US Pre-Market Brief | 2026-08-03 (ET) Monday

Coverage window: 2026-07-31 16:00 ET regular-session close → 2026-08-03 08:30 ET. Includes Friday after-hours, the weekend/overnight (Asia-Europe) sessions, and this morning's pre-market. Single-stock price snapshots are timestamped 08:00–08:07 ET.

---

Data Sources and Methodology Statement (please read first)

⚠️ Local yfinance has now been rate-limited by Yahoo for a third consecutive trading day (YFRateLimitError; direct calls to query1.finance.yahoo.com return HTTP 429, confirmed by live test this session). This is now the third consecutive trading day, which makes it a stable failure of the local environment rather than an intermittent glitch.

• Backup channels that also failed live testing this session: stooq.com (switched to JS proof-of-work anti-scraping), stockanalysis.com via direct curl (returns empty), macrotrends (403).
• Working channels: WebFetch against stockanalysis.com pages (all US equity prices and valuations in this brief); company IR primary filings and SEC EDGAR originals (the cross-verification source for all key financial figures in this brief); akshare (A-share storage-sector data, not a Yahoo channel, working).
• One point of honest disclosure: among this brief's four sub-agents, one did successfully retrieve some yfinance data at a later hour (and it matched stockanalysis on a like-for-like basis). This shows the rate-limiting is intermittent rather than a permanent block, but it cannot be relied on.

Price self-consistency check: for all 20 names, "pre-market price ÷ Friday close − 1" was recomputed one by one against the pre-market change quoted on the page, and all 20 lined up (e.g. ATKR 93.50/72.96−1 = +28.15% ✓; MU 781.78/823.03−1 = −5.01% ✓).

All key financial figures have been checked back to primary sources: Micron FQ3'26 prepared remarks (2026-06-24), Western Digital Q3FY26 deck, SanDisk Q3FY26 press release, Apple FQ3'26 8-K, Palantir Q1'26 8-K, BMY FY2025 annual report PDF, AstraZeneca H1'26 announcement.

Sources excluded / downgraded:

• The ts2.tech aggregator page mixes content from multiple days and contains an obvious error (it says the probability of a September Fed rate hike is 64.5% — what the market is pricing is a rate cut). Not used.
• Fortune and ts2 both report Brent $87.38 (up 71 cents from yesterday morning), which is directionally opposite to the "Brent ≈ $83.3–83.7, down 4.8–5.2%" reported by three independent sources (CNBC / Benzinga / Reuters). Given that the pre-market reaction in energy stocks, airlines, and Treasury yields is uniformly consistent with a sharp oil selloff, we judge $87.38 to be a stale or erroneous print and do not use it.
• The "Recent news" narrative text attached to single-stock pages suffers from staleness (the JETS page is still talking about "Middle East conflict pushing up jet fuel prices"; the DAL page is still talking about "rising fuel costs"). We use only the price fields, not the narrative text.

One important accounting correction (it changes the valuation call): WDC's GAAP net income is severely distorted by the fair-value gain on its retained SanDisk stake (FQ3'26 other income +$2,169M, of which the SanDisk equity gain is +$2,734M). Consequently WDC's TTM P/E of 29.84, net margin of 96.04%, and ROE of 85.92% are all meaningless; any analysis of WDC must use non-GAAP / operating-income measures. This brief has been prepared on that basis.

---

0. Today in One Line

1. Two opposing forces are colliding ahead of the open — this is not a unified risk-on. On one side, Middle East de-escalation → oil down 5% driving a broad cyclical rally (Dow futures +1.05%, Russell 2000 futures +0.76%, airlines JETS +1.63%, healthcare XLV +1.08%); on the other, the ongoing de-rating of memory and semis (SMH −2.32%, MU −5.01%, SNDK −4.92%, WDC −4.38%, STX −3.90%, AMD −3.00%). The result: Nasdaq futures up only +0.15% while Dow futures are up +1.05% — a 7× gap. Today's keyword is not "rally," it is "rotation."

2. First we must correct an error nearly every outlet is making today: the memory selloff did not start today. The actual peak dates were WDC 6/18, STX 6/22, MU and SNDK 6/25 — they topped out one after another within six trading days and have now been falling for about six weeks: MU −35.6% from its peak close, SNDK −50.5%, WDC −30.2%, STX −24.7%, SK Hynix −46.3%. Today's −5% is not "the first red candle at the top," it is another red candle in week six of a bear market. Today's CXMT news is a marginal increment within an ongoing move, not the starting point — but it does carry real substance (the same Reuters piece states CXMT's capacity under construction will more than double, to roughly 600,000 wafers/month), and should not be underrated in the opposite direction either.

3. A piece of evidence that is easily misread today: Chinese memory stocks fell harder than the US ones, yet the actual subject of the news rose. 兆易创新 (GigaDevice) −10.00% (limit-down), 德明利 (Demingli) −9.56%, 佰维存储 (Biwin Storage) −8.49%, 江波龙 (Longsys) −7.84%; on the same day SK Hynix −8.79%, Samsung Electronics −8.76%. But it must be stated alongside that: CXMT's listed entity 长鑫科技 (ChangXin Technology, 688825) closed at RMB 55.04, +1.98%, with turnover above RMB 20 billion. ⚠️ Therefore one cannot conclude from this that "the market has ruled out share transfer" — all four of those A-shares are NOR/MCU makers or memory-module makers, and not one of them is a share beneficiary of this capacity-expansion news; using them as counter-evidence is picking the wrong sample. "Price cycle has peaked" and "share is shifting to a new leader" cannot be distinguished on today's tape, and both may be true at once. Two additional mechanical explanations that have nothing to do with fundamentals also remain unexcluded: ① 8-3 is the first A-share session to digest Friday's −5% to −6% US decline (time-zone catch-up); ② the world's largest memory ETF (Roundhill DRAM) rebalanced on 8-3 to add ChangXin Technology while cutting GigaDevice's weight, creating direct passive selling pressure.

