US Pre-Market Brief | 2026-07-30 (ET) Thursday

Coverage window: 2026-07-29 16:00 ET regular-session close → 2026-07-30 09:00 ET pre-market. Items tagged "carry-over" in §1 occurred before this window; they are retained because they still dominate today's pricing, but their scores have been given a freshness discount. Pre-market data pull timestamp: 2026-07-30 08:00–08:35 ET (yfinance / Yahoo chart API / stockanalysis.com cross-verified) Basis for percentage moves:

• Single-stock pre-market change = pre-market price ÷ 2026-07-29 regular-session close − 1. (yfinance's previousClose field was generally one day stale in this pull and has been discarded.)
• Index futures percentage changes are on the media convention, based on the prior session's futures settlement price, and are not directly convertible to the cash index close.
• Valuation multiples, distance to moving averages, 52-week percentile: MSFT / META are computed on the pre-market price and labeled as such; MU is computed on the 7-29 close and labeled as such. Each is noted individually; there is no single convention.

---

0. Today in One Sentence

1. Yesterday was a double kill of "hawkish Fed + chip collapse": the Fed held rates by a 9–3 vote, with three regional presidents voting to hike, the 30-year Treasury yield rose to a multi-decade high, the Dow fell 1,153.18 points (-2.19%) to 51,594.14, the S&P -1.52% to 7,316.15, the Nasdaq Composite -1.74% to 24,442.94, and the Nasdaq 100 has confirmed entry into a technical correction (about -11.6% from its June peak).
2. This morning the wind reversed, and the inflation data provided support on the macro side: June PCE cooled across the board — core PCE MoM +0.1%, below the 0.2% consensus (this is the genuine downside surprise); core YoY 3.3%, in line with consensus and below the prior 3.4%; headline YoY 3.7% (prior 4.1%), MoM -0.1%, both in line. Futures recovered across the board: Nasdaq futures +1.49% (27,749.25), S&P futures +0.69% (7,402.25), Dow futures +0.47% (52,010), Russell +0.61%; VIX 19.11 (-7.5%); Asian equities rebounded overnight; gold +1.01%. Dollar index: not obtained in this pull, cannot be confirmed.
3. The strongest catalyst is Microsoft, and it is a genuine "guidance win": FQ1 FY27 guidance of Azure +45% (constant currency) vs. consensus of about 41.4%, total revenue guidance $89.85–90.95 billion (+16~17%); Azure +43% in the quarter (constant currency, 40% last quarter, confirming acceleration), full-year Azure revenue crossed $100 billion for the first time, and commercial remaining performance obligations RPO $678 billion (+84%). Pre-market +9.74%.
4. One quality blemish that must be called out: non-GAAP EPS 4.74 (this measure already excludes OpenAI, but retains $3.2 billion of Anthropic investment gains, roughly +0.35/share). Excluding that, about 4.39, making the true beat against the 4.24 consensus about +3.5%, not the headline +11.8%. The quality of this quarter's EPS is inferior to that of the guidance.
5. The strongest negative is Meta: EPS 6.18 badly below the consensus of about 7.17–7.23, the low end of full-year capex guidance raised to $130–145 billion, quarterly free cash flow collapsing to $784 million (-90.8%), operating margin 43.0% → 30.9%, pre-market -9.94%, and it has already recorded its longest losing streak since going public. Same day, same theme (AI capex), two completely opposite market verdicts — this is the main axis of today's pricing.
6. Driver types: earnings and guidance dominate (MSFT/META/QCOM/F/BE), macro second (PCE cooling across the board vs. a hawkish Fed, 30Y at a multi-decade high), overlaid with geopolitics (US forces struck dozens of targets in Iran overnight; Brent 92.10 +1.5%, WTI 85.23 +0.9%). Tonight's after-close earnings from the AAPL and AMZN duo are the real risk node of the day.

---

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 descending by score (higher score = stronger bullish evidence); 3B is sorted ascending by score (lower score = stronger bearish evidence). The same 50 points means different things in the two tables. ⚠️ Evidence-gap deductions: only MSFT / META / MU — three names — completed fundamentals-analyst + yfinance independent fundamental verification this time. All other names take an evidence-gap deduction, and the "moat and fundamentals" component is capped at 8/15 — no data means no high score. ⚠️ Freshness discount: for names whose catalyst occurred before the coverage window (F, BE, CXMT, SK Hynix), catalyst directness is capped at 12/20. ⚠️ Pre-market liquidity warning: pre-market volume is extremely thin, and Yahoo's pre-market volume field is all zeros, so this report cannot provide any pre-market volume — which means that for every percentage move in the tables below, there is no way to judge how much real absorption sits behind it. One case, CI, has a confirmed cross-source quote conflict.

3A. Positive Direction

3B. Negative/Casualty Direction (sorted by strength of bearish evidence, not by score — see the model's self-disclosure in §4 for why)

---

4. Single-Stock Scoring Model (100 points)

Rule: sum of components + risk deduction (range 0 to −15, strictly capped) + evidence-gap deduction (no fundamental verification: −5; even the earnings content unverified: −8). The "moat and fundamentals" component is capped at 8/15 for unverified names.

Model self-disclosure (two known limitations, stated without gloss):

1. META's score is flattened by the model's range. Its uncapped risk deduction is −39, far beyond the model's −15 range. That means META's 52 points understate the density of its negatives — it should not be read as "safer than QCOM's 49."
2. Therefore §3B is ordered by "strength of bearish evidence" (a qualitative judgment), not by score. The scores are listed only for transparency. Where the model's range cannot express an extreme case, the written conclusions in §5, §6 and §9 take precedence.

