90%, software rose, and micro-cap speculation was extremely active, which says capital has not left, it is violently repricing between sectors.
  2. Rates were the true pricing core today, and it was "steepening" rather than a "parallel shift." 2Y −4bp / 10Y +5bp / 30Y +9bp, TLT −1.65% vs IEF −0.42%. A bear steepening hits long-duration assets (high-valuation growth) and spread businesses (financials), which explains why XLF fell under a hawkish Fed — the pre-market mapping of "hawkish → XLF benefits" erred in equating "hawkish" with "the short end rising."
  3. An inflation premium rather than a growth premium: rates↑ + dollar↓ (DXY −0.61%) + crude↑8.12% + gold↑. The dollar falling on a day rates rose is the single most notable thing to record this round — the market's reading of this hawkishness is not "the US economy is stronger" but "the Fed is being pushed along by inflation."
  4. The semiconductor decline came on volume: SOXL −16.02%, NVDA traded 147 million shares and closed at the low of the day, MU 68.5 million shares. Energy and software, the gainers, were relatively subdued. The pattern recorded pre-market ("volume is still on the declining semis, while the rising sectors are absorbed on lower volume") repeated for a third time today, and no style-shift signal of "the rising sectors trading on volume for the first time" has yet appeared.

6. Next-day outlook (2026-07-30, Thursday)

① Theme persistence

Theme Status today Next-day judgment Basis
Energy / geopolitics Main advance, +8.12% Continues, but lower position expectations The driver is still there (Brent above $90), but after a single-day +8% the odds on chasing longs have clearly deteriorated. ⚠️ This theme is event-driven and can reverse on a single diplomatic headline. Watch whether the E&P vs oil-services divergence converges
AI capex trust crisis Intraday climax, partly rebutted after hours ⚠️ Explicitly shifts to "two-way," no longer one-sided bearish MSFT Azure +43% + LRCX double beat + AMAT turning positive after hours are real fuel for a semiconductor bounce. But META's lifted capex floor + QCOM's cost pressure are on the other side
AI power Broad ebb Watch the quality of any bounce, take no directional view Today proved this theme has very high beta to AI sentiment and very low sensitivity to its own fundamentals (VRT raised both guides and still fell −17%). If MSFT's capex spillover shows up tomorrow, BE/VRT/GEV/PWR are the high-beta vehicles — but that would be a sentiment bounce, not a fundamental repair
Rates and duration Bear steepening Adjudicated directly by core PCE tomorrow The 10Y is closing in on 4.70% and the 30Y is already at 5.19%. If core PCE comes in above expectations, the steepening intensifies and today's sector structure plays out again
Momentum breakdown / value reallocation Interrupted Downgraded to "awaiting re-rating" XLV/XLF/XLI all fell today and every vehicle for this theme failed. No stock should be recommended on this basis until it works again
Software vs semiconductor divergence New theme (missed today) ⚠️ Explicitly listed as a next-day priority to track IGV +0.64% vs SMH −4.79%. If MSFT's Azure +43% converts into continued software strength tomorrow, the divergence widens; if semiconductors bounce on LRCX/AMAT, it converges. This is tomorrow's cleanest binary

② Tomorrow's earnings and macro calendar

Time (ET) Event Focus
08:30 pre-market Q2 GDP advance estimate The quality of growth against a backdrop of high rates + AI capex; focus on consumer spending, equipment investment, inventories
08:30 pre-market June PCE / core PCE + personal income and spending Core PCE consensus +0.2% m/m (prior +0.3%). This is the first inflation report card after today's hawkish meeting — and tomorrow's single biggest variable
08:30 pre-market Initial jobless claims The other leg of the Warsh meeting's dispute (employment); the pre-market list already recorded media reporting that the labor market is at multi-year lows
After the close AAPL (苹果 / Apple) iPhone sales, services, gross margin. ⚠️ AAPL was −0.56% today and +0.71% after hours, and it is up 25% this year, so expectations are not sitting low
After the close AMZN (亚马逊 / Amazon) AWS growth and capex guidance — after MSFT delivered Azure +43%, the comparison bar for AWS has been raised significantly. AMZN was −1.82% today
All day Next-day regular-session pricing of MSFT / META / QCOM / ARM / LRCX Whether after-hours prices get confirmed in the regular session. Today's lesson is that the immediate reaction and the reaction 2 hours later differ enormously (MSFT +1%→+4.98%), so tomorrow's difference between the open and the after-hours price likewise needs fresh observation