4. By driver type, today is led by "geopolitics + M&A," not earnings. Three events landed at once: the AZN–BMY merger talks at roughly $400 billion (FT exclusive; BMY +4.58% / AZN −4.23%), Prysmian's all-cash $95/share acquisition of ATKR (+28.15%, definitive agreement signed), and FERG joining the S&P 500 on 8-5 (+7.75%). Only the latter two are confirmed facts; neither party has confirmed the first.

5. Pre-market state: S&P futures 7,562.25 (+0.57%), Dow futures 53,189 (+1.05%), Nasdaq 100 futures 28,446 (+0.15%), Russell 2000 futures 2,960.20 (+0.76%); VIX 15.96 (−0.19%, essentially a carry-over of Friday's close, of limited pre-market value); 10Y 4.688% (about 5bp lower than Friday's 4.738%), 2Y 4.252% (−4bp), 30Y 5.226% (−4bp); the dollar index near 100, USD/JPY down to 156.34 on "joint" US-Japan intervention (about −1% from Friday's close, about −4% from the overnight high above 163); gold 4,115.90 (+0.22%).

6. The biggest risk today is not on the tape, it is in whether the news itself is true: the whole risk-on move rests on the premise of "US-Iran talks resuming Monday," and Iran's foreign ministry publicly denied it this morning. Spokesman Baghaei stated explicitly that "we currently have no direct or indirect negotiations with the United States," and that what is being discussed is only an arrangement with Oman regarding Strait of Hormuz shipping routes. Trump says talks resume, Iran says there are none — every position betting on geopolitical peace today rests on a fact confirmed by only one side.

---

1. News Overview

---

2. Strongest Themes, Descending

---

3. Overall Single-Stock Strength Rankings

Scoring is in §4. The total score is "signal strength"; a high score on a bearish name means a strong bearish signal, not a positive view.

3a. Bullish direction

3b. Bearish direction

Note on this table: it is sorted in descending order of bearish signal strength, and the conclusion labels are independent of the table's direction. NVDA appears here because it is dragged down by the sector rather than by anything wrong at the company; its decline (−1.27%) is materially smaller than SMH (−2.32%), which makes it a relative strength observation.

---

4. Single-Stock Scoring Model (theoretical maximum 95, not 100)

Basis correction: the seven positive components defined in the skill sum to 15+20+15+15+10+10+10 = 95, with a risk deduction of 0 to −15. The actual theoretical maximum is 95, not 100. This brief applies no artificial rescaling; please read the scores on a 95 basis.

Coverage note: §3 scores 20 names in total; this detail table lists 13 of them. The seven not listed — STX / AMD / EBAY / SPCX / XOM / CVX / NVDA — were scored quickly without component breakdowns, so their totals are less reproducible than those above; please note this when citing them.

---

5. Detailed Analysis of Top Names

① PLTR — tonight's verdict: a stock where "meeting guidance equals missing consensus"

• Related news: reports Q2 after the close on 8-3 (17:00 ET call). Consensus revenue $1.812 billion (+80.5%), EPS $0.34.

• Catalyst logic — the most important finding in this brief: the company's own Q2 guidance is $1.797–1.801 billion, midpoint $1.799 billion (+79.2%), while consensus of $1.812 billion sits outside the top end of the guidance range (about $11 million / 0.61% above the $1.801 billion top end; about $13 million / 0.72% above the midpoint). "Fully meeting company guidance" is mathematically equivalent to "missing consensus."** (Confirmed against the SEC 8-K original plus independent third-party verification.) Further: FY26 full-year guidance is $7.650–7.662 billion (+71%), and backing out the midpoint implies second-half revenue of $4.224 billion, YoY of only +63.2% — the company's own full-year number already implies deceleration from ~80% to ~63%. Mitigants: PLTR has beaten consensus 8 times in a row, and in Q1 it raised full-year guidance by 10 percentage points in one go; it has raised guidance every quarter historically, so 63% is most likely a floor.

• Fundamentals (stockanalysis + SEC 8-K primary check; TTM aggregated by quarter by the author and cross-verified):

  • TTM revenue $5.2246 billion, GAAP net income $2.2815 billion, gross margin 84.1%, GAAP operating margin 38.1%
  • Operating leverage is the hardest part of the story: over 5 quarters GAAP operating margin went from 1.3% (Q4'24) → 46.2% (Q1'26); revenue nearly doubled while operating expenses rose only +24%
  • TTM FCF $2.688 billion, FCF/net income 1.18×; CapEx is only 0.45% of revenue; net cash +$7.81 billion, no interest-bearing debt
  • A counter-intuitive fact: in Q1'26 non-GAAP EPS of $0.33 was actually lower than GAAP EPS of $0.34 — non-GAAP deducts $243.6 million of tax at a normalized 23% rate, while the GAAP effective rate was only 1.37%. So whether the "$0.34 consensus" is GAAP or adjusted directly changes the beat/miss verdict, and media figures conflict between $0.34 and $0.35. This is the single most likely thing to be misread tonight.
  • SBC risk is materially smaller than the market narrative: TTM SBC $730.3 million = 14.0% of revenue, and the ratio has fallen for 4 consecutive quarters (15.9%→12.4%); diluted share count is up only +0.71% YoY

• Valuation: P/E (TTM) 138.3 (the lowest level in about 3 years), forward P/E 77.3, P/S 56.5 (about the 45th percentile of the past 3 years, but still absolutely high relative to 16–22x in 2023), EV/EBITDA 142.3 (3-year low). Earnings are catching up with valuation, but P/S is the one metric that has not gotten cheap. Basis note: the "71x revenue" circulating in the media does not match the recomputed value of 56.5x (295,010/5,224.6 = 56.46).

