US Pre-Market Brief | 2026-08-05 (ET) Wednesday

Coverage window: 2026-08-04 (Tue) 16:00 ET regular-session close → 2026-08-05 08:00 ET (20:00 Beijing time on August 5). Scope: Tuesday after-hours earnings (16:00–20:00 ET), the overnight Asian and European sessions, and this morning's pre-market (including pre-market earnings released before 08:00 ET). All times are stated in US Eastern Time (ET). Single-stock pre-market quote snapshots are taken as of 08:06 ET, sourced from stockanalysis.com; pre-market quotes change continuously, and actual opening prints govern.

---

0. One-line summary of the day

1. Strongest catalyst = Shopify (SHOP) pre-market blowout: Q2 revenue $3.58B (+34%) beat, GMV $115.57B (+32%), and Q3 revenue-growth guidance in the "low-30% range" far above the market's 26.3%; pre-market +23.7% to $152.50. This is the day's only combination of "extremely low expectations + extremely good results."
2. Strongest theme = in AI infrastructure, the "water sellers" outrun the "shovel sellers": Arista (networking, pre-market +11.1%), Astera Labs (interconnect, Q3 guidance $540–560M vs $392M last quarter) and Hon Hai (July revenue +54.2%, a record) delivered three simultaneous upward supply-chain readings; meanwhile the chip protagonist AMD fell -9.0% on gross margin. What the market bought today is "the supporting cast of AI," not "the lead role."
3. Driver type = earnings/guidance overwhelmingly dominant + continued geopolitical de-risking: Bessent (Treasury Secretary) said a US–Iran deal on Hormuz could come "today or tomorrow," and CENTCOM said the southbound lane of the strait is "free and open"; oil stabilized after Tuesday's plunge (Brent about $79.7, +0.43%).
4. Two samples of "beat but fell" today: AMD (revenue/EPS/guidance all three beat, yet down about 9% pre-market on a 54% gross margin versus the 56% the market cited — but neither figure's basis has been confirmed in this brief; see the warning box in Section 2) and SpaceX (revenue +92% a huge beat, EBITDA more than double expectations, yet down -11.2% pre-market on a net loss of $541 million + capex of $18.37 billion + the 8/6 lock-up expiry). One possible reading: at current levels, beating expectations is no longer enough to hold a share price up, and the market has started checking gross margins and cash flow.

⏱ Snapshot-basis note (applies throughout): all single-stock quotes below are taken before 08:06 ET, i.e. ahead of the 08:15 ET ADP employment data and the 10:00 ET ISM Services PMI. Those two prints will reset all pre-market pricing, so use live quotes before the open. In addition, for some names (WYNN, PINS, AMGN, GILD, NVO) this brief only obtained after-hours or prior-session data, with no pre-market quote; their conclusions have not been validated against pre-market prices and are flagged individually in the tables.

5. Pre-market state: S&P futures +0.41%, Dow futures +0.35%, Nasdaq futures +0.25%, Russell +0.16%; VIX 16.67 (+1.03%); gold $4,247.10 (+2.28% — risk assets and safe havens rising together, a discordant signal); 10-year Treasury about 4.61–4.63% (as of 8/4; the 8/5 intraday level was not obtained); overnight Nikkei +3.66% to 66,300.44, TOPIX +2.13%.
6. All three major indices set record closing highs on Tuesday: S&P 7,737 (+1.79%), Dow 54,086 (+1.71%, the first close ever above 54,000), Nasdaq 26,585 (+2.59%). Today is "day two after a record high," and the margin for error in chasing strength is materially lower.
7. ⚠️ Two time bombs today: ① 08:15 ET ADP July employment (consensus about 68,000, prior 98,000) and 10:00 ET ISM Services PMI (June 54.0), neither released as of this report's snapshot (08:00 ET); ② tomorrow's 8/6 SpaceX lock-up expiry: about 911.5 million shares enter the sellable-eligibility pool (the pool is about 142.8% of the 638.65 million free float), but the first tranche is capped at 20% of that, i.e. roughly 182.3 million shares near-term, about 28.5% of free float — this is a supply event with a fixed date, but "shares entering the pool" and "shares sellable near-term" are two different numbers and must not be conflated (see §5⑧).

---

0.5 List of data this brief could not obtain (the honesty cost)

The following are items this brief explicitly failed to obtain, and for which the related conclusions have therefore been downgraded or flagged as unverified. They are listed so readers know where the boundaries of this report lie:

---

1. News overview

---

2. Strongest themes, in descending order

⚠️ Gross-margin basis warning (the most important self-check in this brief)

What we independently verified is AMD Q2 GAAP gross margin = 53.77% (≈ the "54%" the market is quoting), and AMD's GAAP quarterly gross-margin path is: Q1'25 50.2% → Q2'25 39.8% (MI308 write-down) → Q3'25 51.7% → Q4'25 54.3% → Q1'26 52.8% → Q2'26 53.8%. In other words, 53.77% is a sequential improvement over last quarter's 52.82%, merely slightly below the Q4'25 peak of 54.3%.

AMD's GAAP and non-GAAP gross margins have historically differed by about 1.5–2 percentage points (acquisition-related intangible amortization + stock-based compensation). Therefore:

• If the 56% consensus is on a non-GAAP basis, AMD's non-GAAP figure most likely lands around 55%±1, and the "miss by 2 percentage points" narrative is significantly overstated;
• If the 56% consensus is itself on a GAAP basis, then it genuinely is a real miss.

These two scenarios point to opposite conclusions, and we did not obtain AMD's actual non-GAAP gross margin or the basis of the consensus. Until confirmed, this brief downgrades the "logical hardness" of this bear case from "hard" to "medium" and recommends deferring to the earnings call / 10-Q original text. | 4 | E-commerce/software: "extremely low expectations + extremely good results" | Bullish | A+ | SHOP revenue +34%, GMV +32%, Q3 guided growth far above consensus, +23.7% pre-market | Hard (company disclosure + explicit next-quarter guidance) | Medium-high (GMV has run at about +32% for four straight quarters, a stable trend) | The market had feared Meta would move directly into e-commerce and erode Shopify; this quarter's data shows no sign of that erosion (which is not disproof) | SHOP | +23.7% pre-market is a historic gap; PE 121 / forward PE 64.8, the valuation is not cheap | | 5 | Middle East de-risking / lower oil | Two-way | A | Bessent says a deal could come "today or tomorrow"; CENTCOM says the southbound lane is free and open; Brent -5.3% on 8/4 to $79.36 | Medium (depends on a negotiation not yet signed, and could reverse overnight) | Medium-low | Beneficiaries: airlines/transport/refining/overall risk appetite; casualties: the "war premium" in E&P, oilfield services and tankers | Casualties: E&P and refiners (VLO, MPC, etc.); beneficiaries: airlines and transport | A collapse in talks reverses everything; the "frozen funds" mechanism and the nuclear framework remain unresolved conditions | | 6 | NAND/storage super-cycle (tested tonight) | TBD (high volatility) | A | SNDK reports FQ4 tonight, with street revenue ($8.3–8.7 billion, inconsistent across sources) above the company's guidance ceiling of $8.25 billion; WDC the same evening; both imply ±15% | Medium (the cycle is hard, but expectations are already extreme) | Medium-high (AI storage demand is real, but cyclical names are already fully priced) | AI data centers → enterprise SSD and the new HBF standard → NAND supply shortage | SNDK, WDC, (reading: MU) | SNDK's 52-week range is $40.53–2,354.39 and it fell 47% in a single month in July; street expectations exceed the company's guidance ceiling, making the bar high — but the size of the gap swings between 0.6% and 5.5% depending on the consensus source, which this brief could not confirm | | 7 | Aerospace/satellites: great story, financials not there yet | Bearish (short term) | A | SPCX revenue +92%, a huge beat, but a net loss of $541 million and capex of $18.37 billion; lock-up expiry 8/6 | Hard (SEC 8-K) | Low (near term dominated by unlock supply; fundamentals have to wait until 2027) | High growth + enormous capex = a cash-flow gap, and the market is unwilling to absorb supply ahead of the unlock | SPCX (bearish), NVDA (beneficiary of the exclusive commitment) | If unlock selling pressure is smaller than feared, a "bad news exhausted" rebound is possible | | 8 | Media consumption: low-expectation repair | Bullish | B+ | DIS adjusted EPS $2.06 beat by about 11%, parks +10%, streaming profitability improved, full-year EPS growth of about 12% reaffirmed | Medium-high | Medium | Park resilience + streaming's first double-digit operating margin → repair of the valuation discount | DIS | Revenue $25.25B came in slightly below expectations; rising sports rights costs still weigh on the Sports segment | | 9 | Platform economy downshifts | Bearish | B+ | UBER revenue missed and Q3 bookings and EPS guidance both came in below consensus; PINS guidance merely "in line" | Medium-high | Medium | Growth moves from "beat" to "in line," and richly valued platform stocks start to be repriced | UBER, PINS | UBER's TTM free cash flow topped $10 billion for the first time; the fundamentals have not deteriorated, growth has simply downshifted |

