US Pre-Market Brief | 2026-07-31 (ET), Friday

Coverage window: 2026-07-30 16:00 ET regular-session close → 2026-07-31 08:05 ET pre-market. Items in §1 tagged "carry-over" occurred before this window; they are retained because they still dominate today's pricing, and their scores carry a freshness discount. Pre-market data capture time: 2026-07-31 08:00–08:05 ET (scraped ticker-by-ticker from stockanalysis.com; local yfinance was rate-limited by Yahoo throughout (YFRateLimitError), failing twice in live testing — no yfinance data is used in this report). Note on the change-basis for price moves (important):

• The data source's Previous Close field lags by one day for some names (lag confirmed on MSFT / NVDA / SNDK / STX / AMZN; no lag on AAPL / WDC / META / AMD / VRT / XOM). This report therefore always back-solves the 7-30 close as "pre-market price ÷ (1 + pre-market % change)" rather than trusting the Previous Close field. The back-solved values reconcile three ways with "7-29 close × 7-30 intraday gain" on SNDK / STX / MSFT, so the method is reliable.
• Index futures percentage moves are the media convention, benchmarked to the prior session's futures settlement price, and are not directly convertible to cash-index closing levels; ETF pre-market moves (SPY/QQQ/DIA/IWM) are appended as a cross-reference. The two conventions differ and must not be mixed.
• Valuation multiples (P/E, P/S) are all on a 7-30 closing-price basis and have not been recomputed on pre-market prices — note this especially for AMZN / AAPL (pre-market +11.58% / −7.65% respectively).

---

0. Today in One Line

1. Today's main axis is an "earnings-quality audit": the EPS beats at both mega-caps were largely fake, and the market held only one of them to account. AMZN's headline EPS was 5.75 vs 1.82 expected (a 216% beat), but it includes $53.4 billion of pre-tax non-operating income, mainly from the revaluation of its Anthropic investment — strip it out and operating EPS is roughly 1.95–2.06, a real beat of just +7% to +13%. AAPL's headline EPS was 2.02 vs 1.89 expected, but it includes a $2.19 billion one-off tariff refund (+0.11/share, +2pp of gross margin) — strip it out and you get 1.91, a real beat of just +1%. Both "beats" are accounting artifacts; the market bought AMZN and dumped AAPL, and what it was actually trading was guidance and forward outlook, not the quarter's EPS. That is the single most worthwhile sentence to remember today.
2. AMZN's substantive positive is not in EPS but in AWS and operating profit: AWS revenue $42.23 billion, +37% (fastest in 18 quarters, 5th consecutive quarter of acceleration), AWS operating profit $16.62 billion +64%, margin roughly 39% (+650bp YoY); AWS backlog $496 billion, up $132 billion sequentially — the sequential increment carries more information than the absolute level. Total operating profit $27.46 billion +43%; operating margin 11.43% → 13.69%. Pre-market +11.58%.
3. But AMZN has two red flags that pre-market sentiment papered over: ① TTM free cash flow has turned negative — company definition −$7.6 billion (vs +$18.2 billion a year earlier), TTM capex of $169 billion now exceeds TTM operating cash flow of $161.4 billion, and full-year 2026 capex guidance was raised from $200 billion to $220 billion (partly attributed to memory price increases); ② Q3 operating profit guidance of $22.5–26.5 billion, midpoint $24.5 billion, is below Q2's actual $27.5 billion — a sequential decline that seasonality cannot fully explain, pointing to accelerating depreciation and amortization eating into profit. This deserves more attention than the revenue-guidance miss.
4. The strongest theme is memory price increases, with three testimonies pointing the same way — but honesty requires saying it: their incentives are not independent: Samsung warned that memory supply is tightening → Cook called memory price increases a "once-in-a-century flood" and cut Apple's gross-margin guidance → AMZN raised capex by $20 billion, partly attributing it to memory costs. ⚠️ An earlier draft of this report called this "cross-validation by three parties with opposing interests, the hardest kind to fake"; that claim was withdrawn after quality review: Samsung talking up tight supply helps it raise prices; Cook, with both gross margin and revenue guidance falling short, attributing the cause to external component costs is the most management-favorable, lowest-cost explanation available; Jassy attributing a capex overrun to memory prices likewise restates "deteriorating returns" as "external price increases". On this specific claim, all three parties' incentives actually point the same way, so this does not constitute independent verification. The genuinely independent evidence would be third-party DRAM/NAND contract and spot prices (TrendForce et al.) — not obtained for this report, flagged as pending verification. Morgan Stanley's "at least +25% in Q3" and JPMorgan's "no meaningful new supply before 2028" are both second-hand relays via 24/7 Wall St.; the original research notes were not seen. Storage names surged as a group on 7-30: SNDK +25.99%, MU +18.36%, WDC +15.37%, STX +11.41%, and are up another +4.4% to +5.8% pre-market this morning.
5. Driver type: earnings and guidance dominate outright (AMZN/AAPL/XOM/CVX/STX, plus ETN reporting pre-market today), layered on a supply-side industry event (memory), with macro as backdrop (08:30 ET ECI, 10:00 ET final University of Michigan consumer sentiment today). Two draft errors on the energy side require correction: ① Brent is at 92.27 (07-31 06:45 ET), +24.43% over the past 30 days and roughly +20% for July as a whole — geopolitics really is pushing oil higher — but it is also still about 23.7% below the 52-week high of 120.88 set on 4-30; both things are true and only presenting one side is not acceptable; ② the DOJ probe into the four major oil companies is old news from late June, not a fresh catalyst today; the draft labeled it as "announced 07-31", which is a factual error, now corrected with the impact level downgraded.
6. Pre-market state: Nasdaq futures +1.11% (28,552.00), Dow futures +0.53% (52,660), S&P futures +0.47% (7,507.75); on an ETF basis QQQ +1.08%, SPY +0.35%, DIA +0.60%, IWM +0.34% (08:00–08:05 ET). Brent 92.27 (−0.41%, 06:45 ET). 10Y Treasury roughly 4.69% (highest since January), 30Y roughly 5.23% (a 19-year high) — this is today's most overlooked drag. ⚠️ Note the timestamps on the bond yields and VIX: the 10Y/30Y readings correspond to 7-30 intraday / around the close; live values as of 08:00 this morning were not obtained. VIX 18.01 (−12.83%) is a 7-30 13:31 ET intraday reading — not a closing value, and not a reading from this morning. The "rates and equities are diverging" call is therefore built on prior-day rate data; confidence has been reduced accordingly, and readers should re-check live yields before the open. Dollar index: not obtained this time, cannot be confirmed.