• Pre-market and technicals: +2.32% ($125.92); 52-week range $106.37–$207.52, 16.5th percentile of the range, −40.7% from the high (calculated on the 7-31 close of $123.06; on the pre-market $125.92 it is 19.3% / −39.3%). Options imply ±10–15%, versus an actual average of 7.39% over the past 4 quarters and 15.4% over the past 14 events. IV crush risk is high.

• Final call: watch closely. Q1 TCV +61% is already slower than revenue +85% — that is the single line to watch tonight, more important than revenue itself.

② The memory sector — today's strongest force, pointing down; but the cause is not today's news

(a) First, correct the timeline: this is week 6 of a bear market, not day 1 at the top

Basis note: "from peak" above is uniformly computed as 8/3 pre-market price ÷ peak close − 1, recomputed row by row. Note that this is not the same basis as the 52-week range highs cited in §3b/§5②(e) — those 52-week highs are intraday prices (MU $1,255.00, SNDK $2,354.39, WDC $799.87, STX $1,145.00), and drawdowns measured from intraday highs would be another 2–5pp deeper. The peak dates and peak closes come from a sub-agent's market-data pull; the author could not independently recompute them (both yfinance and akshare were restricted), so they are flagged as unverified.

These names topped out one after another within the six trading days of 6/18–6/25 (spanning two calendar weeks, not strictly the same week). This is not four separate company events; it looks like a common factor. But the marginal impact of today's news should not be dismissed either: beyond the early-stage financing detail of "seeking at least RMB 60 million from the Beijing economic development zone," the same Reuters piece carries weightier capacity information — once CXMT's Shanghai and Hefei fabs under construction come online, total capacity will more than double from the current level to roughly 600,000 wafers/month, and the second Beijing fab is not included in that figure. So the accurate statement is: the decline started in mid-June, and today's news is a marginal increment within an ongoing move, not the starting point.

(b) Same-day performance in A-shares and Asia: direction is consistent, attribution is not unique

⚠️ A piece of reasoning that must be restrained: at first glance "Chinese memory stocks lead the decline" looks like a refutation of the "share transfer" narrative — but the subject of the news, ChangXin Technology, was up on the day (+1.98%), while the four decliners are all NOR/MCU makers or memory-module makers, and not one of them is a share beneficiary of this capacity-expansion news. Using them to refute "share transfer" is picking the wrong sample.

This brief therefore downgrades its conclusion to: the day's A-share performance cannot distinguish between "the price cycle has peaked" and "share is concentrating toward a new leader," and both may be true at once. In addition, at least two mechanical explanations that have nothing to do with fundamentals remain unexcluded:

• Time-zone catch-up: A-shares closed on 7-31 (03:00 ET) before the −5% to −6% US decline on Friday, so 8-3 is the first session to digest that move.
• Passive flows: the world's largest memory ETF (Roundhill Memory, DRAM) rebalanced on 8-3 to add ChangXin Technology while cutting GigaDevice's weight, creating direct passive selling pressure on the latter; layered on top of sector money concentrating into the leader in the first 5 days after a new listing with an RMB 3.28 trillion market cap.

Basis note: the A-share and Korean data were obtained by a sub-agent via akshare and public market-data sources; on the author's re-check akshare failed to connect (RemoteDisconnected), so independent recomputation was not possible. The ChangXin Technology entry has been cross-verified against public media reports (+1.98% / RMB 55.04 / turnover above RMB 20 billion); the other five entries are flagged as unverified.

(c) Why does "Apple confirming price increases" fail to rescue memory stocks? — a badly underrated contract structure

Most interpretations stop at "downstream confirms price increases = bullish for memory stocks." But Micron put a ceiling on its own price increases in the FQ3 call (primary prepared remarks):

• 16 SCAs (strategic customer agreements) signed, five-year terms (CY2026–2030), take-or-pay
• The largest of these agreements set a price ceiling on existing products "capped at the CQ2 (Q2 2026) market price"
• Once all SCAs are signed, the portion with "fixed prices or price ceilings locked near the CQ2'26 market price" is expected to account for roughly 40% of company revenue
• RPO of about $100 billion; customer cash deposits and commitments of $22 billion

Implication: from CQ2'26 onward, every extra cent Apple pays in memory price increases is a cent Micron cannot collect on roughly 40% of its revenue. The transmission chain of "downstream costs rise → memory makers earn more" has been partly severed by contract structure. This is the most solid fundamental reason for "Apple complains about the cost while memory stocks fall anyway" — the market is not stupid; the market knows Micron has already sold its upside in exchange for downside protection.

And Micron itself concedes that price increases are decelerating (CFO's own words): "Our fiscal Q4 gross margin outlook reflects a meaningful moderation in the rate of price increases." Revenue QoQ growth drops straight from +73.7% to guidance of +20.6%.

(d) The fundamentals really are astonishing, and that is exactly the problem

• MU's quarterly gross margin of 84.56% is higher than Nvidia's — and it makes a commodity. Quarterly net income $28.2 billion, OCF $25.4 billion, FCF $17.6 billion, quarterly ROE about 28% (quarterly net income $28.2 billion / equity of about $102.0 billion).
• 100% of the growth comes from price, not volume: MU DRAM bit shipments up low single digits, prices up in the low 60% range; NAND shipments up mid single digits, prices up in the mid 80% range.
• Judgment: a number high enough to have no historical precedent is an object of mean reversion, not a starting point for extrapolation. DRAM has never historically sustained gross margins above 60% for more than a few quarters.