③ Focus items (ticker + verification point)

# Name Reason to watch Verification point
① MU / SNDK / WDC (memory) 🔴 The strongest logical contradiction found today. QCOM explicitly attributed margin pressure to "rising costs of memory and others," and KLAC likewise called "storage pricing a headwind" — two independent companies confirming from the buy side that memory prices are rising. Yet MU was −9.94% and SNDK −7.32% today, on the grounds of SK Hynix's miss. Buyers complaining memory is too expensive and sellers' shares collapsing cannot both be reasonable at once Whether MU/SNDK/WDC can outperform SMH tomorrow. If they can, the market is starting to price "memory price increases"; if they keep underperforming, what the market has settled on is "demand has peaked" rather than "price," and the contradiction resolves in favor of the demand-side explanation
② SMH / LRCX / AMAT Semiconductors, −4.79% today, got two pieces of positive evidence after hours: LRCX's double beat +4.91% and AMAT +1.96% Whether SMH can turn positive in the first 30 minutes tomorrow, and whether LRCX can hold the after-hours $264.74. ⚠️ Note: TER today used a +14.74% gap to prove that "failing to hold the after-hours price" is the norm
③ MSFT Azure +43% is the hardest positive fact of the day ① whether the regular session can hold the after-hours $409.99; ② exactly what FY27 capex guidance was given on the call (this report could not confirm it, see §7 open item 3) — that number is still unpriced by this report
④ META −8.98% after hours, but the EPS miss was caused by one-time charges of $2.4 billion in legal and $1.2 billion in severance Whether the market strips out the one-time charges tomorrow. If META narrows its decline substantially, "one-time" has been accepted; if it keeps falling, what the market is selling is the Q3 revenue guidance and the capex floor, not EPS
⑤ XLE / MGY / COP The only long theme that delivered, with Brent above $90 Whether Brent can hold $90; and whether OIH/SLB (oil services) catch up. If oil services keep underperforming, this is only an oil-price trade, not the start of an energy capex cycle — a negative judgment on the persistence of energy equities
⑥ CSGP Bounced 15.2% off the low of the day, R3's reversal validated Whether it can reclaim $30.33 (the 7/28 close) tomorrow. If it can, the structure of "cut revenue guidance, protect profit guidance" has been fully accepted by the market, and the pattern can be generalized into a recognition framework

④ Directions to avoid

  1. 🔴 Chasing the names that surged after hours today (MSFT $409.99 / LRCX $264.74 / SBUX $110.49 / CMG $36.29). TER just demonstrated the consequence in the most extreme way today: after hours +13.53% → pre-market +7.0% → open +14.74% → close −0.39% (the low of the day). This pattern has now appeared for the third time this week (details in §6⑤ item 2).
  2. Establishing directional positions in duration-sensitive assets before core PCE is released (08:30 ET). The 10Y is already at 4.657% and the 30Y at 5.193%; today's sector structure was driven entirely by steepening, and a core PCE print above expectations would make today repeat.
  3. Bottom-fishing AI power (BE / VRT / GEV / PWR). Today already proved it: raising both guides and beating consensus (VRT) can still get you killed 17.26%. Within this theme, good news at the company level currently carries no pricing power.
  4. Continuing to use the "front-end capex exposure vs back-end shipment-volume exposure" dichotomy for recommendations. This hypothesis was directly refuted today by LRCX (front-end, double beat, +4.91% after hours), and it should be retired rather than patched (see §6⑤ item 3).
  5. Endorsing a single stock based on the direction of one sector ETF. Today ISRG underperformed inside XLV by 1.79pp while STX rose +2.29% against SMH −4.79% — sector direction as a filter neither blocks the wrong names nor lets the right ones through, so it can only serve as grounds for a downgrade, never for an upgrade.
  6. Treating micro-cap moves (DFNS +109%) as a tradable signal. But it should start being recorded as a risk-appetite reading.
  7. Inferring central-bank behavior from "the drift direction of prediction-market pricing." Today's direct lesson is in §6⑤ item 1.

