US Market Recap | 2026-07-30 (ET) Thursday

Coverage window: 2026-07-30 09:30 ET open → 16:00 ET regular-session close (after-hours window: 16:00 → 18:00 ET, broken out separately in §4). Data capture time: 2026-07-30 18:00–18:20 ET (yfinance daily / 5-minute bars including pre- and after-hours + WebSearch cross-verification). Basis for percentage changes:

• Single stocks / ETFs / indices daily change = 7-30 close ÷ 7-29 close − 1, from yfinance daily bars, consistent with media convention.
• After-hours change (§4) = closing price of the last 5-minute bar at 18:00 ET ÷ 7-30 regular-session close − 1. After-hours volume is extremely thin and still moving; this is a snapshot, not a final print.
• Treasury yields use ^TNX/^TYX (CBOE index convention), in percentage points; may differ from cash-bond terminals by 1–3bp.
• This report did not obtain primary NYSE advancers/decliners data for 7-30; §1's market breadth instead uses RSP (equal-weight S&P) vs SPY (cap-weighted) and the up/down distribution of the 11 SPDR sector ETFs as proxies, explicitly labeled as proxy indicators.

---

0. One-Sentence Recap

Today was risk-on, but an "extremely narrow" risk-on — a victory at the index level, a failure at the breadth level. The S&P rose +1.66%, the Nasdaq Composite +2.78% (ending a six-session losing streak), the Nasdaq 100 +3.36% — but the equal-weight S&P RSP closed down 0.16%, and among the 11 SPDR sector ETFs only 5 advanced, with the defensive trio XLP -2.16%, XLV -1.64%, XLU -0.56% all being drained. Virtually the entire index gain came from just two threads: ① MSFT +15.51% in a single day, the largest single-day market-cap gain in history (roughly $450–490 billion, sources disagree); ② a broad blowout in memory/semiconductors, with the Philadelphia Semiconductor Index SOX +8.19%, SNDK +25.99%, MU +18.36%, WDC +15.37%, STX +11.41%. The strongest theme today was not AI compute, it was memory — and that is precisely the call this report's pre-market list got most wrong. Of the 20 names on the pre-market list, 13 were directionally correct (65%); MSFT/BE/META/AAPL/MO were called right, while MU/SNDK/AMD — three "watch only" names — collectively missed +18%/+26%/+13%, the biggest loss of points on the day. On positioning and sentiment, VIX collapsed 17.28% to 17.09, the dollar fell -0.78%, gold rose +3.17% — but the 10Y rose 4bp to 4.66% and the 30Y rose 7bp to 5.21%, a new high for this cycle: the stocks-up/bonds-down divergence not only failed to heal today, it widened. Tone for tomorrow: after-hours AAPL -6.35% (memory costs crushing September-quarter guidance) and AMZN +9.23% (AWS +36.7%, fastest in 18 quarters) point in opposite directions, so mega-cap tech will tear apart internally at tomorrow's open — and the memory chain this report misjudged in §2 will have to face the first hard evidence of "upstream price hikes = downstream margin destruction" (supplied by AAPL itself).

---

1. Market Overview

1.1 Indices and Volume

Broad-index ETF volume expanded: SPY 65.91 million shares vs prior-5-day average 51.94 million (+27%), QQQ 65 million vs 47.64 million (+36%) — the rally had volume behind it, and that part is real.

1.2 Market Breadth (proxy indicators, not advance/decline counts)

⚠️ The single most important item in this section: RSP (equal-weight S&P) −0.16% vs SPY (cap-weighted) +1.68%. The equal-weight index closed down on a day the cap-weighted index rose 1.68%, which means the median S&P 500 constituent fell today, and the index gain came almost entirely from a very small number of mega-cap names.

Of the 11 SPDR sector ETFs: 5 up / 6 down (full ranking in §5.1). That is inconsistent with a +1.66% index gain — this was not a broad advance, it was a highly concentrated reallocation of capital.

Honest labeling: this report did not obtain NYSE/Nasdaq advance/decline counts for 7-30 (only 7-29's: 1,473 advancing / 3,273 declining, A/D 0.45:1). The breadth judgment above rests on two proxies, RSP vs SPY and the sector distribution; the direction of the conclusion is credible, but it lacks primary advance/decline support.

1.3 VIX / Rates / FX / Commodities

1.4 Sentiment Assessment

Qualitatively: risk-on, but a "narrow, duration-preference" risk-on, not an "economy-is-improving" risk-on.