---

3. Overall single-stock strength ranking (broadly in descending total score; bullish/bearish zones separated)

How to read the total score: it measures "this catalyst + the quality of this company," not "you should buy it today." High scores appearing in the bearish zone (e.g. AMD, SPCX) are a normal and important signal: they mean "the company is not low quality, but today's price direction is down."

⚠️ Note on score traceability: the scoring table in Section 4 breaks down only the 9 names LLY, ANET, DIS, SHOP, NVDA, ALAB, AMD, SPCX and UBER. Total scores for the remaining names (WYNN, AMGN, SNDK, WDC, PLTR, GTE, GILD, PINS, NVO, LCID, VPG) are rough editorial judgments with no line-item support and are not comparable across to the scores in the scoring table.

⚠️ Note on ordering: the table is broadly in descending total score, but positions 9–12 were manually adjusted for differences in evidence strength (WDC scores below PLTR, yet its catalyst occurs today while PLTR's was already realized yesterday), so this is not a strict descending sort by score.

3A. Bullish zone

⚠️ WDC accounting-basis warning (total score cut from 56 to 51 because of this)

WDC's TTM net margin (55.29%) is actually higher than its operating margin (31.26%) — that inversion is itself a red flag. The cause has been verified: the SanDisk stake WDC retained at the spin-off is marked to market through the income statement, and the company itself says the adjustment "does not reflect operating results." In the same quarter GAAP EPS was $8.20 while adjusted EPS was only $2.72.

• Cash flow is the mirror that exposes it: WDC's TTM operating cash flow / net income is only 52% (FCF/net income 46%), and the gap is exactly this non-cash gain.
• Metrics not contaminated at the net-income line: EV/EBITDA 47.75x, EV/FCF 64.57x — clearly showing WDC is not cheap.
• Another anomalous signal: WDC's forward PE (33.5) is higher than its trailing PE (30.0). For a high-growth stock it should normally be the reverse, which itself says the trailing end has been flattered.
• This tailwind has most likely ended: WDC agreed on 2026-06-11 to exchange out its remaining 1,038,681 SanDisk shares, completing on 6-22. And SNDK's share price has fallen since peaking on 6/25 — whether tonight's remeasurement is a gain or a loss over April–July is directionally unknown.
• trailingPEG = 0.48 is a distorted metric (the denominator uses the flattered earnings growth rate) and must not be cited. | 11 | PLTR | Palantir | AI software | Bullish (yesterday's news) | B | 59 | Closed +29.45% at $162.66 after 8/4 earnings; no new catalyst today | Low (not same-day news) | Strong: TTM revenue $6.16B (+78.9%), PE 139.1 / forward 86.1 | Extremely low (already realized in one move) | -0.04% to $162.60 (flat) | High-level consolidation after a +29% single-day move; extremely high valuation | Watch only | | 12 | GTE | Gran Tierra | Energy M&A | Bullish | A | 51 | Selling its Colombia/Ecuador business for $1.33 billion, net cash proceeds of about $315 million, buybacks planned | Extremely high | Weak: a small E&P pivoting to Canada and Azerbaijan post-deal | High | +43.92% to $9.83 (a three-year high) | Poor liquidity, small market cap; closing requires shareholder and regulatory approval and waits until year-end; already +44% pre-market | Watch only |

3B. Bearish / avoid zone

---

4. Single-stock scoring model (100 points)

Deduction rationale in detail:

• AMD −15: a gross-margin miss with no recovery timetable + a -9.0% pre-market gap + de-rating risk at a PE of 132.
• SPCX −15: a certain supply shock from tomorrow's unlock + a cash-flow gap from enormous capex + already below the IPO price.
• SHOP −12: a historic +23.7% pre-market gap + PE 121 / forward 64.8.
• ALAB −12: PE 178; and two consecutive legs down after hours and pre-market = good news that does not lift the stock, the most classic signal that expectations are already overextended.
• ANET −11: already at an all-time high pre-market, so any fill starts from an elevated level; competition from NVDA Spectrum-X and AVGO is a medium-term variable.
• UBER −9: guidance below expectations on both counts, and the stock is already approaching its 52-week low, with a weak trend.
• LLY −8: market cap near $1 trillion, so the denominator of growth is getting larger; already gapped +5.1% pre-market.
• DIS −6: revenue slightly below expectations; cost pressure in the sports segment is unresolved.
• NVDA −7 (cut from −5): beyond "the catalyst is a supply-chain reading rather than its own earnings," two earnings-quality deductions were added — TTM profit includes $25.1 billion of non-operating investment gains (restating lifts PE from 32.5x to about 37.4x), and OCF/net income has fallen four years running to 0.79x, with inventory +128% YoY outpacing revenue at +85%.

---

5. Detailed single-stock analysis (10 names)

Note: this section covers the 10 names whose catalysts today most merit elaboration, which does not exactly match the top 10 of the Section 3 ranking — WYNN (69) and AMGN (66) are not expanded because the evidence is too weak (single secondary source only, no pre-market quotes), replaced by SNDK (58), which has a definite catalyst tonight.