---

5. Detailed Analysis of Top Names

5.1 MSFT｜Microsoft (微软)｜78 pts｜Watch closely

• Related news: 07-29 16:05 ET SEC 8-K EX-99.1⁴; 17:30 ET earnings call (CNBC⁵). FY26Q4 revenue $90.007 billion (+17.7%), Intelligent Cloud $39.3 billion (+32%), Azure +43% (constant currency, 40% last quarter, confirming acceleration), Microsoft Cloud $59.3 billion (+27%), commercial RPO $678 billion (+84%), Copilot paid seats over 30 million. Full-year revenue $331.839 billion (+17.8%).
• Catalyst logic (this is a win on guidance, and it is backed by a current guidance number): FQ1 FY27 guidance of Azure +45% (constant currency) vs. a StreetAccount consensus of about 41.4%, total revenue guidance $89.85–90.95 billion (+16~17%). This is the hardest single line in the print — forward guidance beats by about 3.6 percentage points and points to further acceleration (43% → 45%). RPO of $678 billion (= 2.04× FY26 revenue, +84%) is a stock of accumulated backlog, not guidance, but it lends revenue visibility to the guidance. The two are different in nature and need to be read separately.
• The quality blemish that must be called out: the non-GAAP EPS 4.74 measure already excludes the OpenAI impact but retains $3.2 billion of Anthropic investment gains (about +0.35/share). Excluding that, about 4.39, making the true beat against the 4.24 consensus about +3.5%, not the headline +11.8%. The same holds for the full year: stripping out all investment revaluation gains, "clean net income" grew +17.7% YoY, almost exactly matching revenue's +17.8% — revenue and guidance are hard; the "size of the EPS beat" is amplified by non-cash investment gains.
• Theme and stage: the divergence phase of the AI compute theme, with MSFT on the acknowledged side. Over the past 12 months MSFT is -26.8% (on closing prices) while GAAP EPS is +31.6%, compressing the PE from 36.47× to 20.78× — the market has fully priced "capex eroding FCF," and last night was the first verification in the other direction.
• Fundamental verification (yfinance + SEC 8-K, pulled 07-30 08:01 ET):
  • Gross margin 67.20%, down sequentially for four straight quarters, -185bp cumulatively; Q4 operating margin 45.11%, up only +21bp YoY — operating leverage has decayed to zero this quarter (operating profit +18% = revenue +18%).
  • OCF $182.935 billion (+34.4%), cash capex $115.948 billion (+79.6%), FCF $66.987 billion (-6.5%); FCF/net income fell from 0.84× in FY24 to 0.50×.
  • Definitional correction: of the roughly $46.8 billion increase in OCF, the deferred income tax line swung about $22.6 billion YoY (FY26 +$14.189 billion vs. FY25 −$7.056 billion), about half the increase — this is accelerated tax depreciation created by capex, a timing difference that will reverse in future. Excluding it, adjusted OCF grew +17.8% YoY, the same pace as revenue. (Note: the "adjusted FCF" implied by this lands in a range of −33% to −38% depending on how the base period's convention is handled; we cannot pin it to a single figure, hence the range.)
  • Cash + short-term investments $76.843 billion (-18.7%), total debt $40.294 billion, net cash +$36.549 billion (-28.9% YoY); shareholder returns of $48.716 billion already account for 72.7% of FCF.
  • Valuation (on the pre-market price of 428.57): P/E (TTM) 23.88×, P/S 9.59×, EV/EBITDA 16.24× (the low zone of the past 5 years, about the 30th percentile), but EV/FCF 46.97× (the 70th–75th percentile of the past 3 years).
  • Capex definition (partly clarified on the call): CFO Amy Hood said CY2026 capex is about $175 billion (revised down because some finance leases were reclassified as operating leases), that FQ1 quarterly capex will exceed $50 billion, and that FY2027 will still keep cash flow positive. ⚠️ A figure of "FY2027 capex of $255–260 billion" is also circulating, but that number also appears in pre-earnings sell-side previews and we cannot confirm it came from the call, so this report does not use it as evidence.
• Moat/market position: guiding Azure to +45% on a $100 billion+ base, RPO +84%, 30 million paid Copilot seats — the demand-side moat is widening. But the nature of the moat has changed: from capital-light software (capex/revenue 13%, FCF/net income 0.85×) to capital-heavy infrastructure (capex/revenue 34.9%, FCF/net income 0.50×). "I can spend $175 billion a year" is a moat Amazon, Google and Meta can spend into just as well. The non-AI base is shrinking: More Personal Computing -4%, Windows OEM -7%, Xbox content and services -10%.
• Pre-market and technicals: pre-market 428.57 (+9.74%), range 424.00–430.50, pre-market volume unavailable. On the closing price its position in the 52-week range is about the 20th percentile, on the pre-market price about the 38.5th percentile, still -22.8% from the 52-week high of 555.45; even after +9.7% it is still below the 200-day moving average of 434.13 (-1.3%). Earnings are past, so IV crush risk definitely exists.
• Final judgment: watch closely. The evidence on the guidance side is the hardest in this print (Azure 45% vs. 41.4%), while the evidence on the quarter's EPS is amplified (the clean beat is only +3.5%) — the two should not be conflated. The genuinely unresolved question is the marginal ROIC on this $115.9 billion (and CY2026's $175 billion) of incremental capital: the company does not disclose it, public data cannot compute it, and every observable proxy is deteriorating — the capex required per $1 of incremental revenue rose from $1.34 in FY24 to $2.31 in FY26; revenue/net PP&E fell from 1.37× to 1.06×. EV/EBITDA says it is cheap, EV/FCF says it is not, and the difference is capex. The 200-day moving average at 434.13 is today's observable technical level; whether the +9.7% pre-market gap gets filled is the day's most direct verification point.