⑤ Input notes for the next pre-market list (feedback for retraining)

The 7 items below are what this recap requires tomorrow's /us-premarket to implement point by point and report back on at the end of the document.

  1. 🔴 Prohibit inferring policy outcomes from the "drift direction" of prediction-market pricing. R2 was right to withdraw the stale data today, but it then used "hike pricing retreating from 46.5% to about 30%" to change theme 1 from "bearish" to "two-way" — and the Fed then delivered a hawkish hold with 3 votes for a hike, more hawkish than the retreating pricing. Rule: the level of prediction-market pricing may be used as a reference, but its drift direction must not serve as the basis for a directional judgment; the two-way nature of an event should be expressed by "not taking a position" rather than by "flipping the direction."
  2. 🔴 Codify "the delivering name opens at its high" into a scoring rule — the trigger condition has been met. The pre-market list itself stated "if this occurs a third time this week, it should be codified into a scoring rule." Today was the third occurrence and the most extreme: TER opened +14.74% and closed −0.39% (at the low of the day); BE opened +10.0%, high +11.3%, closed −1.85%; F high +8.9%, closed +2.14%; STX high +9.1%, closed +2.29%; WDC opened 480.40, high 499.19, closed 462.04. Rule: for names up >5% pre-market, deduct additional points in both "trading characteristics" and "risk deductions," cap the conclusion at "watch closely," and give no wording that implies chasing.
  3. Retire the "front-end capex exposure vs back-end shipment-volume exposure" dichotomy. Counter-evidence: LRCX (front-end) double-beat and +4.91% after hours, AMAT +1.96% after hours, while TER (back-end) gave back its entire gap. This axis has had only TER as a single supporting data point since it was created, and it was directly refuted today, so it should be deleted rather than down-weighted.
  4. A theme direction judgment must always be given together with "intra-sector dispersion." Today's biggest omission was that the 5.43pp fissure between IGV +0.64% and SMH −4.79% was masked by the single number XLK −2.64%. Rule: whenever XLK is cited as the direction of the tech sector, SMH and IGV must be listed alongside it; whenever XLE is cited, OIH must be listed alongside it (today E&P rose while oil services fell).
  5. A geopolitical event's theme strength must not be cut on the grounds of "attacks intercepted / no capacity loss." Brent was +8.12% today, and pre-market that very reasoning was used to cut both the logical strength and the persistence of the energy theme, with the result that the only money-making direction of the day was missed. Rule: pricing a risk premium does not require actual capacity loss; "no damage caused" may only be used to judge persistence, never to judge same-day direction and magnitude.
  6. Delete the "hawkish → long financials (XLF/JPM)" mapping. The Fed was hawkish today and XLF was −1.60%, JPM −3.53%. The correct intermediate variable is the shape of the curve, not the policy stance: today was a bear steepening (2Y −4bp / 30Y +9bp), which is unfavorable for spread businesses. Rule: any directional judgment involving the financial sector must first state the direction of change at all three points — 2Y/10Y/30Y — before any conclusion is drawn.
  7. Unresolved gaps (status must be flagged explicitly in tomorrow's list): ① NYSE advance/decline — not obtained for the fourth day running; ② Yahoo/yfinance returned 429 across the board and was unusable all day — an operations item; recommend manually confirming the data quota or switching to a backup source; ③ MSFT FY27 capex guidance unconfirmed; ④ FVRR's EBITDA guidance basis unverified for a second consecutive day; ⑤ CMG's specific earnings figures not obtained.