1礼来LLYTotal 84 | Priority deep-dive · Eli Lilly

• Related news: 08-05 pre-market, company disclosure (Lilly IR⁸). Q2 revenue $23.0B (+48%), EPS $8.38 vs the expected $6.07 (a 38% beat), and full-year revenue guidance raised from $82–85B to $85–87B. Mounjaro $9.9B (+91%, US $4.8B / ex-US $5.2B), US Zepbound $4.9B (+44%). First selling season for the oral weight-loss drug Foundayo (approved in April).
• Catalyst logic: this is a double hit of earnings plus guidance, and the size of the guidance raise ($3 billion) is itself larger than many companies' full-year revenue. It is positive for revenue, EPS and valuation simultaneously. More important is the relative picture: in the same week Novo Nordisk guided full-year growth to 0% to -6%, so the gap between the two is no longer "competition" but "stratification."
• Theme and stage: the GLP-1 theme is in its mid-to-late fermentation stage, but Lilly itself is in a fresh launch phase because oral formulation volumes are only just beginning.
• Fundamental verification (source: stockanalysis.com, as of the 08:06 ET snapshot; the financial basis is labeled item by item below):
  • Income statement: TTM revenue $79,666M (+22.2% vs FY25), TTM net income $26,711M, TTM diluted EPS $29.80, TTM gross margin 83.40%, operating margin 49.72%, net margin 33.53%. Q2 gross margin 85.78% (+3.85pct QoQ, a six-quarter high); gross margin has risen year after year: FY23 79.25% → FY24 81.31% → FY25 83.04% → TTM 83.40%.
  • ⚠️ The EPS basis must be kept straight: the widely discussed EPS of $8.38 vs the expected $6.07 is on an adjusted (non-GAAP) basis; GAAP diluted EPS for the same quarter was $7.94, a difference of $0.44. The two must not be conflated.
  • Cash flow (one quarter behind, as of 2026-03-31): TTM operating cash flow $20,480M, capex -$8,657M, free cash flow $11,823M. Cash conversion OCF/NI = 81.0%, a healthy level.
  • ⚠️ Basis trap: LLY's income statement is updated through 2026-06-30, but the balance sheet and cash flow statement still stop at 2026-03-31 (this morning's 10-Q is not yet in the database). The cash-flow figures above and the debt figures below are therefore all on a Q1 basis and cannot be paired directly with TTM revenue of $79.7B.
  • Debt (as of 2026-03-31): total cash $5,282M, total debt $43,370M → net debt $38,088M (debt exceeds cash); debt/equity 1.39, liabilities/assets 73.2%. TTM dividends $5,586M + buybacks $5,264M = $10.85 billion of shareholder returns, roughly 92% of FCF of $11.82 billion — the leverage is an active choice of "returning everything while expanding capacity," not passive distress.
  • Valuation cross-check: stockanalysis gives a trailing PE of 39.63, but recomputing with its own TTM EPS of $29.80 and the closing price of $1,115.68 gives only 37.44 (that PE appears not to have synced this morning's newly loaded Q2 income statement). This brief uses the self-computed 37.4x; on the pre-market price it is about 39.4x. The forward PE of 31.23 (about 32.6x on the pre-market price) implies forward EPS ≈ $35.7. ⚠️ A secondary source states the company's raised full-year EPS guidance is $35.50–36.50 (if true, the forward valuation already reflects the new guidance rather than the old consensus), but this brief could not verify that EPS range in the company press release — the only guidance we confirmed in company disclosure is the $85–87B revenue range. That EPS guidance is flagged unverified and is not used as a basis for conclusions. Others: P/S 12.49, EV/EBITDA 30.70, ROIC 45.45%. ROE 85.6% (ending-equity basis) to 97.0% (average-equity basis) — ⚠️ this ROE is materially inflated by high leverage (D/E 1.39) and continuous buyback-driven equity shrinkage, and is not directly comparable to low-leverage peers.
• Moat / market position: clearly the stronger side of the GLP-1 duopoly; injectables (Mounjaro/Zepbound) plus orals (Foundayo/orforglipron) cover both formats, with high capacity and channel barriers.
• Pre-market and technicals: +5.12% to $1,172.76, with a 52-week range of $623.78–1,249.45, currently at roughly the 88th percentile of that range and about 6.5% below the 52-week high. Today is an earnings day, so IV crush applies.
• Final judgment: the most solid fundamentals of the day. The risk is not in the logic but in the level — market cap is near $1 trillion and the stock has already gapped 5% pre-market. Label: priority deep-dive; but note it has already gapped 5% pre-market, so the current level is no longer the pre-earnings cost basis.

2Arista Networks | Total 78 | Priority deep-diveANETwait for a fill

• Related news: 08-04 16:05, company disclosure (Seeking Alpha⁶). Q2 revenue $3.036B (+37.7%, expected $2.82–2.83B), adjusted EPS $1.02 (expected $0.88). Q3 guidance revenue $3.3B vs the expected $2.95B (an 11.9% beat), EPS $1.06–1.08 vs the expected $0.92 (about a 16% beat). First quarter ever above $3 billion.
• Catalyst logic: guidance matters more than the current-quarter beat — and this is the textbook case. A 7% beat in the current quarter is unremarkable, but guiding the next quarter 12% above consensus and EPS guidance 16% above means order visibility is improving rather than being pulled forward. Positive for revenue and valuation at the same time.
• Theme and stage: Ethernet AI fabric is in its fermentation stage (not the launch stage, and far from crowded). Once GPU clusters scale out, back-end networking becomes a genuine bottleneck.
• Fundamental verification (source: Arista's official earnings press release + stockanalysis.com, as of 08:06 ET):
  • Income statement: TTM revenue $10,541M (+32.6%), TTM net income $4.04B, TTM diluted EPS $3.16 (+23.9%), TTM gross margin 63.00%, operating margin 43.14%, ROE 31.48%.
  • Q2 (company primary basis): revenue $3.036B (+37.7% YoY, +12.1% QoQ); GAAP gross margin 62.9% / non-GAAP 63.4%; GAAP operating margin 45.4% / non-GAAP 49.9% (48.8% a year ago); GAAP EPS $0.95 / non-GAAP EPS $1.02 (+40% YoY).
  • ⚠️ One detail that must be spelled out: ANET's gross margin actually fell 2.32 percentage points YoY (65.25% → 62.93%), but operating expenses grew only 17.5% against revenue growth of 37.7% — operating leverage fully offset the gross-margin decline, and operating margin actually rose +0.66pct YoY. This is the classic structure of "AI mega-customer volume dilutes gross margin, but margins do not fall." (A direct contrast with AMD: both had gross margins diluted by AI customers, and ANET's expense leverage caught it — so did AMD's — but the market only punished AMD.)
  • Cash flow (extremely high quality): TTM operating cash flow $5.31B, capex only -$151.3M, FCF $5.16B; OCF/net income = 131%, FCF/net income = 128%.
  • Debt: cash and marketable securities $13.343B, essentially zero total debt → net cash +$13.34 billion (about 5.0% of market cap), current ratio 2.96.
  • Deferred revenue of $6.866B = 65% of TTM revenue — this is both a reservoir of future revenue (visibility) and a sign that revenue recognition timing depends on acceptance clauses. We recommend treating the sequential change in deferred revenue as the core tracking item for ANET's growth quality.
  • Valuation: on the close, trailing PE 60.29, EV/EBITDA 48.83, P/S 22.76, EV/FCF 43.94; on the pre-market price of $211.66, trailing PE rises to about 67x, EV/EBITDA to about 54.6x and P/S to about 25.3x. The forward PE of 41.94 reflects the pre-earnings consensus; consensus will necessarily be revised up, and after revision forward PE is about 45x (this is our estimate, not public data).
  • ⚠️ Unverified: the "full-year 2026 guidance raised to $12.6B (+40%)" circulating in the media does not appear in Arista's official press release (which contains Q3 guidance only), and that figure very likely comes from the earnings call. This brief does not cite it as an established fact.
• Moat / market position: the share leader in high-end data center Ethernet switching, with switching costs from the EOS software stack; zero debt + $13.3 billion net cash + FCF/net income of 128% make it the cleanest earnings quality on the board. But the risk must be named: NVIDIA's Spectrum-X Ethernet platform and Broadcom's merchant silicon are moving directly into back-end AI fabric, the largest medium-term uncertainty.
• Pre-market and technicals: +11.10% to $211.65, against a 52-week high of $194.35 — already an all-time high pre-market. Analyst target revisions are clustered upward (TD Cowen to $210, Barclays highest at $289,30-firm consensus target about $192.31, already exceeded by the pre-market price).
• Final judgment: among the hardest logic of the day, but the price has already moved first. Gap-and-go or gap-fill — the volume in the first 30 minutes after the open is the only arbiter.