Three mutually corroborating pieces of evidence:

1. Stocks up + long-end yields up + dollar down + gold sharply up + equal-weight index down — this combination is not a typical recovery trade. A typical recovery trade would be stocks up, dollar strong, gold weak, equal-weight participating.
2. Today's two macro releases pointed in opposite directions: Q2 GDP advance estimate +1.5% (prior 2.1%, below expectations — the expectation figure differs by source, investing-focused media report 2.1%, CNBC reports 1.8%, so this report does not adopt a single expectation value), while initial jobless claims came in at 197,000 with layoffs at a 57-year low (investinglive¹ / BEA²). Growth is slowing, employment is not breaking, inflation (core PCE MoM +0.1%) is cooling — a textbook "bad news is good news" structure, favorable to long-duration assets rather than cyclicals. This is fully consistent with "XLK +5.50% while XLI only +0.98% and XLB −0.19%."
3. Gold +3.17% rising alongside equities usually points to coexisting "expectations of falling real rates + currency-debasement concerns," not a simple recovery in risk appetite.

⚠️ The third risk signal in §8 of the pre-market report ("if the 30Y keeps rising while the index still rallies, this divergence is unsustainable") was today neither falsified nor confirmed — the divergence widened (30Y +7bp to 5.21% while the Nasdaq rose +2.78%). This risk remains outstanding.

---

2. Reconciliation Against the Pre-Market List

2.1 Name-by-Name Check

3A bullish direction (pre-market §3A)

3B bearish / avoid direction (pre-market §3B, §6)

2.2 Hit Rate

Strict directional hit rate: 13 / 20 = 65%.

• Called right (13): MSFT, BE, TSM, AVGO, NVDA (direction), F, GOOGL, META, QCOM, AAPL, MO, STLA, CI
• Called wrong (7): MU, SNDK, AMD (three "watch only" names that surged on the day), VLO (direction reversed), AMZN, MA, REGN (risk control correct but the move was missed)

Classified by "did it cause an actual loss": not a single recommendation would have caused a loss (everything on the avoid list fell or was not recommended for shorting; everything on the recommended list rose), but 6 items caused significant missed gains, among them SNDK +25.99%, MU +18.36% and AMD +13.00%, three names that together sat at the top of the day's market-wide gainers.

2.3 A One-Line Self-Critique (no whitewashing)

This report's biggest pre-market error was not a wrong judgment, it was turning a "two-way theme" into a one-way one — and betting on the wrong side.

Pre-market §2 listed memory as a "two-way S-grade" theme, and §5.3 spent 2,000 words arguing that MU was "a classic trap of peak earnings + low P/E," with the core evidence being "the sequential price increase has already fallen from 60%+ to 13–18% (TrendForce Q3 guidance)" — i.e. the second derivative of the price slope had turned negative. Yet during today's session, Samsung's Q2 conference call directly overturned that premise: Samsung said Q2 DRAM ASPs rose mid-40% QoQ and NAND rose high-60%, that Q3 bit-shipment growth would still be only single-digit, and stated explicitly that "the DRAM shortage in 2027 will be more severe than in 2026," with the shortage lasting into 2028 (Seoul Economic Daily⁵ / OC3D⁶). Layer on Seagate's 7/28 earnings with revenue +48%, a record 52.7% gross margin and order visibility out to 2028, plus UBS's bullish report on agentic AI driving DRAM/NAND bit demand through 2027 (24/7 Wall St.⁷).

Three lessons that must be recorded:

1. "The cycle has peaked" is a multi-quarter judgment and cannot be used as a single-trading-day judgment. The pre-market argument of P/B 8.28× vs 1.26× in the prior cycle may still hold on a cross-cycle horizon, but it has no predictive power whatsoever for today's price action — and yet this report used it as the conclusion for the day ("watch only"). Expensive valuation ≠ it falls today.
2. Pre-market, characterizing SNDK's +7.13% as "a bounce driven by the deepest drawdown, not fundamental repair" is unfalsifiable language — after the fact it explains any outcome, up or down. The close at +25.99% shows that "characterization" was merely a renaming of the gain, with zero information content. Next time facing a deeply oversold name, either state a specific falsification condition (e.g. "if the close is up more than 15%, this judgment is void") or give no characterization at all.
3. The first item of §9③, "avoid chasing any name that gapped up sharply (MSFT +9.7%, BE +8.1%, SNDK +7.1%)" — all three closed far above their pre-market levels (MSFT +5.3%, BE +17.0%, SNDK +17.6%). The rule of thumb that "gaps get filled" broke down entirely on a day of strong catalysts — it applies to sentiment gaps without news, not to gaps driven by earnings or an industry inflection. This report treated a statistical regularity as a universal law.