5.2 META｜Meta｜52 pts (risk deduction is capped; the actual density of negatives is higher)｜Avoid / short watch

• Related news: 07-29 16:05 ET (CNBC⁶). Q2 revenue $60.801 billion (+27.96%), a slight beat, advertising revenue about $59.4 billion; EPS 6.18 vs. consensus of about 7.17–7.23; total expenses $42 billion (+55%), including $2.4 billion legal-related and $1.2 billion of layoffs; quarterly capex $31.078 billion (+83%); FY26 capex guidance raised to $130–145 billion (from $125–145 billion, lifting only the low end); full-year total expense guidance also raised to $165–169 billion; Q3 revenue guidance $61–64 billion.
• Catalyst logic: this is damage from guidance and the cost structure, not from revenue. There is not a crack on the revenue side — ad impressions +14% × price per ad +12%, both up (1.14×1.12−1 = +27.7%, self-consistent with ad revenue growth). Historically, when an ad business is genuinely in trouble the shape is "volume up, price down"; this quarter is nothing like that. All the damage comes from cost and capital: operating margin 43.02% → 30.88% (-12.1pp), or about 36.8% excluding $3.6 billion of one-offs, i.e. roughly half one-off and half structural.
• Theme and stage: the casualty side of the AI capex theme, in a crowded-unwind phase. The 10 straight down days through 07-29 are META's longest losing streak since its 2012 IPO (the prior record was 7 days, computed over the full daily history); if today closes at the pre-market price, it is -22.6% cumulatively from the 7/15 high of 681.31.
• Fundamental verification (yfinance + stockanalysis, pulled 07-30 08:0x ET):
  • Gross margin fell only from 82.13% to 81.37% (-0.76pp) — cloud/compute costs land mainly in R&D and depreciation, not in cost of revenue, so pricing power is undamaged.
  • The segment breakdown overturns the "metaverse drag" intuition: Reality Labs' operating loss was $4.619 billion vs. $4.530 billion a year ago, only +2%, not the reason for the deterioration. Meanwhile Family of Apps operating margin collapsed from 53.0% to 38.7%, and core advertising operating profit fell -6.3% YoY. The margin collapse is 100% inside the business that makes the money.
  • R&D as a share of revenue rose from 27% (Q2'24) to 36%; depreciation and amortization $6.356 billion (+46.4%); stock-based compensation $7.658 billion (+58.4%), 12.6% of revenue and 48% of net income.
  • OCF of $31.862 billion (+24.6%) is still strong — the cash-generating machine is not broken; what is broken is where the cash goes. On the official measure FCF is only $784 million (-90.8%), with FCF margin falling from 19% to 1.3%. This quarter's dividend of $1.353 billion ÷ FCF of $784 million = 1.73×; shareholder returns are no longer paid out of free cash flow.
  • Buybacks have been zero for three straight quarters (Q4'25/Q1'26/Q2'26) while SBC runs at $7.658 billion per quarter — keeping the share count from diluting would require about $7.66 billion of buybacks per quarter, a shortfall of $6.87 billion per quarter.
  • The balance sheet has flipped: cash and marketable securities $90.260 billion, total debt $112.318 billion (+126.6%), net debt of $22.058 billion — five quarters ago it was net cash of +$20.711 billion, a swing of about $42.8 billion. Net new borrowing this quarter was $23.948 billion, with about $53.1 billion issued over the last three quarters. The debt-service cushion is still thick (debt/equity 0.43, interest coverage 27×) — this is not a balance sheet with solvency risk; it is a balance sheet being actively levered up.
  • Valuation (on the pre-market price of 527.25): P/E (TTM) 19.9×, P/S 5.85×, P/B 5.18×, P/FCF 32.7×; annualizing Q2's quarterly FCF, P/FCF ≈ 428×.
• Moat/market position: the demand-side moat is undamaged (volume and price both up, DAP 3.60 billion, ARPP +23.5%). But the cost-side moat is being eroded by its own hand: the business model has been deliberately reshaped from "capital-light + high FCF conversion" (FY24 FCF/net income 0.87×) into a capital-heavy, depreciation-heavy, externally financed form. Even if the AI spend ultimately works, the financial expression of its moat has permanently changed, and the multiples the market is willing to pay (P/FCF, EV/EBITDA) will be re-rated accordingly.
• Pre-market and technicals: pre-market 527.25 (-9.94%), range 526.06–542.55, pre-market volume unavailable. Its position in the 52-week range is about the 2.5th percentile, only +1.3% above the 52-week low of 520.26, and -33.8% from the 52-week high of 796.25; -12.8% below the 50-day moving average and -17.1% below the 200-day. About $148.5 billion of market cap evaporated in a single day.
• Final judgment: avoid. The core contradiction is the collapse in returns on incremental capital: in 2024, operating profit +$22.6 billion against capex of $37.3 billion; in 2025, +$13.9 billion against capex of $69.7 billion; TTM versus FY25, only +$3.65 billion against capex of $89.3 billion — the incremental operating profit per dollar of capex has fallen by an order of magnitude in two years. ROIC fell from 33.4% to 25.8%, not because NOPAT fell ($73.28 billion → $73.02 billion, essentially flat) but because invested capital swelled from $219.5 billion to $283.3 billion (+29.1%). The single most forward-looking number is capex/D&A = 3.93× — the depreciation wave is less than a quarter of the way through. Roughly, for every $25 billion of added annualized depreciation, about $30 billion of incremental revenue is needed just to hold the operating margin flat, equivalent to another 12% of revenue growth spent purely on standing still. Short watch, but extreme care is needed: it has already fallen to within 1.3% of its 52-week low, and oversold-bounce risk is unfriendly to shorts. The next decision point is management's language on 2027 capex — we currently cannot confirm any 2027 figure.