3迪士尼DISTotal 77 | Watch closely · Disney

• Related news: 08-05 07:00, company disclosure (CNBC¹¹). FQ3 revenue $25.25B (+7%, slightly below the expected $25.44B), adjusted EPS $2.06 (expected $1.85–1.86, about an 11% beat). Experiences segment $9.97B (+10%), streaming revenue +11~13% with sharply improved operating profit, reaching a double-digit operating margin for the first time. Reaffirmed full-year adjusted EPS growth of about 12% (excluding the extra week) and buybacks of no less than $8 billion.
• Catalyst logic: a textbook low-expectation repair. Revenue actually missed, but the profit line and segment quality were good enough, and with the stock down 12.1% YTD and 17% over 52 weeks, there is almost no "good news" component in the market's pricing. What is most affected is valuation, not revenue.
• Theme and stage: media consumption is in its launch phase (the repair has only just begun).
• Fundamental verification (stockanalysis.com, 08:06 ET): TTM revenue $98.86B (+4.6%), EPS $4.84 (-24.1%), market cap $170.49B, forward PE 13.28 — the cheapest on the board. ⚠️ The data provider's trailing PE of 15.69 does not reconcile with the 20.3x implied by its own TTM EPS of $4.84 and the closing price of $98.18 (likely a mix of adjusted and GAAP bases), so this brief does not cite that trailing PE. Operating cash flow, free cash flow and debt were not obtained this time — so the "thickest downside protection" judgment below rests only on valuation and share-price position, unvalidated by cash flow.
• Moat / market position: the IP library plus the physical barriers of parks and cruises make it the only "strong moat + low valuation" combination on the board.
• Pre-market and technicals: +2.87% to $101.05, with a 52-week range of $92.19–119.78, at roughly the 32nd percentile of that range — the lowest-positioned strong-catalyst name on the board. The gap is mild and gap-fill risk is comparatively the smallest.
• Final judgment: the best risk/reward of the day — a definite catalyst, low valuation, a low position and a small gap. The upside is less than LLY/ANET, but the downside protection is the thickest. Watch closely.

4Shopify | Total 75 | Watch closelySHOPlargest gap · highest gap-fill risk

• Related news: 08-05 pre-market, company disclosure (Reuters/Yahoo¹⁰). Q2 revenue $3.58B (+34%, expected $3.45B), GMV $115.57B (+32%). Q3 guidance revenue growth in the "low-30% range" (expected 26.3%), gross profit growth in the "mid-to-high-20% range" (expected 24.1%).
• Catalyst logic: two effects. ① Guidance is significantly above expectations; ② this quarter's data shows no sign of Meta erosion — the biggest open question in the market this quarter (note: a single quarter of GMV data cannot disprove a competitive threat, it only shows the threat has not yet materialized this quarter). What is most affected is sentiment and valuation; revenue elasticity is secondary.
• Theme and stage: e-commerce SaaS moves from "fear of disruption" back to "share expansion," which is the starting point of a narrative reversal.
• Fundamental verification (stockanalysis.com, 08-05 08:06 ET): TTM revenue $12.37B (+31.8%), EPS $1.02, market cap $160.41B, PE 121.46 / forward PE 64.76. Cash flow was not obtained this time.
• Moat / market position: merchant-side infrastructure plus payments/logistics network effects; but the moat's "width" still trails LLY/NVDA.
• Pre-market and technicals: +23.68% to $152.50 (close $123.30), with a 52-week range of $94.00–182.19, putting the pre-market price at roughly the 66th percentile of that range. This is the largest gap of the day and the most classic high-risk gap-fill sample.
• Final judgment: the logic holds, but the price already reflects the large majority of it. This brief's focus is on whether the gap can be held, not on the print of the first trade at the open.

5英伟达NVDA / NVIDIATotal 71 | Watch closely

• Related news: two items, both from the evening of 08-04 to the morning of 08-05. ① On the SpaceX earnings call, Musk announced SpaceX will use NVIDIA exclusively for AI infrastructure going forward (Vera Rubin architecture), and is partnering with NVIDIA on the Starmind AI1 satellite compute payload, with 2GW by year-end and 10GW by end-2027 (TechCrunch⁵); ② Hon Hai July revenue NT$946.5 billion (about $29.3 billion), +54.2%, a monthly record (Bloomberg¹³).
• Catalyst logic: this is supply-chain validation rather than its own earnings, so directness is inherently one notch lower. But the two items point the same way: one is "an exclusive commitment from a new large customer" (future orders), the other is "revenue already realized at the contract manufacturer" (current shipments). Positive for sentiment and order visibility, with no direct effect on current EPS.
• Theme and stage: the AI compute theme is in its crowded stage, but today, with AMD hit on gross margin, NVDA benefits on a relative basis, producing a rare "divergence within the leaders."
• Fundamental verification (source: NVIDIA TTM financial statements + stockanalysis.com, as of 08:06 ET): TTM revenue $253,491M (+70.7%), TTM net income $159,613M (net margin 62.97%), TTM diluted EPS $6.53, gross margin 74.15%, operating margin 64.02%, ROE 114.29%, ROIC 107.45% — ⚠️ this ROE is inflated both by the $25.1 billion of non-operating gains described above and by TTM buybacks of $45.3 billion shrinking equity, so it is not a comparable quality metric; cash + short-term investments $80,572M, total debt $12,348M → net cash +$68.2 billion; market cap $5.13 trillion.
• ⚠️ Major earnings-quality flag (independently verified in this brief, and material enough to change the conclusion): of NVDA's TTM pre-tax profit of $189,443M, $25,131M comes from non-operating "fair-value gains on investment securities" — explicitly reversed out of operating cash flow in the cash flow statement as Gain Loss On Investment Securities = -$25,029M, i.e. a non-cash, non-operating gain. Restating at the 15.75% effective tax rate:
  • TTM net income ≈ $138.5 billion (-13.2%), diluted EPS ≈ $5.66 (not $6.53)
  • trailing P/E ≈ 37.4x (not the headline 32.5x), EV/EBITDA ≈ 30.2x (not 26.2x)
  • Corroboration: the EBITDA figure from mainstream data vendors ($165.51 billion) is itself $27.5 billion below reported EBITDA ($192.97 billion), showing that some data sources have already identified this basis difference.
  • Conclusion: any comparison calling NVDA cheap on a headline PE of 32x is contaminated by these $25.1 billion of gains. This brief uses the restated figure of about 37x.
• Another quality metric to watch: NVDA's OCF/net income has fallen for four consecutive years (1.29x → 0.94x → 0.88x → 0.86x → TTM 0.79x); TTM working capital absorbed -$21.06 billion, of which inventory is +128% YoY (to $25.8 billion), outpacing revenue at +85% (receivables turnover has not deteriorated, with DSO of 45 days vs 46 days a year ago). There is also an acquisition of about $14.2 billion (goodwill jumping from $6.26 billion to $20.83 billion), whose target and impact were not obtained this time and remain unverified.
• Moat / market position: the CUDA ecosystem plus a full stack (GPU/networking/software) — the strongest moat of all; tangible book value of $171.46 billion is 88% of shareholders' equity (versus only 34% for AMD).
• Pre-market and technicals: +1.60% to $215.34 (close $211.94, +2.56%), 52-week range $164.07–236.54, at roughly the 71st percentile of that range. The gap is mild and does not constitute chase risk.
• Final judgment: today is not NVDA's "big day," but it is the only large-cap name combining "a positive catalyst + a mild gap + a forward valuation that is not expensive." Watch closely.