---

1. News Overview

---

2. Strongest Themes, Descending

---

3. Overall Single-Stock Strength Rankings

⚠️ Note on the two tables' scales: 3A and 3B share the same 100-point scale but are not comparable across tables. 3A is sorted by score descending (higher = stronger bullish evidence); 3B is sorted ascending (lower = stronger bearish evidence). The same 55 points means different things in the two tables. ⚠️ Pre-market change basis: all figures are "pre-market price vs the back-solved 7-30 close", captured 07-31 08:00–08:05 ET. ⚠️ Unified convention for "positioning" (the draft was inconsistent; unified here): the draft used "distance from the 52-week high" to argue NVDA was well positioned (−17.5%), yet used the same metric for MU (−30.3%), STX (−25.6%) and SNDK (−45.6%) to call them "worst positioned" — the same yardstick produced opposite readings, because the real basis was the two-day cumulative gain, not 52-week positioning. Both numbers are now shown side by side for readers to judge: NVDA −17.5% from the high / +1.2% over two days; MU −30.3% / +23.6%; STX −25.6% / +17.8%; WDC −29.5% / +22.1%; SNDK −45.6% / +32.9%; AMZN −? (52-week data not obtained) / +11.6%. This report's "crowded positioning" calls rest mainly on two-day gains, not on 52-week positioning.

3A Bullish Zone (by total score, descending)

3B Bearish Zone (by total score ascending; lower = stronger bearish evidence)

---

4. Single-Stock Scoring Model (100 points)

＊A special note on AAPL's score: AAPL sits in the 3B bearish zone, and its 42 points is a reading on the bearish scale, not directly comparable to the names in 3A. It is listed in the table above only to show the transparency of its components; its risk deduction (guidance cut, already −7.65% pre-market, gross margin down sequentially, structural Services deceleration) is magnified on the bearish scale. ⚠️ Two flaws in the model itself, disclosed honestly: ① AAPL's −29 breaches the "0 to −15" range declared in the header, meaning one table mixes two scales — the correct approach is a separate table and formula for bearish names, which this report did not do and the next version will fix; ② WDC/AMD/SNDK/VRT in §3A and all six names in §3B have no component breakdown, so their scores lack a reproducible basis — treat them as tier indications only; ③ the "tradability" column gives 10 points to seven of the nine names, which is insufficiently discriminating and close to filler. ＊＊A second correction: the draft stated that "AAPL's fundamentals component of 14/15 is among the highest in the table" — an overstatement. In fact AMZN/AVGO/NVDA/GOOGL are also at 14, and MSFT's 15 is the highest. The accurate statement is: AAPL's fundamentals component is on par with most front-line tech names and clearly higher than its total of 42 would imply — and that is precisely why it is "watch only" rather than "short watch".

---

5. Detailed Analysis of Top Names

⚠️ Note on this section's ordering: this section is ordered by today's importance, which is not the same as §3's score-descending order. AAPL is listed ② (its 42 points belongs to the bearish scale) because it is today's second-largest pricing event; MSFT and META are listed ⑨ and ⑩ because their catalysts occurred on 7-29 and were fully realized on 7-30, leaving them as background today. If sorted purely by score, the order would be AMZN (80) → MSFT (74) → MU (71) → AVGO (69) → STX/NVDA (67) → GOOGL (65) → CVX (63).

1亚马逊AMZNtotal 80, watch closely · do not chase · Amazon

• Related news: 07-30 16:05 ET, Amazon press release¹. Revenue $200.61 billion (+20%), AWS $42.23 billion +37%, AWS operating profit $16.62 billion +64%, backlog $496 billion (+$132 billion QoQ). On the 17:30 ET call: 2026 capex raised to $220 billion.
• Catalyst logic: this report affects the valuation, not the quarter's EPS. The headline EPS of 5.75 includes $53.4 billion of pre-tax non-operating income (mainly the Anthropic investment revaluation) — cross-checked two independent ways: ① the company's press release states it explicitly; ② the pre-tax margin of 40.31% vs the operating margin of 13.69% implies a gap of ×$200.6 billion ≈ $53.5 billion, consistent with $53.4 billion. Stripping it out gives operating EPS of roughly 1.95–2.06, a real beat of +7% to +13% rather than 216%. The real information content is in the backlog rising $132 billion sequentially — contracted, not-yet-recognized revenue, the hardest evidence that "AI spending comes back", and exactly what META lacks.
• Theme and stage: the AI cloud theme is in a maturing phase leaning crowded. MSFT (7-29) → AMZN (7-30) delivered two consecutive guides confirming cloud acceleration, so thematic consensus has formed; but META's counterexample shows the market is still discriminating and has not entered indiscriminate crowding.
• Fundamental verification (stockanalysis.com, captured 7-31; yfinance rate-limited and unusable):
  • Operating profit $27.46 billion vs $19.17 billion a year earlier, +43%; operating margin 11.43% → 13.69% (+2.26pp).
  • TTM free cash flow −$7.6 billion (company definition), vs +$18.2 billion a year earlier; TTM operating cash flow $161.4 billion, TTM capex $169 billion → capex/OCF > 100%. (stockanalysis' gross-capex definition gives −$11.6 billion; both definitions are correct but must not be mixed.)
  • Valuation: P/E (TTM) headline 18.94, roughly 27x once one-offs are stripped out; forward P/E 25.53 (also contaminated by one-offs; roughly 35x after stripping). "Forward P/E above trailing P/E" is itself the fingerprint of one-off gains. P/S 3.27, EV/EBITDA 15.59. Cash $123 billion / debt $223.2 billion → net debt roughly $100.2 billion.
  • ⚠️ Do not take ROE of 30.56% at face value: the numerator likewise includes the $53.4 billion of one-off non-operating income, the same contamination source as the P/E just adjusted above. Since this report's central thesis is an audit of definitions, ROE cannot get a pass — the adjusted true ROE was not computed in this report and is marked "not computed", and 30.56% should not be read as "high-quality profitability".
  • Historical valuation percentile: no reliable data available (a multi-year P/E time series was not obtained).
• Moat / market position: AWS is the global cloud leader with a 39% margin, up 650bp YoY; in-house silicon (Trainium/Graviton) exceeds a $25 billion annualized run rate, and Project Rainier has deployed 500,000 Trainium chips for Anthropic — in-house silicon is a genuine hedge against NVDA's pricing power.
• Pre-market and technicals: $262.78, +11.58% (7-30 close 235.50). This is the largest gap in this report, and the largest chase risk. JPMorgan raised its target to 365 and Wedbush to 310.
• Final call: fundamentals confirm the positive, but pre-market has already priced in most of the good news. Two red flags must be remembered: FCF has turned negative, and Q3 operating profit guidance at a midpoint of $24.5 billion < Q2's actual $27.5 billion (a sequential decline that seasonality cannot fully explain, pointing to accelerating D&A). Do not chase an 11.6% gap; watch whether it is gap-and-go or an opening fade.