5.3 MU｜Micron (美光)｜56 pts (news level A, single-stock conclusion bearish)｜Watch only

• Related news: 07-27 CXMT's debut closed +465.82% (CNBC⁹); 07-28 SK Hynix operating profit +557% but about 6% below consensus, with 2026 capex raised about 50% to at least $31 billion (Bloomberg⁸); 07-29 QCOM said it would raise product prices because of the memory shortage (CNBC⁷). Financial data come from Micron FQ3-26 SEC 8-K EX-99.1¹⁸ and the FQ3-26 earnings call prepared remarks PDF¹⁹.
• Catalyst logic, corrected (overturning the popular attribution): the main leg down was not caused by CXMT. MU fell only 2.25% on 7/27 (CXMT's debut); the real crash came on 7/28 (-8.85%) and 7/29 (-9.94%) — and the fuse was SK Hynix raising capex 50%. The market is pricing two things at once: China's commodity DRAM expansion + a capex arms race among the existing big three, which is precisely the mechanism that has ended every memory cycle in history.
• Theme and stage: the earnings top + valuation-unwind phase of the memory cycle. July MTD -35.98%, -39.10% from the 6/25 peak close, but still +564.27% over 52 weeks.
• Fundamental verification (Yahoo chart API + SEC 8-K + call transcript, pulled 07-30 08:19–08:30 ET):
  • FQ3-26 (ended 2026-05-31): revenue $41.456 billion (+346% YoY, +74% QoQ), GAAP gross margin 84.6% (non-GAAP 84.9%, a company record), operating margin 80.4%, GAAP EPS 24.67. FQ4 guidance: revenue $50 billion ±$1 billion, gross margin about 86%, EPS 30.73 ±1.00.
  • The quality of the growth is the single most important line in this report: DRAM revenue $31.3 billion (+67% QoQ) — bit shipments grew only low single-digit %, prices rose about 60%+; NAND revenue $9.9 billion (+99% QoQ) — bits mid single-digit %, prices about +85%. This surge is almost 100% price-driven, not volume.
  • Counterintuitive segment data: Cloud Memory, which contains HBM, has a gross margin of 83%, the lowest of the three DRAM segments; the commoditized Mobile & Client and the server-facing Core Data Center are both at 87%. The reason is that HBM long-term contracts lock in prices and do not capture this surge. In this particular cycle, the excess profit is growing precisely on the ground CXMT is attacking head-on.
  • OCF $25.388 billion, capex $7.1 billion, FCF $18.3 billion (a quarterly record); FY26 capex about $27 billion, and FY27 quarterly capex above FQ4's $10 billion (annualizing to $40 billion+; management's exact phrasing "above the mid-$40 billion range" is a second-hand relay and pending verification).
  • Net cash +$24.4 billion (net debt of $5.3 billion a year ago), total debt only $5.7 billion, Debt/EBITDA 0.09, interest coverage 257.6×. Inventory of $8.567 billion is nearly unchanged in absolute terms over a year, while inventory/quarterly revenue fell from 93.8% to 20.7% — this is a genuine shortage, not channel stuffing; the flip side is that the buffer is zero, so once prices turn it will be pure price pass-through, and very fast.
  • SCAs (strategic customer agreements) are the new institutional variable this cycle (source: the call prepared remarks PDF above): 16 of them, typically 5-year, take-or-pay, covering about 20% of DRAM volume / one-third of NAND volume, with RPO of about $100 billion; the floor prices correspond to gross margins "well above the peak quarterly gross margin of any prior cycle of ours" (their words). But the price ceiling on about 40% of revenue is pinned to CQ2-26 market prices, capping the upside at the same time; customer prepayments of $22 billion (including about $18 billion of cash deposits) run through financing cash flow and must be returned in future — FQ4's "net cash" will be materially flattered by those deposits, and it is the line most likely to be misread in the next print.
  • Valuation — PE lies, P/B does not (computed on the 7-29 close of 739.00): P/E (TTM) 16.68×, forward P/E (FY27) 5.15×, EV/EBITDA 11.88×, all of which look extremely cheap; but P/B is 8.28×, whereas at the last earnings peak (FY2022) P/B was only 1.26× and the PE only 7.39×. Memory stocks' book value does not collapse the way EPS does through a cycle, so P/B is the most cycle-neutral yardstick, and it shows the current valuation sitting at an absolute historical high rather than a low. For reference: FY2022 peak EPS 7.75 → FY2023 a loss of EPS -5.34 with a gross margin of -9.1%, from peak to loss inside a year.
• Pre-market and technicals: pre-market 772.40 (+4.52%) (767.11 at 08:19; both timestamps are shown side by side to illustrate the jumpiness of thin trading), pre-market volume unavailable. On the 739 close, it is -22.9% below the 50-day moving average of 958.50 and +44.5% above the 200-day of 511.49 — the space between the two moving averages is an enormous vacuum. Beta 2.14; 7/29 volume 68.85 million shares vs. a 20-day average of 46.36 million (a decline on rising volume). Short interest 36.21 million shares (+14.3% sequentially) but only 3.21% of float, with days-to-cover 0.69 — the fuel for a short squeeze is very thin, and the +4.52% pre-market is more likely an oversold bounce than short covering.
• Final judgment: watch only. Every diagnostic feature of the classic "peak earnings + low PE" trap is now present: earnings at a historical extreme, earnings driven entirely by price rather than volume, an extremely low forward PE with an extremely high P/B, a supply arms race already under way, the second derivative of the price-increase slope turning negative (60%+ → TrendForce's Q3 guidance of 13–18%), and the share price topping out while earnings are still rising. Three genuine mitigants (the SCA contract floor, a fortress balance sheet, and a genuine shortage with zero inventory buffer) will make the depth of the downturn shallower than 2018/2023, but the trigger mechanism has not been eliminated at all — 80% of DRAM volume is still at market prices, the company itself says unit costs will rise, and FY27 brings $40 billion+ of capex depreciation behind it. The market is paying 8.3× P/B for this earnings peak, not 1.3× P/B — that is the real asymmetry. To be emphasized: the above positioning of where we are in the cycle is a judgment, not a fact.