⑥ ALAB — Astera Labs | Total 69 | Watch only

• Related news: 08-04 16:05, company disclosure (Investing.com¹⁴). Q2 record revenue $392.4M (+104% YoY, +27% QoQ), EPS $0.80 (expected $0.69). Q3 guidance $540–560M — implying sequential growth of nearly 40% versus last quarter.
• Catalyst logic: on the numbers, this is the report with the strongest earnings elasticity of the day. But the market's reaction was: regular session +12.65% to $361.67 (front-running ahead of the print), after hours -2.3%, and pre-market a further -1.84% to $355.00.
• Theme and stage: AI interconnect is in its crowded stage.
• Fundamental verification (stockanalysis.com, 08-05 08:06 ET): TTM revenue $1.20B (+98.5%), EPS $2.03 (+238.2%), market cap $61.99B, PE 178.15 / forward PE 74.89.
• Pre-market and technicals: 52-week range $97.89–499.48, currently at roughly the 64th percentile of that range.
• Final judgment: this is the most important cautionary tale of the day — revenue doubled, guidance doubled again, and the stock still fell in two consecutive legs after hours and pre-market. "Good news that does not lift the stock" means expectations were already pulled forward to a PE of 178. Being right on the logic does not mean there is money to be made today. Watch only.

7超微半导体AMDTotal 69 | Watch only (No. 1 in the bearish zone) · Advanced Micro Devices

• Related news: 08-04 16:05 plus the 17:00 earnings call (StockTitan¹). Q2 revenue $11,536M (+50.1%, not the +52% circulating in the media), adjusted EPS $1.66 (GAAP diluted EPS $1.38), a double beat; data center $6.7B (year-ago $3.2B, more than double); Q3 guidance midpoint about $13.0B (roughly 2.9% above consensus). The trigger for the drop: gross margin of 54% vs the 56% consensus.
• Catalyst logic: this is the most instructive pricing case of the day. Revenue beat, EPS beat, guidance beat — all three — and the stock still fell about 9%. The market's explanation is that the CFO explicitly said the Helios AI rack (priced at $5–5.5M per rack, in volume production in Q3) carries below-average margins early on and will structurally dilute blended gross margin during the ramp — i.e. "the more you sell, the worse near-term margins get." But see the gross-margin basis warning in Section 2: the GAAP gross margin we verified, 53.77%, is actually a sequential improvement over last quarter's 52.82%, and the basis of the 56% consensus is unconfirmed. The quality of this bear case is itself in question.
• ⚠️ Two things the "gross margin narrative" obscures (independently verified in this brief):
  • ① Operating margin hit a high for the available series: GAAP operating margin Q1'26 14.4% → Q2'26 17.3%; TTM operating margin jumped from 11.65% to 15.71% in a single quarter. Q2 revenue rose +12.5% QoQ while operating profit rose +34.8% QoQ. Expense leverage is delivering powerfully — talking only about the gross-margin miss without mentioning a record operating margin is selective narrative.
  • ② What actually deserves worry is capex and free cash flow, and the market has barely discussed it: Q2 capex was $808M, +107.7% QoQ (Q1 $389M); Q2 free cash flow was $1,558M, -39.3% QoQ — despite revenue rising +12.5% QoQ in the same period. TTM capex as a share of revenue rose from 3.07% to 4.06%; single-quarter OCF/net income fell from the TTM 1.94x to 1.03x. AMD's current P/FCF is about 100x (about 92x pre-market), and the sustainability of that FCF denominator matters far more to the valuation than gross margin does.
• Medium term it is bullish: named commitments include up to 2GW for Anthropic (with AMD simultaneously committing up to $5 billion of equity investment), up to 6GW for Meta, and a 15-year lease of 529MW with Core Scientific (worth more than $14 billion, expandable to about 2.5GW). These are real, named compute demand. But note that YoY growth is downshifting: Q2 +50.1% → Q3 guidance implying +40.6%.
• Fundamental verification (source: stockanalysis.com Q2 data + a pro-forma TTM assembled from AMD's historical quarterly statements; note that most mainstream databases have not yet updated AMD's fundamentals through Q2, so taking them directly produces systematically low figures):
  • TTM revenue $41,305M, gross margin 53.20%, operating margin 15.71%, net income $6,434M (net margin 15.58%), diluted EPS $3.89; ROE 9.57%, ROIC 9.77%.
  • A clean balance sheet: cash + short-term investments $13,111M, total debt $4,276M → net cash +$8.84 billion, with debt/equity of only 6.4%.
  • But equity quality deserves a discount: of $67,224M of shareholders' equity, about $41.8 billion is goodwill and intangibles (the legacy of the Xilinx acquisition), leaving tangible book of only $23 billion, 34% of equity (versus 88% for NVDA). So AMD's P/B of 12.6x corresponds to a P/tangible book of about 34–37x — "P/B is only half of NVDA's" is a book-value illusion.
  • Working capital is actually improving (an overlooked positive fact): DSO fell from 77 days to 54 days; inventory rose +5.3% QoQ against revenue +12.5% QoQ, so turnover improved.
  • Valuation: trailing PE 133.3 (121.9 pre-market), P/S 20.47, EV/Sales 20.26, P/FCF 100.6.
• ⚠️ Forward PE is unusable: two mainstream public data sources give AMD forward PEs that differ systematically by 27–29% (implying forward EPS of $13.92 and $10.99 respectively), and the difference almost certainly comes from a different forecast fiscal year (FY+1 vs FY+2) or a different GAAP/non-GAAP basis. Until the basis is confirmed, this brief adopts no forward-PE conclusion for AMD.
• Pre-market and technicals: about -8.6% to -9.0% (the two public sources' pre-market quotes differ by about 0.4%, a range of $472–474), against Tuesday's close of $518.58 (+7.00%) — giving back all of Tuesday's 7% gain and then some. The 52-week range is $149.22–584.73, putting the close at the 84.8th percentile and the pre-market price at the 74.6th percentile; the pre-market $474 is still 51% above the 200-day moving average of $314.56 and still in the 82nd percentile of the past year — this decline has not changed the "elevated" status. Annualized volatility 71.8%.
• Final judgment: the fundamentals are not merely intact, operating margin hit a record; but there is no timetable for "when gross margin inflects," the basis is itself in doubt, and the FCF erosion from a doubling of capex is the more concrete risk. Until the earnings call / 10-Q confirms non-GAAP gross margin and capex guidance, the evidence does not support a judgment in either direction. Label: watch only.