(e) The two faces of valuation — and one intuition that must be corrected

⚠️ STX's P/B of 89.68 is not comparable: its book equity has been compressed to a very small figure by years of buybacks (about $1.095 billion), so P/B loses meaning; base valuation judgments on fwd P/E 23.93 or EV/Sales. This applies the same standard used elsewhere in this brief for WDC's GAAP distortion and AAPL's ROE distortion.

⚠️ A counter-intuitive but crucial fact: the market values HDD far more richly than DRAM. WDC forward 33.3x, STX 23.9x, versus MU at only 5.71x and SNDK at 6.66x — a 4–6x difference. This shows the market does not treat them as the same asset class at all: DRAM/NAND get "peak-of-cycle profits + very low multiples," while HDD gets "long-term AI storage growth + growth-stock multiples." Corollary: what is embedded in WDC/STX share prices is "HDD is now a permanent growth business," and the bar for falsifying that assumption is far lower than for MU. In other words — if the cycle really turns, the group in greater valuation danger is HDD, not DRAM. This is the opposite of intuition, and the opposite of my own earlier ranking.

"A forward P/E of 6.66 is cheap" is the classic trap in cyclicals: the pricing rule for cyclicals runs the other way — P/E is lowest when E is at its peak, and peak E is the least sustainable. In 2018 MU's forward P/E briefly fell to just 3–4x, and the stock then halved. The market assigning 5.7x is not saying "cheap," it is saying explicitly "I do not believe this E lasts more than 1–2 years." P/B is the number that stings: MU at 9.23x, versus 2–3x at typical historical cycle peaks. The counterargument is that MU's book value per share grows by about $25 a quarter, so over four quarters it could go from $89 to about $190, automatically digesting P/B down to about 4.3x. Whether P/B of 9x is a bubble or just "the denominator hasn't caught up" depends on how many quarters the earnings last — which is precisely the crux of the whole debate.

(f) Do HDD and DRAM share the same downstream drivers? — the divergence is rational, but there is also a case for today's joint decline

Customer mix explains the 7/31 divergence (all figures from company-disclosed segment data):

Apple's complaint lands precisely on "memory sold into consumer electronics": SNDK gets 61.6% of revenue from Edge (the layer Apple sits in) and fell the most; WDC sells 89% to cloud providers, and Apple's handset memory costs have nothing to do with it, so it rose instead. The 7/31 divergence was not noise; it was the market precisely distinguishing end-customer structure.

But there are substantive reasons for today's joint decline too (it cannot simply be dismissed as collateral damage):

1. NAND and HDD are genuinely substitutable, and the direction is unfavorable for HDD: nearline HDD's pricing power over the past two years derives partly from QLC SSDs losing per-TB cost competitiveness as NAND prices rose. If NAND prices collapse, SSDs will immediately reattack the nearline market — a NAND downturn is not neutral for HDD, it is directly bearish.
2. The demand source is the same: nearline HDD demand and HBM demand both come from AI data centers. If what is really being repriced today is "AI capex has peaked," then the two falling together makes complete sense.
3. But the supply side genuinely does differ: CXMT makes zero hard drives; WDC's quarterly capex is only 4.3% of revenue and STX's 5.2% (nobody is expanding HDD capacity); WDC is 100% sold out for 2026, with long-term agreements extending into 2027–2029.

• Final call: avoid MU / SNDK / WDC / STX, all of them. Note that Wednesday's (8-5) earnings create two-way risk — SNDK options imply ±25%, WDC ±16.8% — and MU's forward 5.71x means one strong guide could trigger an instant short squeeze. This is the kind of setup where "the direction may be right but the timing is terrible." The real thing to watch on Wednesday is not Q4 itself (both are near-certain to beat guidance), but: ① whether SNDK's NBM agreements also contain price ceilings like Micron's SCAs (if so, the core thesis of "buy SNDK to play rising NAND prices" is overturned); ② whether WDC's 2027–2028 long-term agreements lock volume only or both volume and price (if price is locked in, then 51–52% gross margin is the ceiling and a 33x forward P/E has no support).

(g) A reverse tail risk that is completely unpriced today

According to CNBC, CXMT is already on the Pentagon's 1260H list of military-linked companies, and an interagency committee has approved adding it to the Commerce Department's Entity List, held back only out of consideration for trade negotiations with China (CNBC, 2026-06-17¹⁸); its 466% first-day listing surge has already triggered a congressional investigation. This item is media reporting; the official list text has not been seen, so its evidence tier is below company disclosure — cite with caution. The market today is pricing CXMT's expansion as purely bearish, but this same news also raises the probability that CXMT gets sanctioned. That tail scenario is not in today's prices by a single cent.

③ BMY / AZN — a mega-merger that punishes the acquirer and rewards the target

• Related news: 8-2 FT exclusive, confirmed by Reuters sourcing, neither party has confirmed.

• ★ One piece of arithmetic that must be spelled out: AZN's market cap of $263.7 billion + BMY's market cap of $133.4 billion = $397.1 billion, almost exactly equal to the reported "$400 billion." This shows the FT figure is simply the sum of the two current market caps, with no acquisition premium included. Applying the 30–40% premium typical of pharma M&A implies real consideration of roughly $173.4–186.8 billion, and a total transaction size of roughly $437.2–450.5 billion (author's own estimate; this is a sum-of-consideration basis, not a post-merger market cap, and acquirers' shares typically fall in tandem). Every dilution calculation based on "$400 billion" will understate the impact.

• Catalyst logic (the crux of the disagreement): Jefferies says it is "a bit perplexed," and states bluntly that "if there is one company that does not need financial engineering, it is AstraZeneca." BMY trades at about 11x 2027E (14.38x on a TTM basis, 10.19x forward), AZN at about 15x (15.68x forward). Buying a low-multiple asset with high-multiple stock may be EPS accretive near term, but the valuation multiple gets dragged toward the lower-multiple target. UBS (Buy, target 17,600p) said it was "surprised" by the report and warned that BMY would bring AZN an earlier-arriving cliff. Union Investment fund manager Markus Manns was more direct: "A merger with Bristol makes no sense either strategically or financially."