2苹果AAPL4th in the bearish zone, watch only, not a short candidate · Apple

• Related news: 07-30 17:00 ET, Apple press release⁵ and earnings call. Revenue $109.4 billion (+16%, strongest June quarter ever), EPS 2.02 (+29%), iPhone $54.25 billion beat, Mac $10.35 billion a big beat; Services $30.74 billion +12%, below the $31.22 billion expected. Q4 guidance $111.69–113.74 billion vs $114.84 billion expected; gross-margin guidance 47–48%.
• Catalyst logic: the quarter's beat is fake and the guidance miss is real — a textbook case for Iron Rule No. 2 of this report. The $2.19 billion tariff refund contributed +0.11/share and +2pp of gross margin; stripping it out gives EPS of 1.91 vs 1.89 expected, a real beat of just +1%, meaning roughly 85% of the beat came from a one-off.
• The key question: is it a demand problem or a supply/cost problem? — judged to be cost and supply; Cook's account is credible. Three independent pieces of evidence:
  1. Financial evidence: revenue beat, with iPhone and Mac both beating; what was cut was the margin, not demand strength; Q4 revenue guidance of +9–11% is still double-digit growth.
  2. The CFO quantified it directly: "more than 100% of the sequential change in gross margin is explained by memory costs" — an extremely strong attribution.
  3. External cross-check: on the same day AMZN raised capex by $20 billion, partly attributed to memory prices. Memory inflation is an industry-wide event, not an Apple-specific demand story.
• Quantifying the memory hit: the Q4 gross-margin guidance of 47–48% includes roughly 1pp of tariff refund → underlying 46–47%, down 1.1–2.1pp sequentially from this quarter's underlying 48.1%. Management said memory costs will keep rising in Q4, only partly offset by remaining low-cost inventory — implying greater pressure in the first half of FY27 (once the cheap inventory is exhausted).
• Credit where it is due: this quarter's underlying gross margin of 48.1% compares with 46.49% a year earlier, still an improvement of roughly 1.6pp YoY — core profitability has not deteriorated.
• What deserves more worry than memory: Services growth slowing to +12%, which management attributed to currency, a high base from last year's F1 film, weak mobile gaming, and App Store regulatory changes — and the regulatory change is structural and irreversible, a longer-lived variable than the memory cycle.
• Fundamental verification: TTM operating cash flow $146.7 billion, capex only $10.04 billion → FCF $136.7 billion, capex/OCF only 6.8%. This is an extreme contrast with AMZN's >100% and is the most fundamental divide between the two right now (Apple's AI strategy runs asset-light, on external compute). Net income converts to cash at 1.13x; earnings quality is clean. Cash $146.5 billion / debt $84.3 billion → net cash roughly $62.2 billion. Valuation: P/E (TTM) 38.25 (on the 7-30 close of 333.43) / roughly 35.3x on the pre-market 307.91, forward P/E 36.20, P/S 10.49. ⚠️ A reminder on definitions: when debating "whether to buy the pre-market dip", use 35.3x rather than 38.25x — using the valuation from before the −7.65% to argue about the level after the −7.65% is a mismatch. Even at 35.3x, it remains well above Apple's median across most of the past decade (precise percentile: no reliable data available).
• ⚠️ Important correction (the draft cherry-picked one side; now fixed): the draft only wrote that "Greater China at $18.8 billion came in below the $19.6 billion expected", omitting the other half of the same public data set — Greater China revenue was $18.816 billion, up from $15.5 billion to $18.8 billion YoY, growth of +22.4% (MacObserver³⁰ / TelecomLead³¹). The conclusion must be rewritten accordingly: China is an expectation-gap problem, not a share-loss problem. A +22.4% YoY growth rate is strong among Apple's regions; it is simply that the sell-side set the bar higher (roughly $19.6 billion). In a report that places AAPL in the bearish zone, presenting only "below expectations" while hiding "+22.4% YoY" is a materially one-sided narrative, corrected here. (Segment revenue still defers to the 10-Q segment tables.)
• Pre-market and technicals: $307.91, −7.65% (7-30 close 333.43, itself −1.41% that day). The 52-week range is 201.50–344.57, and even after the drop it remains in the upper half. 46 analysts have a consensus "buy" with an average target of roughly 321.73–328.69 — the targets are already below the 7-30 close, so the sell-side was already cautious.
• Final call: "watch only" rather than "short watch". The reasons: net cash of $62.2 billion, FCF of $136.7 billion, underlying gross margin still +1.6pp YoY, and Greater China +22.4% YoY — the fundamentals have not broken; what broke is the near-term margin outlook and an overly high sell-side bar. Shorting a company with cash flow this strong is a poor risk/reward. But equally, this is no time to buy the dip: memory cost pressure runs at least into FY27H1, the Services regulatory factor is irreversible, and roughly 35.3x on the pre-market price still offers no margin of safety. Wait for a margin inflection or valuation digestion.