5.4 BE｜Bloom Energy｜55 pts｜Watch closely

• Related news: 07-28 after the close (carry-over), Q2 revenue $1.065 billion (+166%), driven mainly by product sales to data center customers; FY26 revenue guidance raised to $3.9–4.2 billion (Investing.com earnings call transcript¹³).
• Catalyst logic: its own earnings + guidance, both up. More importantly, its logic is decoupled from the AI compute return-on-capital dispute — whether or not MSFT/META can earn back their capex, the power bottleneck at data centers is a precondition, and the money has to be spent here first.
• Theme and stage: the incubation phase of the data center power theme (not the start, not yet crowded).
• Fundamental verification: ⚠️ Not verified. Both revenue +166% and the guidance raise are the company's own disclosed figures; gross margin, free cash flow, net cash/debt and valuation percentiles all cannot be confirmed for now. The fuel cell business has historically had earnings-quality and subsidy-dependence issues, so please do not infer earnings quality directly from "revenue +166%." This is the largest evidence gap in this entry, and the reason "moat and fundamentals" gets only 8/15 plus a −5 evidence-gap deduction.
• Pre-market and technicals: pre-market 177.00 (+8.09%) (7/29 close 163.75), pre-market volume unavailable. Already gapped +8%, so gap-fill risk is high.
• Final judgment: watch closely. The logic is one of the cleanest among today's positives, but no independent fundamental data at all + already +8% pre-market — the two together do not support a stronger label. Whether the gap fills and whether volume absorbs it are today's verification points.

5.5 AMZN｜Amazon (亚马逊)｜54 pts｜Watch only

• Related news: reports Q2 2026 tonight at 17:00 ET. Consensus revenue about $197 billion (+about 17%), adjusted EPS 2.26; AWS growth expected to accelerate from last quarter's 28% to 32%; options price in an earnings-day move of about ±6.3%, above the recent single-day average of 5.4% (Seeking Alpha²⁰ / Amazon IR¹⁷).
• Catalyst logic: MSFT's FQ1 Azure guidance of +45% raised the bar for AWS, it did not lower it. That is tonight's key game structure — MSFT has already proved cloud demand is real, so if AWS only manages 30–31%, the market's attribution will lean toward "AWS is losing share" rather than "industry demand is weak." ⚠️ Note: this report obtained no cloud market share data at all; "losing share" is an inference, not a fact. At the same time AMZN's capex is equally aggressive, and META's script (revenue on target, capex over target, FCF collapse → -10%) can be replicated on AMZN in full.
• Fundamental verification: ⚠️ Not verified. AWS segment margin, absolute capex, FCF and valuation percentiles all cannot be confirmed for now.
• Pre-market and technicals: pre-market 233.33 (+2.95%) (7/29 close 226.65), pre-market volume unavailable.
• Final judgment: watch only. Three verifiable observation thresholds: ① is AWS growth ≥32%; ② is capex guidance revised up as at MSFT/META; ③ is FCF eaten by capex. Implied volatility of ±6.3% means tonight is pure event risk, and a directional call before earnings does not meet this list's evidence standard.

5.6 QCOM｜Qualcomm (高通)｜49 pts｜Avoid

• Related news: 07-29 16:30 ET (CNBC⁷). FQ3 revenue $9.95 billion beat the $9.67 billion estimate but was -4% YoY, adjusted EPS 2.21, a slight miss vs. 2.23; FQ4 adjusted EPS guidance 2.05–2.25, midpoint 2.15, far below the 2.35 estimate (-8.5%); revenue guidance $9.7–10.5 billion, midpoint $10.1 billion, slightly above consensus. Segments: handset revenue $5.09 billion, -20% YoY; autos $1.59 billion (+61%, 23 straight quarters of double-digit growth); IoT $1.83 billion (+9%). Management said it will raise product prices because of the memory shortage.
• Catalyst logic: a textbook "revenue beat, guidance miss" — exactly the case iron rule 2 of this list requires be called out. The revenue beat means nothing here, because the FQ4 EPS guidance midpoint is 8.5% below consensus while the revenue guidance is slightly above it — subtracting one from the other implies margins are being compressed. This is the other side of the same coin as "we will raise prices because of the memory shortage": Qualcomm's gross margin is being eaten by upstream memory price increases, and its own price increase is a forced attempt to pass that through, not evidence of pricing power.
• Theme and stage: the casualty side of the memory price-increase cycle — this is today's most easily overlooked transmission chain: part of MU's excess profit is being taken straight out of the gross margin of downstream players like QCOM.
• Fundamental verification: ⚠️ Not verified. The gross margin trend, net cash and valuation percentiles cannot be confirmed for now. Segment data are the company's disclosed figures.
• Pre-market and technicals: pre-market 147.19 (-5.45%) (7/29 close 155.68), pre-market volume unavailable.
• Final judgment: avoid. The -20% YoY in the handset base business is a structural problem, and autos +61% is off too small a base ($1.59 billion vs. handsets' $5.09 billion) to offset it. Shorting requires care — it is already down 5.45%, and autos/IoT together at +28% provide some support.