• ★ BMY's patent cliff (the most important verification in this brief, with a counter-intuitive conclusion):

  • Eliquis $14.443 billion (30.0%) + Opdivo $10.049 billion (20.9%) + Opdivo Qvantig subcutaneous formulation $238 million (0.5%) = $24.730 billion = 51.3% of FY2025 total revenue, the bulk of which hits LOE in 2027Q4–2028. (BMY official FY2025 annual report PDF; note this includes Qvantig — Eliquis + Opdivo alone are $24.492 billion / 50.8%)
  • Eliquis US generics come earliest on 2028-04-01 (Federal Circuit ruling); Europe in mid-2027Q4, and the company has guided to a resulting $1.5–2.0 billion step down in 2027.
  • The Revlimid cliff is essentially behind it (FY2025 $2.95 billion, 6.1% of revenue, already down from a peak above $12 billion) — the real cliffs are Eliquis and Opdivo.
  • ★ Counter-intuitive finding: Eliquis Q2'26 global revenue was $4.481 billion, up +21~22% YoY, and the company raised its 2026 Eliquis growth guidance from 10–15% straight to 20–25%. Why would a drug about to go off patent, and which was just cut 56% by the IRA, grow 22%? Because of the IRA's "nominal price cut, actual net price increase" effect: the large cut to list price simultaneously eliminated the CPI penalty rebates based on AMP inflation.
  • Judgment: this growth spurt raises the base from which the 2027–2028 cliff falls (Q2 annualized is already about $17.9 billion vs FY2025's $14.4 billion). The flattering near-term EPS (forward P/E of only 10x) is precisely the final surge before the cliff, not sustainable earnings. If the report is accurate, it also explains why BMY would entertain talks with its stock at 52-week highs — negotiating a share exchange using financials temporarily flattered by an unexpected IRA windfall is, from a seller's timing perspective, rational. (Again: neither party has confirmed the transaction itself.)

• Why the acquirer AZN fell sharply instead (three reasons beyond the valuation multiple):

  1. Narrative damage exceeds numerical damage: AZN's valuation premium rests on "organic growth + the 2030 revenue target of $80 billion." Acquiring BMY amounts to admitting organic growth is insufficient — the signal the market reads is "management is not that confident in its own 2030 target."
  2. ★ AZN was already a laggard, not falling from a high: before this news AZN was already 20% below its 52-week high, up only about 9% over the past year, versus BMY +49%, MRK +65%, LLY +51% over the same period. The market's patience with it was already running out; this news was merely the trigger.
  3. Antitrust is a real issue: Opdivo ($10 billion) and Imfinzi are both PD-(L)1 checkpoint inhibitors and overlap directly; UBS notes duplication across solid tumors, hematology, and cardiovascular, which will invite deep scrutiny and probable divestitures — and what gets divested is often exactly where the deal's synergies come from.

• Peer valuation positioning (stockanalysis, 7-31 close): the market splits big pharma into two groups — growth names (LLY forward 32.2x, MRK 21.1x) and cliff names (PFE 8.8x, BMY 10.2x). AZN's 15.7x sits exactly in between, reflecting "growth that is real but already decelerating" (Q2'26 revenue +6.41%, the lowest in about 6 quarters). A cautionary sample: NVO trades at a TTM P/E of 11.09 but a forward P/E of 15.29 — forward above trailing means the market expects earnings to decline. That is exactly the shape of the "low P/E trap," and worth holding up against BMY.

• Pre-market and technicals: BMY $68.30 (+4.58%), already through its 52-week high of $65.66, at the 98.5th percentile of the 52-week range — no overhead supply, but it also means the good news is being priced quickly. AZN $162.46 (−4.23%), having traded −6% to −7.8% at one point in London.

• Final call: BMY watch closely (if the deal is confirmed, +4.6% is far from reflecting a control premium; if it falls apart, the 52-week high of $65.66 loses its supporting rationale). AZN watch only — the consideration structure is completely unknown (possibly cash + stock), and AZN's FCF after its rigid dividend leaves only about $5.8 billion/yr of discretionary cash, which makes an all-cash acquisition nearly impossible; it would necessarily be mostly stock plus large borrowings, so dilution risk is substantive. A possibility that cannot be ignored: UBS speculates the report may in fact describe a narrow single-product/franchise collaboration rather than a whole-company acquisition. If that is clarified during the week, AZN would recover quickly.

④ AAPL — a story the media has collectively told wrong

• Two popular claims that must be corrected (both checked against the 8-K primary source):

  1. The "double beat" needs a large discount: EPS of $2.02 vs consensus $1.89 looks like a $0.13 beat, but the 8-K explicitly states FQ3 included +$0.11 of EPS and about +2pp of gross margin from a tariff refund. Stripping that out, operating EPS ≈ $1.91, essentially in line with consensus.
  2. "Greater China is weakening" is wrong: the 8-K states Greater China revenue of $18.816 billion vs $15.369 billion a year ago, +22.4% YoY; +30.0% for the first nine months of FY26. Cook's own words were that every geographic segment grew double digits. The correct statement is "high growth, but about 4% below already-elevated sell-side expectations (consensus about $19.6 billion)."