3美光MUtotal 71, watch closely · wait for a pullback · Micron

• Related news: 07-30, 24/7 Wall St.⁹. Samsung warned that memory supply is tightening; Morgan Stanley expects memory prices to rise at least +25% from Q2 to Q3, with shortages possibly worse in 2027–2028; JPMorgan says there is no meaningful new supply before 2028.
• Catalyst logic (substantially revised after quality review): the three testimonies point the same way — the seller (Samsung) says supply is tight, and two large buyers (Apple's Cook, Amazon's Jassy) admit they are being hurt by the price increases and are therefore cutting guidance / raising capex. ⚠️ But the draft's claim that this was "joint confirmation by three parties with opposing interests, the hardest kind to fake" is a flawed argument and is withdrawn here: a buyer blaming external component costs when results fall short is precisely the highest-incentive, lowest-cost explanation available (far better than admitting weak demand or deteriorating returns on investment), not "the hardest evidence to fake". On this specific claim, all three parties' incentives point the same way and do not constitute independent verification.
• What this chain genuinely lacks: third-party DRAM/NAND contract and spot prices (TrendForce et al.) — the only independent price anchor, not obtained for this report and flagged pending verification. Morgan Stanley's "at least +25% in Q3" and JPMorgan's "no meaningful new supply before 2028" are both second-hand relays via 24/7 Wall St.; the original notes were not seen. Until independent price data is in hand, this theme's strength should be treated as "medium-high" rather than "very high".
• Theme and stage: late in the maturing phase, approaching crowded. +18.36% on 7-30 and another +4.41% this morning, roughly +23.6% over two days.
• Fundamental verification: FY25 revenue $37.38 billion +48.85%, net income $8.54 billion (+997.56% YoY); market cap $987.8 billion. 45 analysts at "strong buy", average target 1,507.38.
• ⚠️ The P/E of 19.74 and the figures above are on incompatible bases (found in quality review, clarified here): market cap $987.8 billion ÷ FY25 net income $8.54 billion = 115.7x, 5.9 times away from 19.74. The reasonable explanation is that the P/E is on a TTM basis while net income is on an FY25 annual-report basis (Micron's FY25 ended in 2025, and profits jumped sharply in the quarters since), but the two cannot be used to corroborate each other in the same sentence. This report did not obtain MU's TTM net income / TTM EPS, so the judgment that "19.74x is the low-P/E trap of a cyclical at peak earnings" cannot be independently verified by readers from the data given here; treat it as a qualitative flag rather than a proven conclusion.
• Pre-market and technicals: $913.20, +4.41% (7-30 close 874.66). 52-week range 103.38–1,255.00 — a more than 12x range within the year, extreme volatility.
• Final call: hardest logic, worst positioning. A P/E of 19.74 is the classic low-P/E trap of a cyclical near peak earnings — the denominator is an EPS pushed to a high by the price cycle, not sustainable earnings. The theme deserves close watching, but +23.6% over two days is not an entry point. Wait for a pullback, or express the same theme through names that are not over-priced, such as AVGO or NVDA.

4博通AVGOtotal 69, priority deep-dive · Broadcom

• Related news (all carry-overs, nothing new today): no earnings of its own; a thematic beneficiary. Partnership support: 07-08, at least $30 billion committed with Apple to co-design and produce US-made custom chips and wireless connectivity technology (expected to deliver over 15 billion US-made chips, CNBC³²); 07-25, an agreement with Samsung worth over $200 billion running through 2030, covering HBM4 memory supply, 2nm foundry and advanced packaging (CNBC³³ / Fortune³⁴). A key qualifier: the Samsung deal is a memorandum of understanding (MOU), not a binding contract; specific volumes, pricing and delivery await follow-on agreements — it must not be treated as locked-in orders.
• Catalyst logic: MSFT and AMZN raised cloud capex at the same time (AMZN to $220 billion), and custom AI ASICs are one of the most direct beneficiary paths; AMZN's ramp of in-house Trainium is also a positive signal for the ASIC ecosystem.
• Fundamental verification: FY25 revenue $63.89 billion +23.87%, net income $23.13 billion (+292.30%); TTM P/E 64.55 but forward P/E only 24.61 — the best value in this table. Dividend 2.60 (0.67%). Average target 527.00.
• Pre-market and technicals: $392.30, +1.15% (7-30 close 387.84, +4.73% that day). 52-week 281.61–495.00, in the upper-middle of the range, no excessive gap.
• The gap between a TTM P/E of 64.55 and a forward 24.61 has been verified and is explainable: the main cause is amortization of acquired intangibles from the $61 billion VMware acquisition completed in November 2023 — that item alone was $1.969 billion in FY26Q1, the largest component of the gap between GAAP and non-GAAP net income (which was $2.836 billion that quarter). This is a non-cash expense that will decline year by year as the amortization period runs, so a forward P/E far below the trailing P/E is structural rather than accounting manipulation. One caveat is warranted, though: non-GAAP also excludes stock-based compensation (SBC), which is a real cost to shareholders, so "24.61x" is still somewhat optimistic and the true center of gravity lies between the two. (Broadcom IR³⁵ / Calcbench breakdown³⁶)
• Final call: priority deep-dive. Within the highest-certainty theme of "AI capex being revised up", AVGO is one of the few names that did not gap sharply pre-market and whose forward valuation holds up after verification. Its main weakness is no catalyst of its own today (the partnership news is a 7-08 / 7-25 carry-over), so today's move will ride entirely on the thematic sentiment AMZN is driving.