5.7 AAPL｜Apple (苹果)｜42 pts｜Watch only

• Related news: reports FY26Q3 tonight at 17:00 ET. Consensus (about 27–31 firms) revenue about $108.86 billion (+16%), diluted EPS 1.89 (+20%); iPhone revenue expected at about $53.97–54.8 billion (+21~23%). This is Tim Cook's last earnings call — hardware engineering chief John Ternus takes over as CEO on September 1. The options market implies a market-cap swing of about $170 billion (AppleInsider¹⁶ / Benzinga²¹).
• Catalyst logic: three independent uncertainties stacked into one evening — ① the results and guidance themselves; ② the strategic language around the CEO handover; ③ the impact of memory price increases on BOM cost. Apple is one of the world's largest buyers of DRAM/NAND, and against the current memory shortage its language on memory costs on the call is worth listening to. (There are also circulating reports that "Apple is testing CXMT memory," but we did not obtain a reliable source and do not use it as evidence.)
• Fundamental verification: ⚠️ Not verified. Gross margin, the services share of revenue, net cash and valuation percentiles cannot be confirmed for now.
• Pre-market and technicals: pre-market 336.19 (-0.59%) (7/29 close 338.19), pre-market volume unavailable — falling slightly against the tape with Nasdaq futures +1.49% is a relative-weakness signal worth noting this morning.
• Final judgment: watch only. If guidance is soft, AAPL could become the third case of a "guidance kill" this week (after META and QCOM); conversely, if iPhone and services are both strong, it is the strongest boost to a Nasdaq rebound. The implied $170 billion market-cap swing leaves a directional pre-earnings call without an evidentiary basis.

5.8 GOOGL｜Alphabet｜40 pts｜Watch closely

• Related news: no new catalyst of its own. Caught up in the sentiment around AI capex return-on-capital doubts at META.
• Catalyst logic: it belongs to the "heavy capex camp" and is filed on the to-be-questioned side of today's thematic split. But the key difference from META is that Google Cloud has verifiable external revenue like Azure's, which in theory puts it closer to MSFT's position than META's.
• Fundamental verification: ⚠️ Not verified, cannot be confirmed for now.
• Pre-market and technicals: pre-market 334.80 (-0.57%), pre-market volume unavailable — falling against the tape with Nasdaq futures +1.49%, making it, along with AAPL, one of this morning's relatively weak mega-cap techs.
• Final judgment: watch closely. This morning's relative weakness says the market is provisionally putting it on the "capex on trial" side rather than the "MSFT side." Once AMZN's cloud segment numbers land tonight, the relative pricing of the three cloud giants will be reshuffled.

5.9 TSM｜TSMC (台积电)｜47 pts｜Watch closely

• Related news: no new catalyst of its own. MSFT's FQ1 quarterly capex above $50 billion + Micron's FY27 capex of $40 billion+ + SK Hynix's $31 billion — a substantial part of these three capex budgets ultimately becomes orders for TSMC and the equipment makers.
• Catalyst logic: this is the one uncontested beneficiary logic inside the "capex dispute" — whatever the return on the hyperscalers' AI investment, as long as they keep spending, foundry and equipment get paid first. The moat is the hardest in the semiconductor supply chain.
• Fundamental verification: ⚠️ Not verified, cannot be confirmed for now.
• Pre-market and technicals: pre-market 386.51 (+3.16%), following SMH's +3.13% sector bounce, pre-market volume unavailable.
• Final judgment: watch closely. Today's +3.16% is a sector oversold bounce, not its own catalyst, and needs confirmation via SMH co-movement and volume. If the SOX rebound lacks volume absorption, this entry does not hold.

5.10 F｜Ford (福特)｜52 pts｜Watch closely

• Related news: 07-29 pre-market (carry-over), Q2 adjusted EPS 0.42 vs. est. 0.35, revenue $48.3 billion vs. $45.8 billion; FY26 adjusted EBIT guidance raised from $8.5–10.5 billion to $10–11 billion, adjusted FCF raised to $6–7 billion (an extra $1 billion of FCF with the capex plan unchanged); $1 billion of material and warranty cost reductions on track. Citi upgraded the rating and Jefferies raised its price target (third-party ratings are listed as news only and are not counted in this report's catalyst-directness score).
• Catalyst logic: guidance up + FCF up, and "capex unchanged while FCF rises" is the highest-quality kind of raise (it comes from cost and production efficiency, not from spending more). In an environment where money is leaving the AI theme, low-valuation value names with improving cash flow are a natural landing spot for rotation.
• Theme and stage: the delivery phase for traditional cyclical value — the move was mostly realized on 7/29 (about +6% that day).
• Fundamental verification: ⚠️ Not verified. The EBIT and FCF guidance ranges are company disclosures; gross margin, net debt (note the treatment of Ford Credit's financial liabilities) and valuation percentiles cannot be confirmed for now.
• Pre-market and technicals: pre-market 15.36 (+0.52%), pre-market volume unavailable — it has barely followed through pre-market today; the momentum was released yesterday.
• Final judgment: watch closely. The news quality is high but the expectation gap was digested yesterday, and today's +0.5% pre-market says incremental buying is limited. Track it as a "rotation candidate once money leaves AI," not as a same-day catalyst name. Also note: the Dow fell 2.19% yesterday, and value stocks did not actually work as a haven.

---

6. Negative/Avoid List

---

7. Within-Theme Rankings

Theme 1: The "return-on-capital trial" splits AI capex

Theme 2: Memory super-cycle vs. China's commodity DRAM expansion

Theme 3: A hawkish Fed meets cooling inflation

Note: this theme is expressed mainly through indexes/ETFs. We verified no rate-sensitive single stocks, so none are listed.

Theme 4: Data center power and its supply chain

Note: PWR/GEV/VRT are beneficiary paths on the theme's logic. We obtained no news or fundamental data on any of them today; they are listed only to mark the transmission path and do not constitute a strength judgment.