• The real risk is a third point, and it is quantifiable:

  • FQ4 guidance of 9–11% → $111.69–113.74 billion, midpoint $112.71 billion; consensus is $114.9 billion (implying +12.1%), above the top end of guidance. The gap is −$2.19 billion / −1.9%; even hitting the top end still leaves −1.0%.
  • Gross margin path: adjusted 49.3% (March) → 48.1% (June) → 46.5% (September guidance), a cumulative −280bp over two quarters; the CFO said more than 100% of the sequential gross margin change is explained by memory costs (confirmed in the full Six Colors transcript).
  • EPS sensitivity (author's calculation): FQ4 revenue $112.7 billion, tax rate 16.5%, share count 14.65 billion → each 1pp of gross margin ≈ $0.064 of EPS; −160bp ≈ −$0.10 of EPS (about 4–5% of FQ4 EPS).
  • Cook described memory prices as a "100-year flood," and Apple has already been forced to raise iPad and Mac prices.
  • The hardest behavioral evidence (8-K balance sheet, primary): inventory surged 94% in half a year, from $5.718 billion to $11.092 billion — Apple is spending real money stockpiling memory and locking in capacity.

• Fundamentals: TTM revenue $466.8 billion, net income $128.9 billion, gross margin 48.65%, operating margin 33.17%; TTM FCF $136.7 billion, FCF/net income 1.06×; net cash +$62.2 billion. FQ3: iPhone +21.7%, Mac +28.7%, Services +12.1%, iPad −5.9%; Services gross margin 75.6% vs Products 40.1%. Accounting note: ROE of 148.75% is an accounting artifact of years of buybacks thinning equity, and equity is now recovering from $73.7 billion to $107.5 billion (+45.8%), so ROE is structurally declining and YoY comparisons are distorted — use ROA of 27.08% instead.

• Valuation (the falsification point for the "it's fully corrected" claim): after a 7.35% decline, P/E (TTM) of 35.44 is still around the 85th percentile of the past 3 years; P/S of 9.66 is around the 92nd percentile of the past 3 years, only 0.5% below the 3-year high of 9.71. Of the +47.8% gain over the past year, about one-third came from multiple expansion (P/E 30.5→35.4, +16.1%) and the remaining roughly +27% from earnings growth — i.e. the gain was mainly earnings-driven, but the multiple was expanding at the same time.

• Pre-market: +0.84% ($311.50), 52-week range $201.50–$344.57, still at the 75.1st percentile and −10.3% from the high (percentile and distance-from-high calculated on the 7-31 close of $308.91; on the pre-market $311.50 they are 76.9% / −9.6%). This is a pullback from a high, not a breakdown from a low.

• Final call: watch only. The size of the bounce does not match the magnitude of the negative, and the valuation percentile offers no margin of safety.

⑤ ATKR — the highest certainty, and the least meat on the bone

• News: 8-2 company release, Prysmian all cash at $95.00/share, EV about $3.8 billion, roughly a 23% premium to the 90-day VWAP, targeted to close by year-end 2026.
• Arb arithmetic: pre-market $93.50 vs consideration $95.00 → only 1.60% of remaining spread; if it closes at year-end (about 5 months), that is roughly 3.9% annualized, below the short-end risk-free rate.
• Fundamentals: TTM revenue $2.93 billion, TTM EPS −4.85 (loss-making), but Q3 adjusted EPS of $1.92 beat the $1.70 consensus; pre-market $93.50 is already above the 52-week high of $90.16.
• Call: watch only. A definitive agreement means maximum certainty, but the upside has been eaten entirely by the +28.15% gap.

⑥ FERG — mechanical buying, but today is not the entry point

• News: 7-31 after hours, official from S&P DJI, joins the S&P 500 before the open on 8-5, replacing Electronic Arts, which was acquired by the PIF/Silver Lake/Affinity consortium.
• Fundamentals: market cap $45.4 billion, P/E 23.35, TTM revenue $31.06 billion (+2.8%), EPS $10.04 (+25.2%); plus a $1.6 billion acquisition of FloWorks; reports earnings on 8-10.
• Call: watch only. The announcement is out and the stock has already gapped +7.75% pre-market; index fund rebalancing is typically completed in the closing auction on the trading day before the effective date (8-4), so the buying ends once 8-5 opens, and the 8-10 earnings create a second risk. Historically the "inclusion effective date" is often a local top.

⑦ BABA — the model capability is real, the monetization path is distant

• News: Qwen3.8-Max, 2.4 trillion parameter MoE with 95 billion active, multimodal, 1 million token context. On Arena.AI it became the highest-ranked Chinese text model, but it still trails several Anthropic models.
• Fundamentals: market cap $278.9 billion, P/E 19.16; FY2026 revenue RMB 1.02 trillion (+2.74%), profit RMB 105.9 billion (−18.20%). Revenue is close to stagnant and profit is declining, a clear gap versus the "AI leader" narrative.
• Pre-market: +4.15% ($127.32), −34% from the 52-week high.
• Call: watch closely. It is in a "narrative-first" phase.

⑧ UAL / DAL — the most direct beneficiaries, but on the most fragile premise

• Logic: fuel is one of the largest cost items for airlines, and the cost elasticity to a 5% oil decline far exceeds the broad market. JETS +1.63% (pre-market ETF volume is thin, so treat this as directional only — it is not enough to "confirm" the sector reaction).
• Data: UAL $124.27 (+2.42%), P/E 11.35; DAL $88.80 (+1.56%), P/E 14.51.
• Call: watch closely, but tie it to geopolitical verification. The entire logic rests on "US-Iran talks resuming," which Iran publicly denied today. If harder rhetoric emerges this afternoon or oil rebounds, airlines will be the fastest group to give back gains. They have already gapped pre-market, so chasing carries significant risk.