5英伟达NVDAtotal 67, priority deep-dive · Nvidia

• Related news: no earnings of its own (8/26 earnings). Chip stocks rebounded on 7-30 on MSFT's strong results.
• Catalyst logic: cloud capex being revised up across the board (MSFT roughly $175 billion for CY2026, AMZN $220 billion) maps directly to GPU orders.
• ⚠️ A correction on the "in-house silicon cannibalization" figure: AMZN's original disclosure was that "the AI business and the in-house silicon business each exceed a $25 billion annualized run rate", and the in-house silicon definition covers both Trainium and Graviton — Graviton is a general-purpose CPU and is not a substitute for GPUs. The draft turned that number directly into "Trainium reached $25 billion" and used it to argue GPU share loss, which inflated the definition; corrected here: Amazon does not separately disclose Trainium's run rate, so the magnitude of cannibalization for NVDA cannot be quantified. Directionally the cannibalization is real (Project Rainier has deployed 500,000 Trainium chips for Anthropic and "will keep adding"), but the magnitude cannot be derived from that $25 billion.
• Fundamental verification: P/E 29.87 — actually not expensive among current AI leaders (versus AMD's 161.86). Average target 302.83.
• Pre-market and technicals: $197.35, +1.18% (7-30 close roughly 195.05). 52-week 164.07–236.54, mid-range and roughly −17% from the high, one of the few AI names in this report whose positioning is not crowded.
• Final call: priority deep-dive. Theme leader, reasonable valuation, positioning not extended, no pre-market gap. The risk is the roughly one-month news vacuum before 8/26 earnings, during which only thematic sentiment can support it, and in-house silicon cannibalization is a real medium-term variable.

6希捷STXtotal 67, watch closely · Seagate

• Related news: after the close on 07-30, FQ4 delivered record revenue and earnings, net income +116.7% YoY, driven by AI infrastructure storage demand; multiple banks raised targets to an average of 1,106.09.
• Catalyst logic: the only name in the storage theme backed by its own earnings — catalyst directness higher than the "industry transmission" of MU / WDC / SNDK.
• Fundamental verification: P/E 61.27, market cap $193.3 billion.
• Pre-market and technicals: $900.50, +5.73% (7-30 close roughly 851.68, +11.41% that day). Roughly +17.8% over two days. 52-week 138.30–1,145.00.
• Final call: watch closely. Delivering on its own results makes it sturdier than SNDK/WDC, but a P/E of 61 plus +17.8% over two days means there is no margin of safety left. HDD's long-term structural position (against SSD) is an unresolved medium-term question mark that this report did not examine in depth.

7谷歌GOOGLtotal 65, priority deep-dive · Google

• Catalyst logic: an AI cloud beneficiary that also supports Anthropic's data-center buildout. Among all the mega-cap tech names in this report, it is the only one not repriced by yesterday's or today's events — only +1.43% pre-market.
• Fundamental verification: P/E 16.74, the lowest valuation among mega-cap tech; TTM revenue $445.87 billion. 64 analysts at "strong buy", average target 427.59.
• Risk note: the YoY increase in TTM net income of $244.12 billion is anomalous, and there is strong suspicion it likewise includes AI investment revaluations and other non-operating income (the same issue as AMZN/MSFT). If so, the "lowest valuation" of 16.74x is equally distorted — this report did not independently verify the item, lists it as the highest-priority to-do, and cautions against using "P/E 16.74" as a reason to buy.
• Pre-market and technicals: $338.43, +1.43% (7-30 close roughly 333.66). 52-week 187.82–408.61.
• Final call: priority deep-dive, but verify income-statement quality first. Its expectation gap is the best in this report (8/10), but whether its valuation appeal is genuine is unsettled.

8雪佛龙CVXtotal 63, watch closely · Chevron

• Related news: 07-31 06:30 ET, BusinessWire¹³. Q2 net income $12.1 billion (EPS 6.11), adjusted EPS 6.06 vs 5.57 expected (beat), the highest in six years; upstream $8.2 billion (+200%), refining $4.9 billion (+500%). Also signed a 20-year West Texas data-center power supply agreement with Microsoft.
• Catalyst logic (corrected after quality review): geopolitics (the US-Iran conflict) is pushing oil higher → a double hit to upstream and refining. The full oil picture (missing from the draft, filled in here): Brent is at 92.27 (07-31 06:45 ET, −0.41% on the day), +24.43% over the past 30 days (from 74.15), roughly +20% for July as a whole, +27.31% YoY; but still about 23.7% below the 52-week intraday high of 120.88 set on 2026-04-30, with a June average of $85. Two facts coexist: oil has rebounded strongly on a monthly scale and is at the same time far below the April peak. The implication for Q3 is neutral-to-positive — quarter-to-date Brent (≈92) is above Q2's June average (85), so a sequential deterioration is not a given, contrary to the intuition that "profits are purely a rear-view mirror".
• Getting the timing of the regulatory factor right (a factual error in the draft, now corrected): Trump's order for the DOJ to investigate gas-station pricing at Exxon/Chevron/Shell/BP occurred in late June 2026, not today. The draft labeled it "announced pre-market 07-31" with an impact level of A+, which is factually wrong — news the market has digested for roughly 5 weeks is not incremental bearish news today, and it has been downgraded to a carry-over factor (B). CNN's 7-31 coverage is background for the earnings.
• The most notable market signal (but note that this is inference, not fact): a clear beat (6.06 vs 5.57) yet only +0.07% pre-market. There are at least four explanations, which this report cannot distinguish pre-market: ① DOJ probe risk offsets the earnings positive; ② the profits from high oil prices were already anticipated (oil prices are publicly visible, so an earnings beat is not genuinely new information); ③ the market expects oil to retrace from its geopolitical premium and therefore will not capitalize cycle-peak profits; ④ pre-market liquidity is thin, and large-cap blue chips are often not fully priced until after the open. "Profits surge but the stock does not move = regulation is already priced in" is one explanation among several and should not be taken as a conclusion. The first 30 minutes of price and volume after the open is the real test.
• Fundamental verification: dividend yield 3.70% (7.12), market cap $380.3 billion, average target 215.00. ⚠️ The P/E of 33.45 rests on a questionable basis and should not be used to judge cheap or expensive: market cap $380.3 billion ÷ 33.45 implies TTM net income of roughly $11.37 billion, which is less than this single quarter's net income of $12.1 billion — suggesting the TTM window likely does not yet include this record quarter, or that market cap and P/E were taken at different points in time. Using a TTM multiple that does not reflect a record quarter to describe a company posting its highest profit in six years will simultaneously overstate the valuation and understate the earnings inflection. The multiple is flagged pending verification and is not used as a basis for any conclusion here.
• Final call: watch closely; relatively favored within energy (versus XOM: beat vs miss). The 20-year power agreement with MSFT is an intersection of the energy and AI data-center themes and may be an underappreciated long-term variable. But the DOJ probe is a sword overhead, and it is unwise to express this in size before the nature of the investigation becomes clear.