One thing done right must also be recorded: pre-market §5.7 judged AAPL's key risk to be "memory price hikes pushing up BOM cost," and placed AAPL on the harmed downstream side of the memory theme. Tonight's Apple call confirmed it: September-quarter guidance weakened due to supply constraints and memory cost, with gross margin guided to 47–48%. Across the whole memory chain, this report got the upstream backwards but got the downstream right — which shows the framework for supply-chain transmission was correct; what was wrong was locating where we currently sit along that chain.

---

3. Theme Verification for the Day

3.1 Did the Pre-Market Report Identify the Strongest Theme?

No. The three strongest themes named in pre-market §9② were ① AI capex divergence ② the memory-top standoff ③ a hawkish Fed meeting cooling inflation.

• The "roster" of theme rankings was right (memory was indeed in the top two), but the direction label was wrong — pre-market labeled memory "top standoff / watch only," when it was in fact the strongest long theme of the day.
• What was genuinely called right was ①, and called precisely (two opposite verdicts on the same theme on the same day).
• ③ picked the wrong expression vehicle (chose IWM, should have chosen QQQ).

3.2 Surprise Themes the Pre-Market Report Missed

---

4. After-Hours Earnings Moves (16:00–18:00 ET, catalysts for tomorrow)

⚠️ After-hours volume is extremely thin; the following are 18:00 ET snapshots, not closing prices, and will keep moving before the open.

---

5. Flows and Sentiment

5.1 Sector Rotation (11 SPDR sector ETFs + key broad indices, sorted by today's gain)

The shape of the rotation is extremely clear: an extreme, concentrated allocation flowing one way, out of defensives and value and into technology and semiconductors. Of the 5 sectors that rose, 4 gained less than 1% — in practice only XLK rose at all; of the 6 that fell, 4 fell on expanding volume (XLU/XLRE/XLV/XLP), meaning this was active position reduction, not a passive drift lower for lack of bids.

5.2 VIX and Treasuries

• VIX 17.09 (−17.28%): a single-day collapse of 3.57 points, from 20.66 back to around 17. It is already back in the recent-month low range with limited room for further compression — which means if negative news appears tomorrow, VIX's upside elasticity exceeds its downside, an asymmetry that works against the bulls.
• 10Y 4.66% (+4bp), 30Y 5.21% (+7bp), 5Y 4.38% (+3bp): the long end rose more than the front end, the curve keeps steepening. On a day when Q2 GDP was only 1.5% and core PCE MoM only +0.1%, long-end yields rose 7bp rather than falling — this is not growth/inflation pricing, it looks more like term-premium pricing (supply, fiscal policy, Fed independence).
  • ⚠️ This report cannot prove the attribution above. It can only confirm the fact: the data was dovish while the long end was hawkish, and the two diverged.
• Dollar −0.78% + gold +3.17% + rising long-end yields appearing together is a combination that warrants caution. In the textbooks it usually corresponds to "the market demanding higher compensation for holding dollar assets." This report only records it and draws no verdict — a single day of data is insufficient to support a structural judgment of that kind.

5.3 Risk-On / Risk-Off Assessment

Conclusion: in form it is risk-on; structurally it is a "preference for duration and scarcity," not a "preference for economic strength."

In one line: the money that bought today was real money (it had volume), but it was not "new money" — it was "old money" pulled out of defensive sectors and the equal-weighted median stock. The durability of this kind of tape depends on how long the draining can continue, not on macro improvement.

---

6. Outlook for Tomorrow (2026-07-31, Friday)

① Theme Persistence

② Tomorrow's Earnings and Macro Calendar (ET)

Macro data

(Calendar source: Markets Today calendar⁹. ⚠️ This source is a social-media restatement; this report did not verify release times against BLS or the UMich official site as primary sources — use your own terminal as the reference.)

Pre-market earnings (before the 07-31 open)

• The energy duo: XOM, CVX (Chevron confirmed for pre-market release) — with Brent down −1.42% while XLE edged higher, these two reports will set the short-term direction of the energy sector
• Pharma: ABBV, MRNA (with XLV −1.64% today and SYK −8.06% after hours, healthcare has been under sustained pressure; ABBV is the stabilizer)
• Industrials/materials: LIN (Linde), ETN (Eaton) — ETN is an important cross-check for the data-center power theme; strong ETN guidance would add another brick to the power theme in §6①
• Consumer/other: CL, SONY, D, CCJ, CBOE, ARES, TROW, AN, IMO

(Sources: Markets Today earnings calendar¹⁰ / MarketBeat¹¹. ⚠️ Other than CVX, this report did not verify release times one by one against company IR pages.)