⑨ CRCL — today's cleanest rating-driven negative

• News: Morgan Stanley (James Faucette) Equalweight → Underweight, price target $106 → $38 (a 64% cut). Rationale: shrinking USDC supply exposes the sensitivity of reserve income; USDC assumptions cut −33% for 2027 and −44% for 2028, with GAAP EPS estimates about 3% and 20% below consensus; tokenized money market funds and deposits threaten both USDC balances and the take rate.
• Data: $58.85 (−6.01%); TTM revenue $2.86 billion (+51.5%) but net income −$14.26 million; −69% from the 52-week high; reports earnings on 8-5.
• Call: avoid. The $38 target is far below the current price. Note the extreme dispersion: the average target across 27 analysts is $118.26, and TD Cowen just initiated at Buy with $82 — a spread from $38 to $118 is itself a high-risk signal.

⑩ SPCX — a double supply shock on a locked-in calendar

• News: first post-listing earnings report on 8-4; lockup expires 8-6, releasing about 911.5 million shares (about $116 billion), with a further 458 million shares potentially released under specified conditions; Musk and core insiders are locked up until mid-2027.
• Data: $106.16 (−2.04%); TTM revenue $19.3 billion (+33.2%), net income −$9.36 billion; down about 52.9% from the post-IPO high of $225.64 (pre-market basis; on the 7-31 close of $108.37 it is −52.0%).
• Call: avoid. Two major events within two days, and the company is still deeply loss-making. The average target of 35 analysts, $236.71 (sell-side view, stockanalysis aggregate, as of 7-31), diverges sharply from the share price — a textbook case of sell-side estimates not yet finished being cut.

---

6. Bearish / Avoid List

---

7. Within-Theme Rankings

Theme 1: Memory / semiconductors (direction: bearish)

★ Correction to the previous version's ranking: looking only at TTM P/E (MU 18.57, WDC 29.84, STX 61.59) would lead to the conclusion that "STX has the thinnest valuation cushion and MU the thickest." But WDC's TTM P/E is distorted by the SanDisk equity gain, and the forward basis completely inverts that ranking — the real valuation risk sits with the HDD group (WDC 33.3x / STX 23.9x), not the DRAM group (MU 5.7x / SNDK 6.7x).

Theme 2: Falling crude (direction: two-way)

Theme 3: Pharma M&A (direction: two-way)

---

8. Open-Bell Verification Signals

Pre-market (before 09:30)

• The single most important one: whether the internal divergence among the four memory names reappears. If WDC/STX (HDD) declines narrow noticeably while MU/SNDK keep falling, the market is pricing by customer structure (HDD sells 89% to cloud and is unaffected by Apple); if all four widen their declines together, it confirms the market is trading the common factor of "AI capex has peaked" — which is far more serious.
• The Dow vs Nasdaq gap: currently YM +1.05% vs NQ +0.15%. If the gap narrows (Nasdaq catching up), the memory de-rating is a localized event; if it widens, money is systematically leaving tech.
• Whether BMY can hold above its 52-week high of $65.66 (pre-market $68.30) — the dividing line for whether real money is picking up the merger expectation.

Intraday (first 30 minutes)

• Gap-and-go vs gap-fill: ATKR (+28.15%) and FERG (+7.75%) are "fully priced event" gaps, with a higher probability of filling than continuing; BMY (+4.58%) is more likely to continue because the deal is unconfirmed and the upside is unknown.
• Sector ETF linkage: whether SMH −2.32% keeps weakening; whether JETS +1.63% / XLV +1.08% can hold. If SMH's decline narrows while JETS/XLV go flat, today is just sector rotation rather than falling risk appetite.
• Volume confirmation: AAPL fell on heavy volume of 132.5 million shares Friday; today's +0.84% bounce is an invalid bounce without comparable volume.

Macro events (today)

• 09:45 ET S&P Manufacturing PMI final (prior 53.8)
• 10:00 ET ISM Manufacturing PMI: consensus 54.0 vs prior 53.3; new orders 57.0 vs 56.0; employment prior 49.7 (still below the expansion line)
• 10:00 ET ISM prices paid: consensus 70.0 vs prior 73.0 — the most underrated data point today. With oil having just collapsed 5%, if prices paid falls as expected it will reinforce the "cooling inflation + September rate cut" narrative, which is bullish for long-duration assets (and can partly hedge the memory-driven negative in tech); if it unexpectedly rebounds above 73, today's simultaneous rally in bonds and equities loses its support.
• 10:00 ET construction spending (consensus +0.2% vs +0.1%)
• The rest of this week's calendar (equally important for today's pricing): 8-4 JOLTS + trade balance; 8-5 ADP + ISM services; 8-7 (Friday) July nonfarm payrolls (unemployment rate expected to hold at 4.2%, with June adding only 57,000). Both of this brief's conclusions — the "September rate cut narrative" and "prices paid is today's most underrated data" — are ultimately adjudicated by Friday's payrolls. Also watch long-end supply (the quarterly refunding) as a separate variable, against the backdrop of the 10Y at its highest since January 2025.

Options sentiment

⚠️ Evidence-tier note: the implied volatilities in this section (PLTR ±10–15%, SNDK ±25%, WDC ±16.8%) and all consensus figures in this brief (PLTR $1.812 billion/$0.34, AAPL FQ4 $114.9 billion, ATKR $1.70, SNDK $839–842 million/EPS $34.45, WDC $3.70 billion/EPS $3.32) all come from secondary financial media accounts, without a stated exchange or precise capture timestamp, and the figures differ across sources. They are key inputs into this brief's "expectation gap" scores and its "meeting guidance equals missing consensus" conclusion, so treat them as secondary evidence. Logged as a follow-up item.

• PLTR implies ±10–15% (vs an actual average of 7.39% over the past 4 quarters and 15.4% over the past 14 events)
• SNDK implies ±25% for Wednesday (vs an actual average of 6.45% over the past four quarters), WDC ±16.8%, AMD tomorrow night
• This is a heavy IV-crush week; you can be right on direction and still lose.