⑧ SPCX — SpaceX | Total 68 | Avoid

• Related news: 08-04 16:05 SEC 8-K plus the earnings call (SEC³). Q2 revenue $7.81B (+92%, expected about $6.93B); Starlink $4.3B (+66%), subscribers doubled to 12 million, Starlink operating profit $1.66 billion (the only profitable segment); adjusted EBITDA $3.5B (expected $2.0B, a 75% beat). But: a net loss of $541 million and quarterly capex of $18.37 billion (about 6x YoY).
• Catalyst logic: revenue and EBITDA were both overwhelming beats, but the market is looking at only two things: ① $18.37 billion of quarterly capex versus $7.81 billion of quarterly revenue — this is a company burning cash far faster than it earns it; and ② tomorrow's lock-up expiry.
• Unlock details (today's most certain risk, and more extreme once quantified than it first appears): on 8/6 (Thursday) about 911.5 million shares enter the sellable-eligibility pool, with insiders able to sell at most 20% of that in the first tranche; and the free float we verified is only 638.65 million shares, 4.85% of shares outstanding.
  • Two numbers that must be kept separate and never conflated:
    • Shares entering the pool: 911.5 million ÷ 638.65 million = 142.8%, i.e. the pool is about 1.43x the existing float;
    • Actual near-term sellable amount: subject to the "at most 20% in the first tranche" cap = 911.5 million × 20% ≈ 182.3 million shares, about 28.5% of free float.
    • This brief's risk discussion uses the near-term 28.5%; the 1.43x figure is the ceiling once all staged unlocks are complete, not tomorrow's quantity.
  • ⚠️ Two conflicting public figures exist for the float basis: free float of 638.65 million shares (4.85% of shares outstanding) coexists with "public float of under 280.1 million shares," from different sources on unreconciled bases. This brief consistently uses the 638.65 million float basis (more conservative, producing a smaller multiple); if the true float is closer to 280.1 million, the ratios above would more than double. This is an unresolved denominator problem in this brief, and the prospectus/S-1 lock-up terms should govern.
  • This is the classic unwinding of a "small float propping up the price" structure. The IPO price was $135 (listed 2026-06-12), the current price is already below it, and the stock is about 43% off its peak.
  • ⚠️ One clarification: SPCX's revenue and adjusted EBITDA this quarter both beat by a wide margin, and the company did not cut guidance. The negative reaction comes from the net loss, capex and unlock expectations, not from an earnings blow-up.
  • Insiders' actual willingness to sell is unknown and cannot be inferred from financial data — so "the unlock will definitely crush the stock" is not this brief's conclusion; our conclusion is only that "this is a supply event with a fixed date and meaningful size, and the uncertainty cannot be removed before the unlock date."
• Fundamental verification (source: stockanalysis.com, as of 08:06 ET):
  • Income statement: TTM revenue $23,044M, TTM net income -$8,218M, TTM gross margin 51.87%, operating margin -14.92%.
  • Q2 actually had bright spots: revenue went from $4,694M to $7,814M QoQ (+66.5%), while the operating loss narrowed from -$1,943M to -$143M, and gross margin went from 49.13% to 55.27% (an all-time high). At the operating line, the quarter was close to break-even — which does not contradict the "$541 million net loss," since the difference sits mostly below the operating line (long-term debt of $29.118 billion newly issued in Q2, interest expense, etc.).
  • Cash and debt (unexpectedly strong): total cash $100.009 billion (cash $93.522 billion + short-term investments $6.487 billion), total debt $39.364 billion → net cash +$60.645 billion; current ratio 5.12, debt/equity of only 0.31. In one quarter cash jumped from $23.7 billion to $100.0 billion and equity from $41.6 billion to $127.2 billion, almost entirely from IPO proceeds + $29.1 billion of long-term debt issued in the quarter.
  • But cash flow is the sore point: TTM operating cash flow $9.9 billion, TTM capex -$42.42 billion → TTM free cash flow -$32.52 billion. At the current burn rate, $60.6 billion of net cash lasts roughly two years (a rough calculation in this brief; it is a static extrapolation that ignores revenue growth, changes in the capex cadence and refinancing capacity, and should not be used as a conclusion).
  • ⚠️ All per-share metrics are distorted and not comparable across periods: public data shows two SPCX TTM EPS values simultaneously, -$0.65 and -$2.52, and both are valid — the denominators differ: the former uses the current share count of 13.174 billion shares, the latter the pre-IPO weighted average of about 3.261 billion shares. The IPO increased the share count by 304%, so any SPCX per-share metric spanning 2026-06-12 (EPS, BVPS, PE, PEG) is non-comparable. (The book-value-per-share series confirms it: Q4'25 $0.87 → Q1'26 $8.89 → Q2'26 $21.70.)
  • ⚠️ The forward PE of 100.54 cannot serve as a valuation anchor: it implies forward EPS of about $1.25, i.e. SPCX swinging from a loss of $0.62 per share to net income of about $16.5 billion within the next 12 months. That is an extremely strong assumption and can only be read in reverse, as "what the market is implying." More reliable are P/S of 71.7x and EV/Sales of 69.1x; even annualizing Q2 revenue ($7,814M × 4 = $31.3 billion), P/S is still 52.8x.
  • ⚠️ TTM revenue itself contains constructed data: SpaceX disclosed only semi-annual data before listing, and public databases split the $10,536M of H2 2025 into two nominal "quarters" by averaging (the two quarters are identical line by line). Of the $23.044 billion of TTM revenue, $10.536 billion (45.7%) comes from that semi-annual split — the total is accurate at the half-year level, but any SPCX quarter-over-quarter trend line (other than Q1'26 and Q2'26) is pseudo-data, and the denominator of P/S deserves a question mark too.
  • Annual data only goes back to FY2023. Worth noting: FY2024 net income was positive at $791M, FY2025 swung to -$4,937M, and TTM widened to -$8,218M — the losses widen with the Starship/AI capex cycle rather than being a chronic condition, but only three years of data support that.
  • Company-basis data we could not independently verify: Starlink subscribers of 12 million, Starlink operating profit of $1.66 billion and adjusted EBITDA of $3.5 billion — these three are company-defined segments on a non-GAAP basis with no corresponding fields in standard financial data, and this brief labels them "company-disclosed basis."
• Pre-market and technicals: -11.20% to $111.29 (Tuesday close $125.33, +9.43%). The 52-week range is $104.83–225.64, putting the close at the 17.0th percentile and the pre-market price at only the 5.4th percentile — near the lowest since listing; it is 44.5% off the 52-week high (50.7% pre-market). Implied pre-market market cap is about $1.466 trillion.
• Final judgment: the gap between the company narrative (a $100 billion annualized run-rate by year-end, $1 trillion by 2030) and the current financials is too wide, and tomorrow brings a supply shock with a fixed date. Avoid ahead of the unlock date; after it, if selling pressure is lighter than feared, that becomes a "bad news exhausted" watch point.

9优步UBERTotal 63 | Watch only · Uber

• Related news: 08-05 pre-market (CNBC¹²). Q2 revenue $14.19B (+12%, below the expected $14.24B), adjusted EPS $0.81 (in line); gross bookings $58B (above the expected $57.23B), mobility $28.99B (+22%), delivery $27.46B (+26%); TTM free cash flow topped $10 billion for the first time. Q3 guidance: bookings midpoint $59.25B (expected $59.33B), adjusted EPS $0.84–0.88 (expected $0.89), both below consensus.
• Catalyst logic: a quarter of "good structure, downshifting growth." Bookings beating expectations shows demand is fine, but the revenue miss plus guidance below on both counts shows take rate and margins are under pressure. Guidance carries more weight than the current quarter, so the direction is set as bearish.
• Fundamental verification (stockanalysis.com, 08:06 ET): TTM revenue $55.23B (+16.7%), EPS $4.58, market cap $146.54B, forward PE 19.45. ⚠️ The data provider's trailing PE of 17.80 is inconsistent with the 15.7x implied by EPS $4.58 @ $71.99, so the bases are not aligned. One further note: the "forward PE above trailing PE" phenomenon was shown in this brief's WDC case to stem from one-off non-operating income at the trailing end; this brief did not run the same check on UBER's trailing end, so we do not assert from it that "the market expects EPS to decline."
• Pre-market and technicals: -3.53% to $69.45 (close $71.99), 52-week range $65.41–101.99, at roughly the 11th percentile of that range — already near the 52-week low.
• Final judgment: the valuation is not expensive and free cash flow set a record, but both the trend and the guidance point down. Fundamentals-driven investors can research it; trend followers should not touch it today. Watch only.