9微软MSFTtotal 74, watch only · Microsoft

• Related news: FQ4 results and FQ1 guidance after the close on 7-29 (Azure +45%, RPO +84%). Up +15.51% in a single day on 7-30, adding roughly $450 billion of market cap — a record single-day market-cap gain for an individual stock.
• Why a high score yet "watch only": its fundamentals and guidance quality are the strongest in this report (FY26 revenue $331.8 billion +17.8%, net income +31.3%, P/E 25.13, average target 560.95), but the catalyst was fully realized on 7-30. It is −0.37% pre-market this morning, its expectation-gap component scores only 2/10, and its risk deduction is −9.
• Final call: watch only. A good company ≠ a good entry. Chasing after +15.5% in a day means taking profit-taking risk, not fundamental risk. Reassess if it pulls back toward the gap.

10MetaMETA5th in the bearish zone · watch only

• Related news: continued declines since the 7-29 results, closing 7-30 at 539.03 (−7.95%), near the 52-week low of 520.26; AI data-center lease obligations reportedly reach $279 billion; multiple banks cut targets.
• Catalyst logic: the contrast with AMZN is the core of today's theme 2 — AMZN has $496 billion of backlog proving the spending comes back; META has only $279 billion of lease obligations proving the spending happened. Both are spending; the market's verdict is opposite.
• Fundamentals: P/E 20.31, market cap $1.37 trillion.
• Pre-market and technicals: $544.15, +0.95% — only +0.95% against a broad backdrop of Nasdaq futures +1.11%, a weak rebound that cannot keep up with the market, and it is hugging the 52-week low.
• Final call: watch only. A weak rebound near a 52-week low is unreliable; a turn requires verifiable evidence of capex returns (backlog-style disclosure). The reason it is not listed as a short watch: the valuation has already fallen to 20x and the decline has run for several days, so the odds on a right-side short are poor.

---

6. Bearish / Avoid List

---

7. Within-Theme Rankings

Theme 1: Memory / Storage Super-Cycle

Theme 2: The Return-on-Investment Trial in AI Cloud

Theme 3: AI Compute Hardware

Theme 4: Energy

---

8. Open-Verification Signals

Pre-market (before 09:30 ET)

• The single most important thing to watch: whether AMZN's +11.58% gap is gap-and-go or gap-fill. Estimating AMZN's NDX weight at roughly 4–5%, its contribution to the Nasdaq futures' +1.11% is about +0.5pp (a rough weight × move calculation, not a precise attribution) — the main contributor but not all of it. ⚠️ Wording correction: the draft said "if AMZN fills half its gap within 30 minutes, the entire AI cloud theme's move today is essentially falsified" — using a single stock's 30-minute action to 'falsify' a theme this report itself defines as 'structural, spanning 2026–2028' is a mismatch of time scales. The correct formulation: that price action is only an observation point for the day's sentiment and does not falsify a structural theme.
• The four storage names (MU / SNDK / WDC / STX) have similar pre-market gains (+4.4% to +5.8%). ⚠️ The draft inferred from this that "the same logic must be driving them", which is circular reasoning — using price synchronization to prove a shared fundamental. In fact STX/WDC are HDD (AI nearline storage capacity demand) while MU/SNDK are DRAM/NAND (Samsung's supply tightening and contract price increases); the driver chains differ (see the 1a/1b split in §2). The more reasonable explanation is that money is buying the "storage" label wholesale. So the correct observation runs the other way: if the HDD group and the DRAM/NAND group diverge after the open, that in itself shows the earlier synchronization was label-driven rather than a shared fundamental.
• Whether AAPL's −7.65% attracts dip buyers: AAPL's weight is very large; if the decline narrows quickly after the open, it will meaningfully support the S&P; if it keeps weakening, the divergence between the Dow/S&P and the Nasdaq will widen.

Intraday (first 30 minutes)

• ETF confirmation: QQQ's pre-market +1.08% is materially stronger than SPY's +0.35% — this is an extremely bifurcated tape, not a broad rally. If SPY cannot keep pace with QQQ after the open, the advance is driven entirely by AMZN alone plus semis, which is "lone-runner" risk.
• Whether IWM (+0.34%) follows: small caps are the most sensitive to long-end rates. With the 30Y at a 19-year high, IWM lagging all day would confirm that rate pressure is still at work.
• Volume confirmation: gapping names must draw volume to hold. For names with extreme 52-week ranges like SNDK / STX, a rally on no volume easily reverses.