Already-occurred after-hours events to be priced at tomorrow's open: AAPL −6.35%, AMZN +9.23%, RDDT −11.12%, SYK −8.06%, COIN −5.86%, BE +5.02%, MU +3.30% (all 18:00 ET snapshots)

③ Key Focus for Tomorrow (Ticker + verification point)

④ What to Avoid Tomorrow

1. ⚠️ The most important item: do not chase memory higher just because today's call on memory was wrong. After SNDK +25.99%, MU +18.36% and WDC +15.37% in a single day, tomorrow's risk/reward is entirely different from what it was at today's open. Today's lesson is "do not use valuation to judge a single day's direction," not "valuation does not matter." The two conclusions must not be conflated.
2. Avoid chasing MSFT. After +15.51% and the largest single-day market-cap gain in history, it is already −0.69% after hours. The two quality blemishes flagged in pre-market §5.1 (a clean beat of only +3.5% for the quarter, FCF/net income down to 0.50×) were not falsified today, they were merely drowned out by the strength of the guidance.
3. Avoid consumer electronics and PC/handset downstream hardware. AAPL has itself confirmed that memory cost will significantly compress September-quarter gross margin, and that cost pressure works in the same direction for every downstream manufacturer — and Apple is the one with the most bargaining power.
4. Avoid names that "beat but with a blemish in visibility." Tonight's RDDT (revenue +61%, a double beat, strong guidance, crashing 11.12% purely on one line about "search referrals were choppy") and SYK (EPS beating by 5.8% yet falling 8.06%) show that the market currently weights "visibility" far more heavily than "the quarter's numbers." These are three faces of the same rule that lifted MSFT on guidance and crushed META on its FCF collapse.
5. Avoid running volatility-sensitive long positions with VIX at 17.09. VIX is back at recent-month lows, with little downside and high upside elasticity.
6. Avoid using defensives and high dividends as a "hedge" allocation. Today proved that in this kind of extremely concentrated tape, defensive sectors are not a safe harbor, they are the source of funds (XLP −2.16%, XLV −1.64%, all falling on expanding volume).
7. Avoid using AMZN's EPS of 5.75 in any valuation calculation (see the Anthropic mark-to-market caveat in §4).
8. Avoid linear inference from "oil price → oil stocks." Falsified two days running: yesterday oil up and stocks down, today oil down and stocks up.

⑤ Input Notes for Tomorrow's Pre-Market List

Seven inputs that must be handled in the 07-31 pre-market report:

1. The AAPL September-quarter guidance must be addressed at the top (9–11% revenue growth vs the expected 12%, gross margin 47–48%, supply constraints spanning iPhone/iPad/Mac, Cook calling memory a "once-in-a-century flood"). This is tomorrow's single largest pricing event, and it pushes the memory theme into phase two.
2. Both sides of the memory chain must be repriced from scratch: upstream (MU/SNDK/WDC/STX/SK Hynix) and downstream (AAPL/QCOM/DELL/HPQ/Lenovo, etc.). This report's pre-market framework was right (splitting upstream from downstream); the positioning was wrong (assuming the upstream had already peaked). Tomorrow should keep the framework, reset the positioning, and write out explicit falsification conditions.
3. The macro dimension missed today must be added back: gold 4,162.80 (+3.17%), the dollar 100.01 (−0.78%), the 30Y at 5.21% (+7bp). In particular, the combination of "stocks and gold up together + dollar down + long-end yields up" needs either an explanation or an explicit "no explanation available at this time."
4. Market breadth must be written into §0: the 184bp gap between RSP and SPY is today's most important structural fact, and tomorrow's pre-market report should make "equal-weight vs cap-weighted" a standing monitoring item.
5. AMZN's clean EPS must be verified (on a basis excluding the Anthropic mark-to-market), otherwise no score may be assigned to AMZN. This must be consistent with how this report treated MSFT on 7-30 — we cannot be strict with MSFT and lenient with AMZN.
6. AMZN's capex guidance was not obtained by this report and must be filled in. Of the three thresholds pre-market §5.5 set for itself, "② whether capex guidance is revised up" still has no answer. Ahead of the print, sell-side expectations had already risen from $200 billion to $210 billion (BofA, on rising memory cost).
7. The cross-source discrepancy on the 10Y yield remains unresolved (this report's ^TNX convention gives 4.66%; TradingEconomics reported 4.70% in the 7-30 pre-market report). Tomorrow should fix one convention and state it.