---

8. Opening Verification Signals

Pre-market (09:00–09:30 ET)

• The single most important item: whether MSFT can hold the 200-day moving average of 434.13. At 428.57 pre-market it is still -1.3% below it. If it pushes above 434 at the open and holds, that is technical confirmation that "the de-rating is over"; if it stalls before 434 and rolls over, last night's +9.7% looks more like a one-off repricing after earnings than a trend reversal.
• Is META gap-and-go or gap-fill: pre-market 527.25, only +1.3% above the 52-week low of 520.26. A break below 520.26 at the open means trend continuation (extending the losing-streak record further); a quick recovery above 540 means -9.94% was already the bottom of an emotional flush.
• Does the semiconductor bounce have volume: SMH +3.13% pre-market, MU +4.52%, SNDK +7.13%, AMD +4.58%, TSM +3.16%. A first bounce after moves of this size very easily runs without volume. It must be confirmed with volume, not with the percentage gain — and since this report cannot obtain pre-market volume (see below), this item can only be executed after the open.
• The counter-trend weakness in AAPL and GOOGL: against Nasdaq futures +1.49%, the two are -0.59% and -0.57% pre-market. If this divergence persists after the open, it says money inside mega-cap tech is concentrating into MSFT alone rather than lifting the group — which would markedly reduce the rebound's staying power.

⚠️ The most important execution constraint in this section

Yahoo's pre-market volume field is all zeros, so this report cannot provide pre-market volume for any name. That means the core claim "confirm with volume, not with the percentage move" cannot be executed pre-market, and must instead become observing actual volume in the first 15–30 minutes after the open. For every pre-market percentage move (including MSFT +9.7%, BE +8.1%, SNDK +7.1%) there is no way to judge how much real absorption sits behind it — this is the single most important gap in this report.

Intraday (09:30–10:00 ET)

• The first 30 minutes' trend: focus on whether money is "MSFT rising alone" or "the AI compute group rising together." A lone runner = poor rebound quality.
• Sector ETF co-movement: SMH (semis), QQQ (pre-market +1.48%), XLE (-0.65%, diverging from oil), IWM (Russell futures +0.61%, watch whether hike pricing eases).
• Volume confirmation: MU traded 68.85 million shares yesterday vs. a 20-day average of 46.36 million (a decline on rising volume). If today's bounce comes on volume markedly below yesterday's, it is a textbook volume-less bounce.

Options sentiment

• VIX 19.11 (-7.5%), back from yesterday's 20.63 — the panic is easing but has not gone.
• IV crush risk: MSFT, META and QCOM are all past earnings, so pre-market option prices contain implied volatility that is about to collapse, and the risk/reward of buying options for direction is markedly worse today.
• Implied moves for tonight's two prints: AMZN about ±6.3% (above the recent average of 5.4%); AAPL an implied market-cap swing of about $170 billion.
• MU's short interest is only 3.21% of float with days-to-cover of 0.69 — the fuel for a short squeeze is very thin; do not count on short covering to drive the move.

Risk signals (each one reduces the credibility of today's rebound)

1. Reversal after a gap: MSFT +9.7%, BE +8.1%, SNDK +7.1% are all large gaps, and with thin pre-market trading, gap fills are the norm rather than the exception.
2. A lone runner with no group follow-through: if MSFT is strong while GOOGL/AAPL/AMZN weaken, the AI theme has not repaired as a whole.
3. Divergence between index futures and Treasuries: futures are rebounding across the board, but the 30Y is still at a multi-year high of about 5.24%. One month of cooling in June PCE cannot offset the fact of years of above-target inflation + three presidents voting to hike. If the 30Y keeps rising while indexes still rally, that divergence is unsustainable.
4. Divergence between oil and energy stocks: Brent 92.10 (+1.5%) but XLE -0.65%, XOM -0.94%, CVX -0.72% — one of today's most anomalous signals.
5. Disagreement on the 10Y yield: we pulled about 4.62%, TradingEconomics reports 4.70%. Please re-check against your own terminal before the open.
6. Tonight's 17:00 ET AAPL + AMZN: any overnight position held today is exposed to two prints that could replicate META's script.
7. A self-reminder on causality: this report attributes the macro pivot of this morning's rebound to cooling PCE, but the futures quote pull window (08:00–08:35 ET) straddles the 08:30 data release, so we cannot strictly separate "a rebound after PCE" from "a rebound driven by the MSFT-specific event." The dispersion of Nasdaq futures +1.49% versus the Dow's +0.47% looks more like an MSFT-specific event than a broad macro rally — please do not conflate the two.

---

9. Final Conclusions

① The 5 Names Most Worth Watching Today

② Today's Three Strongest Themes

③ What to Avoid Today, and Why

1. Avoid chasing any name with a large gap (MSFT +9.7%, BE +8.1%, SNDK +7.1%). Pre-market trading is extremely thin, and this report has empirically hit one quote conflict (CI) and cannot obtain any pre-market volume — this is not a theoretical risk.
2. Avoid "revenue beat, guidance miss" names: QCOM is today's specimen (revenue guidance actually above consensus while EPS guidance is 8.5% lower, which amounts to admitting margins are being compressed). Guidance always matters more than the current quarter's beat.
3. Avoid treating a current-quarter EPS beat as a positive by itself: MSFT is the cautionary tale inside a positive case — the 4.74 non-GAAP EPS includes $3.2 billion of non-cash Anthropic investment gains, so the clean beat is only +3.5%. What is genuinely worth buying is the guidance, not the quarter's EPS.
4. Avoid reading "memory price increases" simply as a positive: part of MU's excess profit is being taken straight out of the gross margin of downstream players like QCOM and AAPL. The same piece of news points in opposite directions up and down the supply chain.
5. Avoid refiners (VLO): rising oil compresses crack spreads, and refining is the casualty side rather than the beneficiary.
6. Avoid forming a directional view ahead of AAPL / AMZN earnings: implied moves of ±6.3% and a $170 billion market-cap swing, and META's script is already in front of us.
7. Avoid high-dividend substitute assets (MO and the like): with the 30Y at a multi-year high of about 5.24%, high dividends are losing appeal relative to the risk-free rate.
8. Avoid names inferred purely back from the pre-market price (MA / REGN / STLA / MO): this report has withdrawn the scores for STLA and MO because the earnings details were unverified, and MA and REGN are listed only as "pending verification." Using a price you have yourself judged unreliable to generate a conclusion is circular reasoning.
9. Beware post-earnings IV crush: MSFT / META / QCOM are all past earnings, and the risk/reward of buying options for direction today is markedly worse.