Macro events (two times today)

• 08:30 ET Employment Cost Index (ECI, Q2) — today's only macro risk event. With the FOMC having just produced "three votes to hike", an above-expectation wage-inflation print would directly reinforce hike pricing and weigh on every risk asset today, especially the already-gapped high-valuation names.
  • Comparable priors (be sure to read the print against these two numbers): Q1 2026 (three months ended March), total compensation costs for all civilian workers +0.9% QoQ, +3.4% YoY; private sector +0.9% QoQ over the same period, of which wages and salaries +0.7% and benefit costs +1.3% (BLS²⁴, released 4-30).
  • ⚠️ Sell-side consensus was not obtained for this report. TradingEconomics' model estimate of roughly +0.40% QoQ sits too far from the 0.9% prior to look like a usable consensus, and this report does not recommend using it to judge a beat or miss. Without a consensus, the practical approach is to read the print's direction and magnitude relative to 0.9%/3.4%, and to watch the bond market's immediate reaction (10Y/30Y and rate futures) rather than equities.
• 10:00 ET final University of Michigan consumer sentiment — less impactful, but the inflation-expectations component is worth a look in the current environment.
• 11:00 ET ETN earnings call — the only real catalyst in the data-center power theme today; the results were released before the open (the figures had not been obtained as of this report's completion).

Options sentiment

• Both AMZN and AAPL face IV crush after earnings: even with the direction right, buying options can lose money as implied volatility collapses. AAPL's decline has released most of the event risk, so its IV may fall faster than AMZN's.
• IV is already rising ahead of earnings for AMD (8/4) and WDC / SNDK (8/5), so options are not a suitable way to express a directional view on these three names before their reports.

Risk signals (the three most worth guarding against in this report)

1. The divergence between rates and equities: Nasdaq futures +1.11% coexisting with the 30Y Treasury at 5.23% (a 19-year high). Historically such a divergence rarely persists — either rates fall back or equities catch down. This is today's most overlooked risk.
2. Reversal after the gap: AMZN +11.58% and the four storage names +4% to +6% are all gap openings. The first hour after a gap-up open is the easiest time of the day to get trapped.
3. "Lone runner, sector doesn't follow": if AMZN rallies hard while GOOGL/MSFT are flat or weaker, this is a single-event move rather than a theme move, and its durability is questionable.

---

9. Final Conclusions

① The 5 Names Most Worth Watching Today

② The 3 Strongest Themes Today

③ What to Avoid Today, and Why

• Avoid every storage name up more than 20% over two days (SNDK +33%, MU +23.6%, WDC +22%): right logic ≠ right level. The theme is correct but already borrowed from the future, and WDC/SNDK report on 8/5.
• Avoid chasing AMZN's 11.58% gap: most of the good news is priced, and the two red flags (FCF turning negative, Q3 operating profit guided down sequentially) have not been digested by pre-market sentiment.
• Avoid buying the AAPL dip: memory cost pressure runs into FY27H1, Services contains irreversible regulatory factors, and a 38x P/E offers no margin of safety. But equally, do not short it — net cash of $62.2 billion and FCF of $136.7 billion mean the fundamentals have not broken.
• Energy: do not add, but there is no need to treat it as bearish either (a clear softening versus the draft). Three corrected facts: ① the DOJ probe is old news from late June, digested for roughly 5 weeks, and is not incremental bearish news today; ② Brent is at 92.27, +24.43% over 30 days and roughly +20% in July, with the quarter-to-date average above Q2's June average of 85 — so the claim that "profits are purely a rear-view mirror" does not hold; ③ but oil is still about 23.7% below the 4-30 high, and XOM did post an EPS miss. Note: do not read "CVX beat but did not move" as "regulation is already priced in" — thin pre-market liquidity, expectations of oil mean-reversion, and profits already being anticipated are all equally reasonable explanations (see §5⑧). Conclusion: within the sector CVX is favored over XOM, but the group lacks an incremental catalyst today; observation is enough.
• Avoid high-valuation names within 48 hours of earnings: AMD (8/4, P/E 161.86), WDC/SNDK (8/5).
• Avoid initiating positions before the 08:30 ET ECI: with three votes for a hike in the background, an above-expectation wage-inflation print would directly weigh on all risk assets.

④ The Final One-Line Judgment

Today is the day "accounting beats" and "real guidance" parted ways: AMZN's and AAPL's quarterly EPS beats were almost entirely illusions manufactured by one-off gains ($53.4 billion of Anthropic revaluation / a $2.19 billion tariff refund), and the market saw through that layer accurately — paying only for AWS's $496 billion backlog and penalizing only Apple's cut gross-margin guidance — while the thread tying the two together is the same thing now reshaping the whole industry's cost structure: memory price increases. But the evidentiary boundary of this main thread must be stated at the same time: the three testimonies on memory price increases (Samsung, Cook, Jassy) point the same way yet their incentives are not independent, and the genuinely independent price anchor (DRAM/NAND contract prices) was not obtained for this report — that is the softest patch of ground beneath today's strongest theme. Thematically, the AI cloud logic is the hardest (backlog is verifiable contracted revenue), but the storage names are already badly positioned and AMZN's gap is already very large, so today's more likely opportunities lie in names not repriced pre-market (AVGO, NVDA, GOOGL) and in ETN, the only name with a real catalyst today and one the market is ignoring, rather than at the top of the gainers list. There are two biggest blind spots: first, the divergence between the 30-year Treasury at 5.23% and Nasdaq futures at +1.11% (and that rate reading is from the prior day; this morning's live value was not obtained); second, the 08:30 ET ECI — after "three votes to hike", that data point will determine today's close more than any earnings report, and this report could not obtain its consensus, so please judge it against the 0.9%/3.4% priors yourself.

---

Appendix: Data-Confidence Statement and Verification Backlog

Methodology statement

• Local yfinance was rate-limited by Yahoo throughout (YFRateLimitError; .history(), .info and .fast_info all failed, repeatedly in live testing), so this report uses no yfinance data; all quotes and valuations come from stockanalysis.com (captured 2026-07-31 08:00–08:05 ET), and all earnings figures come from company press releases / SEC filings / call transcripts.
• The data source's Previous Close field lags by one day for some names, so the back-solve method was used instead (see the note at the top), with three-way cross-validation completed on SNDK / STX / MSFT.