---

Appendix: Evidence Gaps in This Recap (must be disclosed honestly)

1. NYSE/Nasdaq advance/decline counts for 7-30 were not obtained. The breadth conclusion in §1.2 rests on two proxies, RSP vs SPY and the sector distribution. The direction is credible, but primary advance/decline data is missing.
2. SMH and SOXX give conflicting volume signals (0.67× vs 1.30×); this report cannot determine which is more representative, and therefore drew no sector-level conclusion about whether the rally had volume. Single-stock volume readings (MU 1.15×, SNDK 1.17×, MSFT 3.3×) are reliable.
3. AMZN's EPS of 5.75 has not been decomposed for one-off items. Its stake in Anthropic is known (carrying value $74.2 billion at the end of Q1) and BofA flagged large mark-to-market swings ahead of the print, but this report did not obtain the specific Q2 breakdown, so §4 only flags the caveat and makes no quantitative adjustment. AMZN's capex guidance was likewise not obtained.
4. Some AAPL segment data is second-hand restatement: iPhone $54.25 billion (+21.7%), Services $30.98 billion (+12.1%, below the expected $31.22 billion), Greater China roughly $18.82–18.86 billion (+22%, below the expected $19.58 billion) come from media (Yahoo Finance¹² / MacObserver¹³); Apple's press release itself contains no segment breakdown, and this report did not check the 10-Q. Revenue of $109.4 billion, EPS 2.02 and gross margin 50.1% come from Apple's official press release¹⁴ and are reliable.
5. Sources disagree on AAPL's after-hours decline: this report's 18:00 ET snapshot is −6.35%, a CNBC headline says −8%, and other media say −4%. The after-hours price fell from 318.68 all the way to a low of 305.07 between 17:05 and 18:00 before rebounding to 312.25 — extremely volatile, so all three numbers may be correct, just at different times.
6. Sources disagree on MSFT's single-day market-cap gain: Bloomberg reports about $450 billion, Fortune about $480 billion, and other headlines say $490 billion. This report uses the unified phrasing "roughly $450–490 billion" and does not use a single figure. "The largest single-day market-cap gain in history (surpassing NVDA's prior record of about $440 billion)" follows Bloomberg's convention.
7. Sources disagree on the market's expectation for Q2 GDP: investinglive¹ reports an expectation of 2.1%, CNBC¹⁵ reports 1.8%. The actual value of 1.5% and the prior of 2.1% are uncontested; this report does not adopt a single expectation figure.
8. Tomorrow's calendar (earnings roster + macro timings) comes mainly from a social-media restatement of a calendar, and other than CVX was not verified name by name against company IR / BLS / the UMich official site.
9. Samsung Q2 conference call data is second-hand restatement (DRAM ASP +mid-40% QoQ, NAND +high-60% QoQ, "2027 more severe than 2026," shortage lasting into 2028). This report did not obtain Samsung's original IR text, but the news is highly consistent with the day's tape (SOX +8.19%).
10. "Memory price hikes → downstream margin destruction" is this report's inference, not a fact. Only two facts are confirmed: QCOM cut guidance last week and cited the memory shortage as a reason for price increases; AAPL tonight guided September-quarter gross margin to 47–48% and attributed it to supply constraints and memory. For the remaining downstream companies (DELL/HPQ/Lenovo, etc.) this report obtained no data at all.
11. The statement in §5.2 that "the long-end rise is term-premium pricing" is an inference this report cannot prove. Only the factual layer can be confirmed: the data was dovish (GDP 1.5%, core PCE MoM +0.1%) while the 30Y rose 7bp.
12. The fundamentals of BE, PWR, GEV and VRT still have not been independently verified by this report (the same gap as in the 7-30 pre-market report). The judgments in §6① on the data-center power theme are based entirely on price and news, with no fundamentals.
13. Earnings details for RDDT / SYK / COIN are second-hand restatement; this report did not check their 8-K filings.
14. How the 65% hit rate is calculated: on the basis of "whether the directional label given pre-market matched the day's closing direction," with "watch only" counted as bearish/avoid. This basis is strict on "watch only" — it was meant as a risk-control label, not a short recommendation. Measured by "whether an actual loss was caused," the hit rate would be 100% (0 losses), but that basis carries no information, so this report uses the strict basis while disclosing both.

---

⚠️ Risk disclaimer: this recap is a post-close information review and observation only and does not constitute investment advice. Data may differ in timeliness or convention; company disclosures and SEC filings prevail, and this must not be used directly as a basis for trading.