7. Open items (flagged proactively by this report)

  1. Two readings for the 10Y yield: the intraday closing basis was 4.657% (+5bp) (close to, but not equal to, another media basis of "+7bp to above 4.67%"), while CNBC's 17:05 ET after-hours quote showed 4.687%. This report uses 4.657% as the close in §1 and flags the after-hours drift to 4.687%; it was unable to obtain a single authoritative closing snapshot. ⚠️ Also: CNBC's 2Y field shows last 4.266% / prev 4.236%, which is offset relative to the media basis of "the 2Y fell 4bp to 4.236%"; this report adopts the latter (because it is self-consistent with the relative performance of TLT/IEF), but the conflict is unresolved.
  2. NYSE advance/decline — not obtained for the fourth day running. This report substituted two proxy indicators (9 of 11 sectors down, RSP outperforming SPY by 0.64pp) and has clearly flagged that they are not equivalent to A/D counts.
  3. 🔴 MSFT's FY2027 capex guidance is unconfirmed. Microsoft's press release explicitly stated that forward guidance would be given on the call, and the release contains no FY27 capex figure. The "$255–260 billion" circulating online comes from an "Earnings Preview" article published before the print, making it a forward expectation rather than disclosed guidance, and this report refuses to cite it as fact. ⚠️ It also conflicts with the market expectation recorded in the pre-market list (about $220 billion, +20~30%) — meaning even "what tonight's bar actually is" exists in two different versions. One media aggregation also mentions "capex guidance of $41 billion," which, set alongside the company-disclosed Q4 actual capex of $35.8 billion, is more likely next quarter's single-quarter guidance rather than an annual figure, and this report does not adopt it.
  4. Whether FVRR also maintained/raised EBITDA guidance — unverified for a second consecutive day. FVRR was −20.53% today, delivering on the bearish judgment, but "the conclusion was right" is not the same as "the evidence was sufficient at the time"; this blind spot (the same type as CSGP) remains open.
  5. CMG's specific earnings figures were not obtained, only the after-hours price of +5.99%.
  6. ✅ Closed — the conflicting basis for crude's prior close (pre-market open item 6). CNBC shows the Brent September contract's prior close at $84.09, consistent with the value implied by today's sources, proving that the $85.00 recorded in yesterday's recap was wrong. This report also completed, as required pre-market, the ETF cross-check on a no-roll basis: USO +7.32%, BNO +7.89%, consistent with the futures' +8.12% in both direction and magnitude.
  7. ✅ Closed — missing real-time readings for VIX / 10Y / DXY (pre-market open item 7). Closing values for all three were obtained today.
  8. The conflict between the Nasdaq 100's technical pullback and "rotation rather than panic" — pre-market open item 1, partly answered by this report but not completed. Today's evidence points to a third description, "targeted de-rating": VIX breaking 20 (evidence on the panic side) and the equal-weight S&P down only −0.90% with software rising (evidence on the rotation side) hold simultaneously. A full re-assessment still has not been done.
  9. The nature of QCOM's FY26Q2 one-time net income of $7.370 billion — pre-market open item 12, no progress today.
  10. BE's implied share-count assumption behind its FY26 per-share guidance, and its Q2 gross margin / operating cash flow — pre-market open items 11 and 10, neither of which could be advanced because the Yahoo endpoints were unavailable. The 10-Q must be checked.
  11. ⚠️ Global deviation in the data path: all price data in this report comes from stockanalysis.com and CNBC, without yfinance verification (Yahoo returned 429 across the board). Company earnings figures were traced back to press releases wherever possible (MSFT verified), but the financial figures for META / QCOM / ARM / LRCX / SBUX are from media aggregations and were not confirmed first-hand from company press releases.

⚠️ Risk disclaimer: this recap is only a post-close information review and observation, and does not constitute investment advice. Data may differ in timeliness or basis — please defer to company disclosures / SEC filings, and do not use this directly as a basis for trading.