8 items corrected after quality review (the draft was wrong; this version supersedes it)

1. The DOJ probe into the four major oil companies is old news from late June 2026; the draft mislabeled it as "announced pre-market 07-31" with an impact level of A+. It has been changed to a carry-over factor and downgraded to B, and the related causal chain in the energy section has been rewritten throughout.
2. The draft omitted crude quotes entirely while making "oil prices" the causal core of the energy theme. Added: Brent 92.27 (07-31 06:45 ET), +24.43% over 30 days, roughly +20% in July, about 23.7% below the 4-30 high of 120.88.
3. One-sided use of AAPL's Greater China data: the draft wrote only "$18.8 billion, below the $19.6 billion expected", omitting +22.4% YoY ($15.5 billion → $18.8 billion). Now filled in, with the conclusion changed to an expectation-gap problem rather than a share-loss problem.
4. The "three-way cross-validation" claim on the memory theme has been withdrawn: the three parties' incentives are not independent (see §0 item 4). Logic robustness downgraded from "very high" to "medium-high".
5. HDD and DRAM/NAND were wrongly merged into a single theme and have been split into two independent sub-chains, 1a and 1b; §8's circular reasoning that "price synchronization proves a common logic" has been rewritten.
6. Incomplete coverage of today's pre-market earnings: the draft had only XOM/CVX and missed ETN (the only real catalyst in the data-center power theme today) as well as ABBV/CL/CCJ/LEA. Now added, with theme 6 re-anchored on ETN.
7. §1 lacked standalone entries and primary links for MSFT / META, which are the core basis for the scoring and comparative arguments. Added as rows 17 and 18.
8. Other in-place corrections: the over-inference that "CVX beat but did not move = regulation is already priced in", the inflated Trainium/Graviton definition, AAPL's use of a closing-price P/E to argue about a pre-market dip buy, AMZN's ROE not being adjusted for one-offs, the §4 footnote's overstatement of "highest fundamentals in the table", §8's scale mismatch in "falsifying a structural theme in 30 minutes", and the inconsistent convention for 52-week positioning.

5 items that still must be verified (do not draw strong conclusions on top of them)

1. Third-party DRAM/NAND contract and spot prices (TrendForce et al.) — the only truly independent verification source for the "memory price increase" theme, not obtained for this report. Morgan Stanley's "+25%" and JPM's "no new supply before 2028" are both second-hand relays; the original notes were not seen. Highest priority.
2. ETN's actual Q2 results today (not obtained as of 08:35 ET). Today's conclusion for the data-center power theme depends entirely on them; do not presume a beat.
3. The tax treatment of AMZN's $53.4 billion Anthropic gain: this report back-solved using the overall effective tax rate of 22.5%, but unrealized investment gains often carry deferred tax at a potentially different rate. This determines whether adjusted EPS lands at 1.95 or 2.06. Requires checking the 10-Q's "Other income (expense), net" and the income tax footnote (SEC EDGAR CIK 1018724).
4. Whether GOOGL's P/E of 16.74 is contaminated by investment revaluations: the YoY increase in TTM net income of $244.12 billion is anomalous. If it shares the same source as AMZN / MSFT, the "lowest valuation" is distorted. This report has already lowered its confidence in GOOGL's fundamentals accordingly and does not use that P/E as a basis for recommendation.
5. Several valuation multiples rest on unsettled bases: AMZN/AAPL historical percentiles (no multi-year time series), MU's TTM net income (the P/E of 19.74 and FY25 net income are not mutually convertible), and CVX's P/E of 33.45 (the implied TTM net income of $11.37 billion is less than the single quarter's $12.1 billion, so the window is questionable). None of these single-point snapshots can substitute for judgment.

Known limitations of the scoring model (not fixed in this version; to be handled next)

• AAPL's risk deduction of −29 breaches the 0 to −15 range declared in the header, meaning one table mixes two scales; bearish names should get a separate table and formula.
• WDC/AMD/SNDK/VRT in §3A and all six names in §3B have no component breakdown, so their scores lack a reproducible basis — treat them as tier indications only.
• The "tradability" column gives 10 points to seven of the nine names, which is insufficiently discriminating.

Known definitional conflicts (must be cited explicitly; do not mix)

• Two definitions of AMZN's free cash flow: company definition TTM −$7.6 billion; stockanalysis' gross-capex definition −$11.6 billion. Both are correct; the difference comes from the treatment of finance leases and incentive adjustments.
• Two conventions for index moves: futures (benchmarked to the prior day's futures settlement) and ETF pre-market (benchmarked to the prior day's ETF close). This report presents both side by side; they are not interchangeable.
• VIX 18.01 (−12.83%) is a 7-30 13:31 ET intraday reading, not a closing value and not a reading from this morning. This morning's VIX was not obtained.
• The dollar index was not obtained this morning and cannot be confirmed.
• MU's 7-30 gain: this report uses stockanalysis' +18.36%; 24/7 Wall St. reported +15%, and the discrepancy between sources is unresolved.
• 10Y 4.69% / 30Y 5.23% correspond to 7-30 intraday through around the close; live values as of 08:00 this morning were not obtained. The "rates and equities are diverging" call therefore rests on prior-day rate data, and confidence has been reduced accordingly.
• The timestamp on the $3.93/gallon average gasoline price is described inconsistently across sources ("end of June" vs "7-29"), so this report does not cite that figure as evidence for any argument.
• AAPL analyst price targets: three different "averages" appear in this report — 321.73 / 327.81 / 328.69 — drawn from different sources at different capture times. They must not be treated as a range on a single basis; they support only the qualitative fact that "the sell-side average target is already below or close to the 7-30 close".
• A note on the "label dictionary": the labels used in the conclusion columns — priority deep-dive / watch closely / watch only / avoid / short watch — are this report's observation-intensity labels, not trading instructions; phrasings in the text such as "wait for a pullback" or "unwise to express in size" are likewise statements of observation intensity, consistent with the disclaimer at the end, and do not constitute investment advice.

---

⚠️ Risk warning: this list is a pre-market information review and observation exercise only and does not constitute investment advice. US equities carry high volatility and pre-market gap risk, and post-earnings IV crush and guidance reversals are common. Automatically generated content may contain stale information or factual errors; company disclosures and SEC filings prevail, and this must not be used directly as a basis for trading.