Risk signals (reverse confirmation)

1. Geopolitical reversal (highest probability): Iran's foreign ministry has already denied talks with the US today. Any further hard-line statement → oil rebounds → airlines give back gains immediately, energy stocks rally.
2. ★ Today's real cognitive contradiction runs one layer deeper than "Apple complains about cost vs memory stocks fall": the surface contradiction is already explained by Micron's SCA price ceilings (about 40% of revenue is locked at CQ2'26 prices, so further downstream price increases do not reach the memory maker). The genuinely unresolved contradiction is: memory prices are still rising (TrendForce still has 3Q26 at +13~18%), yet global memory stocks have fallen for 6 weeks and Chinese memory stocks hit limit-down today. What the market is trading is the second derivative of price and the duration of earnings, not the price level itself.
3. Lone-runner risk: if BMY's rally is not followed by XLV and other big pharma (LLY/PFE/MRK), it is purely event-driven rather than a sector re-rating, and durability will be poor.
4. Yen carry unwind: joint US-Japan intervention has pulled USD/JPY back from the overnight high above 163 to 156.34 (about −4% cumulatively). If the yen keeps appreciating rapidly, a carry trade unwind could become an exogenous shock to global risk assets — a tail risk today's tape has not priced at all.
5. Rates are still high: the 10Y closed Friday at 4.738% (the highest since January 2025) and has only eased to 4.688% today. Today's equity rally rests on geopolitics rather than rates, so the foundation is not solid.

---

9. Final Conclusions

① The 5 names most worth watching today

② The 3 strongest themes today

③ What to avoid today, and why

1. All memory stocks — the direction may be right, but with SNDK implying ±25% and WDC ±16.8%, and MU at a forward 5.71x, two-way volatility risk ahead of Wednesday's earnings is very high.
2. Do not assume the DRAM group is more dangerous just because "TTM P/E is low" — on a forward basis the genuinely expensive group is HDD (WDC 33.3x / STX 23.9x), not MU (5.7x). This is the opposite of intuition.
3. Fully priced event-driven gaps (ATKR +28.15%, FERG +7.75%) — ATKR is only 1.60% below the offer, an annualized return below the risk-free rate.
4. High-multiple names on the eve of earnings (AMD P/E 158.78, CRCL, SPCX) — this is a heavy IV-crush week.
5. Do not chase the airline gaps — the entire logic rests on a premise Iran has publicly denied.
6. Do not short energy just because oil fell 5% — XOM/CVX are down only about 1%, poor risk/reward for shorts.

④ The final one-line judgment

Today is not a day of "up" or "down," it is a day of rotation: geopolitical de-escalation has pushed money out of energy and tech into airlines, pharma, and cyclicals, and the 7× gap between Dow futures at +1.05% and Nasdaq futures at +0.15% is the shipping manifest for that move. But what most needs correcting today is the timeline nearly every outlet is repeating — memory stocks did not start falling with today's CXMT fab news; they topped out on June 18, have been falling for about six weeks, and have lost roughly a quarter to a half of their value. Today's news is a marginal increment within an ongoing move, not the starting point. As for whether this is "the price cycle peaking" or "share shifting to a new Chinese leader" — today's tape offers no answer: China's module and NOR makers are falling (GigaDevice limit-down), while the subject of the news, ChangXin Technology, is itself up (+1.98%). A one-sided narrative in either direction is missing a piece of evidence today. Until SanDisk's and Western Digital's Wednesday earnings deliver a verdict, everyone with a position betting on "geopolitical peace" today should also remember: Trump says talks resume Monday, and Iran said this morning that no such thing is happening.

---

Appendix: Follow-Up Items From This Brief (to be picked up in the recap)

1. yfinance has been rate-limited for a third consecutive trading day (HTTP 429) and must be treated as an environment failure; but intermittent availability was observed this session. Recommend evaluating a replacement data source (a paid API, or formalizing stockanalysis WebFetch + company IR primary filings as the main channel).
2. Do SNDK's NBM agreements contain ceiling prices? How much revenue do they cover? — This is the single most valuable piece of information in Wednesday's earnings. Micron has explicitly stated its SCAs set a CQ2'26 ceiling on about 40% of revenue. If SNDK's 5 NBM agreements are structured the same way, the core thesis of "buy SNDK to play rising NAND prices" is overturned.
3. Do WDC/STX's 2027–2028 long-term agreements lock volume only, or both volume and price? — The HDD group's forward P/Es of 33.3x / 23.9x rest entirely on "high gross margins are sustainable." If price is locked in, 51–52% gross margin is the ceiling.
4. The basis of PLTR's consensus EPS is unsettled ($0.34/$0.35, GAAP or adjusted). Because PLTR's non-GAAP EPS is below its GAAP EPS (which is rare), this distinction directly changes the beat/miss verdict.
5. The AZN–BMY consideration structure is completely unknown (cash/stock mix), and it is the prerequisite for judging dilution. UBS's alternative explanation — that this may be only a narrow product collaboration — must also be tracked.
6. The specific year of BMY's Opdivo US LOE comes from secondary sources (multiple sources point to 2028), and no primary patent table has been seen; this requires pulling the patent section of Item 1 in BMY's FY2025 10-K (CIK 14272).

---

⚠️ Risk disclaimer: this list is a pre-market information review and watchlist only. It does not constitute investment advice and contains no position, buy/sell, or price-target recommendations; third-party price targets and ratings appearing in the text are sell-side views, with sources and dates noted. Prices in the text are pre-market snapshots from 08:00–08:07 ET, about 90 minutes before the 09:30 open, and may have changed materially before the open. US equities carry high volatility and pre-market gap risk, and post-earnings IV crush and guidance reversals occur; automatically generated content may contain stale information or factual errors. Please rely on company disclosures and SEC filings, and do not use this as a direct basis for trading.