10SanDisk | Total 58 | Watch onlySNDKno position ahead of earnings

• Related news: reports FQ4 and full-year results tonight at 16:00 ET, with the call at 16:30 ET; an investor day follows on 8/13 (Sandisk IR³⁰). The company previously guided revenue to $7.75–8.25 billion and non-GAAP EPS to $30–33; the street is at about $8.7 billion / EPS $34.8–35.5.
• Catalyst logic: this is a high-bar setup where street expectations sit above the company's guidance ceiling — what the company has to beat is not its own guidance but the line the market raised on its own. The options market prices an implied one-day move of about ±15% for the 8/7 expiry, with a call/put ratio of about 1:1 (no clear directional bias).
• Theme and stage: the NAND/storage super-cycle is in a post-crowding choppy phase — already 38.9% off the 2026-06-25 high and 16.3% below the 50-day moving average ($1,704.83); but over the last 5 trading days it has ripped from $1,015.89 to $1,427.62 (+40.5%), a classic high-volatility front-run into earnings.
• Fundamental verification (source: stockanalysis.com + SanDisk company guidance, as of 08:06 ET):
  • TTM revenue $13,184M (+82.8%), TTM net income $4,507M, TTM diluted EPS $29.71, gross margin 56.04%, operating margin 41.60%, ROE 39.30% (⚠️ this sits at a cyclical profit peak with capacity held in an unconsolidated joint venture, so this ROE, like the P/B below, is not comparable across cycles); TTM operating cash flow $4,639M, OCF/net income 103%.
  • The single-quarter ramp is extraordinarily steep (precisely the dangerous pattern in cyclicals): 2025-03 revenue $1,695M / gross margin 22.54% / EPS -$13.33 → 2025-06 $1,901M / 26.20% / -$0.16 → 2025-10 $2,308M / 29.77% / $0.75 → 2026-01 $3,025M / 50.94% / $5.15 → 2026-04 $5,950M (+251% YoY) / gross margin 78.35% / operating margin 69.09% / EPS $23.03.
  • Debt: cash $3,735M, total debt of only $207M → net cash +$3.53 billion; TTM net debt repayment of about $2 billion.
  • ⚠️ The unusually low capex ($179M TTM, only $45M in the quarter) is structural, not "asset-light": SanDisk's wafer capacity sits mainly inside the Flash Ventures joint venture system with Kioxia (unconsolidated), so expansion funding and guarantees are not fully reflected in its own capex or in the "$207M total debt." FCF of $4.46 billion therefore overstates true free cash generation; the scale of off-balance-sheet commitments and guarantees was not obtained this time and remains unverified.
  • ⚠️ The "cheapness" of a 7.82 forward PE rests entirely on extrapolating the cycle peak: 7.82 implies forward EPS of $182.6 (another data source implies $212.95), i.e. quarterly EPS of $45–53; whereas the company's own FQ4 guidance is only non-GAAP EPS of $30–33. PEs computed off different anchors diverge wildly: annualizing last quarter's GAAP EPS = 15.5x; annualizing company guidance = 10.8–11.9x; annualizing the street's Q4 = 10.4x (using the single source with EPS $34.24; another source gives $34.8–35.5, which this brief could not reconcile); on the street's FY27 = 7.8x. This is the classic cyclical structure of "peak earnings, low PE" — the lower the PE, the closer you often are to the earnings top.
  • Other valuation metrics: P/S 16.04, EV/Sales 15.77, EV/EBITDA 37.67, P/B 15.34 (historically NAND makers have traded at 1–3x P/B even at cycle peaks; but post-spin the equity base is small and the wafer assets sit in an unconsolidated JV, so P/B is less comparable than before the 2025 spin-off and should not be used alone to draw a conclusion).
• Pre-market and technicals: -1.66% to $1,403.94 (Tuesday close $1,427.62, +10.84%). 52-week range $40.53–2,354.39, with the close at the 59.9th percentile of that range; up about +3,305% over the past year; +66.9% relative to the 200-day moving average. RSI 47.69.
• Final judgment: tonight is a purely binary event, with street expectations above company guidance, implied volatility of ±15%, and an inevitable IV crush after the print. Before the results are out, a judgment in either direction has no evidentiary basis. Label: watch only; this brief will revisit it with the actual results in tomorrow's pre-market.

---

6. Bearish / avoid list

The last column below follows this brief's fixed labeling system. "Short watch" is an observation category, not a trading recommendation; labeling a name "short watch" only means its downside drivers are relatively clear and it merits tracking, not that this brief believes a short position should be established.

---

7. Within-theme ranking

Theme 1: AI infrastructure supporting layer (networking / interconnect / contract manufacturing)

Theme 2: GLP-1

Theme 3: NAND storage (tonight's binary event)

Theme 4: Low-expectation repair (media consumption)

---

8. Opening verification signals

Pre-market (before 09:30 ET)

• Confirming the main line: watch whether ANET is gap-and-go or gap-fill. ANET's pre-market $211.65 is already above the 52-week high of $194.35, and if it cannot hold the $194–200 zone within the first 30 minutes, this brief leans toward reading today's "AI networking" move as a one-day affair.
• Gap quality grading (the larger the gap, the higher the probability of a gap fill): SHOP +23.7% (highest risk) > ANET +11.1% > LLY +5.1% > DIS +2.9% > NVDA +1.6%. The small gaps in DIS and NVDA are actually the healthiest shapes.
• Absorption on the bearish side: AMD is -9.0% pre-market at $472. $472 roughly corresponds to where Tuesday's intraday rally began — if it holds after the open and turns green, the gross-margin negative is only a sentiment disturbance; if it breaks lower, the market is systematically repricing "the gross-margin assumption for AI chips," which would spread to the entire semiconductor sector.

Intraday (09:30–16:00 ET)

• The first 30 minutes' trend: focus on whether volume confirms in SHOP and ANET. The low of the first 30 minutes after a gap is one of the key reference levels this brief recommends tracking (not the sole arbiter).
• Sector ETF linkage: if AI networking is a genuine main line, ANET's gains should be accompanied by AVGO/MRVL following; if ANET rises alone while its sector does not follow, that is capital rotation rather than a sector move.
• Divergence within semiconductors: today features a rare NVDA up / AMD down. Watch whether this is "money rotating from AMD into NVDA" (sector-neutral) or "the whole semiconductor complex weakening with NVDA merely holding up better" (sector-negative) — check the SOX index and the breadth of the Philadelphia semiconductor components.
• Two macro timestamps:
  • 08:15 ET ADP (consensus about +68,000, prior +98,000): weak data = rising rate-cut expectations = supportive for valuations; but given three officials favored a hike at the 7/29 FOMC, strong data is the more dangerous outcome.
  • 10:00 ET ISM Services (June 54.0): a drop below 50 would directly change risk appetite for the whole day.

Options sentiment

• Names with rising IV today: PCG, DOCU, GME, TDC, COGT, OC, EIX, JCI and others.
• Implied one-day moves for tonight's earnings: SNDK ±15%, WDC ±15%, XYZ (Block) ±10% (call/put 1:2.4, clearly bearish-skewed), EBAY ±9%; tomorrow ABNB ±8%.
• Post-earnings IV crush: LLY, SHOP, DIS and UBER, which all reported today, have entered the IV crush phase — implied volatility drops rapidly once the event lands. This is an options-pricing phenomenon that puts extra pressure on these names' option prices independent of direction.

Risk signals

• Risk assets and safe havens rising together: gold +2.28% to $4,247.10 alongside index futures up across the board and a VIX of only 16.67. This divergence usually means some portion of the money does not believe the current risk appetite.
• Day two after a record high: all three major indices set record closing highs on Tuesday, so every act of chasing strength today lacks protection from "position."
• Geopolitical reversibility: the Hormuz deal is not yet signed. What Bessent said was "possibly today or tomorrow." A single statement in the other direction could send oil and energy stocks back up overnight.
• This week's real test comes Friday: July nonfarm payrolls at 08:30 ET on 8/7 (consensus +91,000, unemployment 4.2%→4.3%). With three votes inside the Fed already favoring a hike, both "too strong" and "too weak" are problematic — Wednesday's and Thursday's action may be no more than a bridge to Friday's data.
• Tomorrow's certain event: the 8/6 SPCX unlock.