④ The Final One-Sentence Judgment

Today is not a "rebound day," it is "day one of pricing the AI capex split" — June PCE cooling across the board (core MoM +0.1% below consensus, headline YoY down to 3.7%) gave the macro side room to breathe, MSFT used forward guidance of Azure +45% to prove that "capex with contracts and guidance behind it" can be repriced, and META used FCF -90.8% to prove that "capex with no verifiable return" will be punished without mercy; but the 30-year Treasury is still at a multi-year high of about 5.24%, three Fed presidents voted to hike, oil is up while energy stocks are down, AAPL and GOOGL closed lower against a Nasdaq futures tape up +1.5%, and the dispersion of Nasdaq futures +1.49% far above the Dow's +0.47% says this looks more like an MSFT-specific move than a broad sector rally — every one of these divergences says the rebound's foundation is not solid, and tonight's 17:00 ET AAPL and AMZN prints will decide whether this is a repair of the theme or a convergence of it. There are only three executable things today: wait to see whether the gaps fill, confirm the sector with post-open volume rather than pre-market percentage moves, and read "guidance" and "the current quarter's EPS" separately.

---

Appendix: This Report's Evidence Gaps (must be labeled honestly)

1. Only three names — MSFT, META, MU — completed fundamentals-analyst + yfinance independent fundamental verification. AMZN, AAPL, BE, F, QCOM, TSM, NVDA, AVGO, GOOGL, MA, REGN, CI, SNDK, AMD, WDC, STX, PWR, GEV, VRT are all unverified; their revenue/gross margin/cash flow/valuation percentiles are uniformly labeled "cannot be confirmed for now" in this report, and they have taken evidence-gap deductions with the "moat and fundamentals" component capped at 8/15. Only 3 of 22 names were verified; please discount every score accordingly as you read.
2. STLA and MO have had their scores withdrawn (only a price reaction, earnings content entirely unverified); MA's and REGN's levels are mainly inferred back from the pre-market price and are labeled "pending verification."
3. CI has a cross-source pre-market quote conflict: we pulled 284.04 (-4.19%) at 08:0x, while media report a beat + raised guidance and +1.65%. The two quotes are from different times (the media's timestamp is unknown), and sign reversals in thin pre-market trading are normal, so the opening trade price must govern.
4. Pre-market volume is unavailable across the board (Yahoo's pre-market volume field is 0). This is the report's single most important gap — the report's core actionable claim is "confirm with volume, not with the percentage move," and that claim cannot be executed pre-market.
5. MSFT's "FY2027 capex of $255–260 billion" was not adopted as evidence: that figure also appears in pre-earnings sell-side previews and we cannot confirm it came from the 7/29 call. This report uses only what is confirmable from the call: CY2026 about $175 billion (including the finance-lease reclassification adjustment), FQ1 quarterly capex above $50 billion, and FY2027 keeping cash flow positive. There is also a claim of "building depreciable life 15 → 25 years," which is second-hand and not adopted.
6. MSFT's balance sheet has about $83 billion in total of unexplained moving line items (other current assets +$22.9 billion, equity and other investments +$20.9 billion, other long-term liabilities +$19.9 billion, investing activities Other net -$19.9 billion); the 8-K has no notes, so the FY26 10-K must be awaited.
7. MSFT's adjusted FCF can only be given as a range (−33% to −38%), because the base period's deferred-tax convention can be handled in different ways and we cannot pin it to a single figure.
8. There is a major cross-source conflict on CXMT's market cap: search summaries report about $85.5 billion, while another calculation (IPO price ¥8.66 → close ¥49.00, +465.82%) gives about ¥3.28 trillion ≈ $484.5 billion. The two differ by about 5.7×, and we cannot determine which is right, so the body text uses only the two more reliable figures — "raised about $8.56 billion" and "+465.82% on debut" — and uses no CXMT market-cap figure at all.
9. MU does not separately disclose the absolute quarterly HBM revenue, so "how much revenue the HBM moat insulates" cannot be quantified; it can only be inferred that non-HBM is about three-quarters. MU's "FY27 capex above the mid-$40 billion range" is a second-hand relay and pending verification.
10. Sources disagree on the 10Y Treasury yield: we pulled about 4.62%, TradingEconomics reports 4.70%. The 30Y at about 5.24% is the media figure; "highest since 2007" has not been verified against a primary data source, so this report uses "multi-year high" instead.
11. The following figures carrying precise percentages were not obtained from primary sources and serve only as qualitative reference: SOX July MTD about -26.7%, the Nasdaq 100 about -11.6% from its June peak, SNDK July MTD about -55%, the 7/29 closing levels of the four major indexes, the specific drawdowns for WDC/STX, and MO's dividend yield of about 5.8%.
12. Share data are entirely missing: this report contains statements such as "is AWS losing share" and "CXMT/Hynix change the supply landscape," but no current-period cloud market share or DRAM share figures were obtained, so all such statements are inferences rather than facts.
13. The dollar index and the specific overnight moves in Asia and Europe were not obtained (only the qualitative description "Asian equities rebounded overnight" is known).
14. A limitation on causal attribution: the futures quote pull window (08:00–08:35 ET) straddles the 08:30 PCE release, so we cannot strictly prove this morning's rebound was triggered by PCE. Nasdaq futures +1.49% is markedly stronger than the Dow's +0.47%, and that dispersion points more toward an MSFT-specific event.

---

⚠️ Risk warning: this list is a pre-market information review and watch list only and does not constitute investment advice. US equities carry high volatility and pre-market gap risk, and after earnings there is IV crush and the risk of guidance reversals; automatically generated content may contain stale information or factual errors. Company disclosures and SEC filings govern, and this must not be used directly as a basis for trading.