---

9. Final conclusions

① The 5 stocks most worth watching today

② The 3 strongest themes today

③ What to avoid today, and why

1. SPCX and every trade tied to its unlock — the 8/6 expiry is a supply shock with a fixed date and a size far exceeding the float. However good the earnings, they cannot offset the arithmetic of 911.5 million shares entering the pool against a float of 280.1 million shares.
2. SNDK / WDC, which report tonight — an implied one-day move of ±15%, and street expectations already above the company's guidance ceiling (SNDK street revenue is between $8.3 billion and $8.7 billion, inconsistent across sources; street EPS about $34.2–35.5, versus a company guidance ceiling of $8.25 billion / EPS $33), so the bar was raised by the market itself. Worse, neither name's valuation metrics can be read at face value: SNDK's forward PE of 7.8x implies quarterly EPS of $45–53 (company guidance is only $30–33), extrapolating the cycle peak; WDC's trailing PE of 30x is flattered by the fair-value remeasurement of its SanDisk stake, and is about 41–60x once stripped out. Before the results are out, neither name's direction is judgeable.
3. Every name gapping more than 10% pre-market (SHOP +23.7%, ANET +11.1%, GTE +43.9%) — being right on the logic does not make the price right. The main way to lose money today is not picking the wrong direction, it is buying at the wrong price.
4. Names like ALAB where good news does not lift the stock — revenue +104%, guidance another +40% QoQ, and yet the stock fell in two consecutive legs after hours and pre-market. This is the clearest signal of overextended expectations.
5. Energy E&P and refining — if the Hormuz deal lands, the war premium disappears systematically. But the reverse risk is equally real: the deal is not yet signed.
6. Small caps like VPG with "an elevated level + an earnings plunge + unclear details" — down 27.7% pre-market, and we do not even have a reliable source for what numbers it actually reported. The uncertainty itself is reason to avoid.

④ The one-line final judgment

The core message the market is sending today is not "who went up" but "AMD beat on three metrics and still fell about 9%, and ALAB doubled its revenue and still fell in two consecutive legs after hours and pre-market" — at levels where the three major indices have just set record closing highs, beating expectations is no longer a sufficient condition, and the market has begun scrutinizing the basis of gross margins, the quality of cash flow, and whether expectations have been pulled forward. Accordingly, this brief believes the names most worth tracking today are those with "a definite catalyst but a mild gap" (DIS, NVDA); those gapping more than 10% (SHOP, ANET) have already priced in most of the logic, so the focus should be on whether the gap holds; and SPCX's uncertainty ahead of the 8/6 unlock cannot be removed by fundamentals. The real variable this week is the 8/7 payrolls print, and everything in today's session could be rewritten by Friday's data.

---

⚠️ Data-collection failures and self-check log for this run (not for clients)

A. yfinance rate limiting (occurrence N)

• The main flow's direct python3 -m yfinance calls were rate-limited by Yahoo: all four tickers DIS / SHOP / UBER / PLTR returned YFRateLimitError: Too Many Requests, zero successes; a bare curl against the Yahoo chart endpoint returned HTTP 429 Edge: Too Many Requests.
• Root cause identified: a sibling process on this machine (yfretry.py PID 3473866 / /tmp/yfretry_fund.py) was hammering Yahoo at high frequency at the same time, getting the machine's whole IP rate-limited at the CDN layer. This is our own processes fighting each other, not a unilateral Yahoo ban. Check for leftover retry processes before the next run.
• A working workaround (empirically validated by a subagent; recommended for the toolchain): inject curl_cffi.requests.Session(impersonate="chrome") into yfinance with 12–20s exponential backoff and at most 5 retries. With this method quoteSummary (.info) passed on the first try, and chart and fundamentals-timeseries were all retrieved after 4–5 backoffs. fast_info uses an endpoint that is not blocked and should be preferred.
• ⚠️ Silent-failure trap (the most dangerous one): under rate limiting, quarterly_financials returns an empty DataFrame with shape (0,0) rather than raising. In a scheduled job, without an explicit row-count check, this silently produces empty data.

B. What the three subagents actually contributed (important: they corrected substantive errors in the main flow)

• Three fundamentals-analyst subagents were launched at 08:1x ET and all returned after 9.5 / 13.3 / 13.2 minutes respectively. While waiting, the main flow had already drafted the report using stockanalysis.com, and after the subagents returned, 12 revisions were made to the body based on their conclusions, 5 of which were substantive corrections:
  1. AMD's Q2 revenue growth is +50.1% YoY, not the +52% circulating in the media (cross-checked against two independent sources). The draft said "+50~52%"; corrected.
  2. "Gross margin 54% vs expected 56%" is very likely a basis mismatch between GAAP actual and non-GAAP consensus. GAAP 53.77% is in fact a sequential improvement over last quarter's 52.82%. The draft treated this as a confirmed negative; it has been downgraded to "logical hardness = medium" with a warning box added.
  3. NVDA's TTM profit includes $25.1 billion of non-operating investment gains, and restating lifts PE from 32.5x to about 37.4x. The draft said "forward PE 21.2, relatively cheap," which is wrong; corrected, and NVDA's total score cut from 73 to 71.
  4. WDC's TTM net income is heavily flattered by the fair-value remeasurement of its SanDisk stake (the inversion of net margin 55.3% > operating margin 31.3%; GAAP EPS $8.20 vs adjusted $2.72; OCF/net income of only 52%). The draft copied PE 30.0 / forward 33.5; an accounting warning box has been added and the total score cut from 56 to 51.
  5. SPCX's unlock size can be quantified as 1.43x float (911.5 million shares vs a float of 638.65 million), which is more accurate than the draft's "911.5 million vs float <280.1 million" (280.1 million is a different basis). Switched to the float basis with both figures retained.
• Conclusion: this time the subagents were not a nice-to-have, they prevented 3 valuation conclusions that would have misled clients. Do not skip them in future just because you cannot wait.

C. Data-source reliability findings (recommended for the toolchain)

• yfinance .info's fundamental fields are broadly untrustworthy and should be recomputed from ttm_income_stmt / ttm_cashflow / quarterly_balance_sheet instead. Errors measured: NVDA.freeCashflow reported 46.34B vs 119.076B in the statements (a 157% difference), AMD.enterpriseValue uses the prior day's close, trailingEps disagreed with both companies' own TTM statements, and AMD.grossMargins differed from the same-period statement by 2.8pct. Price-type fields (marketCap / fiftyTwoWeekHigh/Low / preMarketPrice) were verified as usable.
• Forward PE differs systematically by 27–29% between the two data sources (AMD implied forward EPS $13.92 vs $10.99; NVDA $12.89 vs $9.99, both ratios about 1.28), almost certainly because the forecast fiscal year differs (FY+2 vs FY+1). Until the fiscal year and the GAAP/non-GAAP basis are confirmed, forward PE should not go into body-text conclusions. AMD was handled this way in this brief.
• Third-party databases construct SPCX's H2 2025 by splitting the semi-annual figures into two averaged quarters (Q3/Q4 identical line by line), and 45.7% of TTM revenue comes from that construction. SPCX's quarter-over-quarter trend line is essentially unusable.

D. Fetch failures

• WebFetch against CNBC, Benzinga, Schwab and TipRanks all returned HTTP 403; the relevant content was instead obtained via WebSearch summaries and cross-checked on other sites. As a result, a few items (pre-market quotes for WYNN / PINS / AMGN / GILD) have after-hours data only, and the body text says "no reliable data available."

E. Basis traps already avoided (logged for reference)

• The SanDisk "Q2 FY2026 revenue $3.03B / Q3 guidance $4.4–4.8B" in search results is the fiscal quarter released on 2026-01-29, not the FQ4 being reported tonight; it was nearly misused.
• The ANET "full-year 2026 guidance raised to $12.6B" circulating in the media is not in Arista's official press release (which contains Q3 guidance only) and is suspected to come from the earnings call. The body text flags it as unverified and does not cite it as fact.

---

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