US Pre-Market Brief | 2026-09-01 (ET) Tuesday

Coverage window: 2026-08-31 16:00 ET regular-session close → 2026-09-01 08:30 ET. Covers the 8/31 after-hours session, the overnight Asia/Europe sessions, and this morning's pre-market.

Data conventions (source and read timestamp stated item by item)

• Single-stock/ETF pre-market prices and pre-market volume: stockanalysis.com quote API, read at 08:00–08:08 ET, field Pre-market. All percentage moves are computed in-house by this report as (pre-market price − previous close) ÷ previous close, and are cited only where they agree with the API's own field.
• The "notional" column is computed in-house by this report = pre-market price × cumulative pre-market shares. Pre-market volume is cumulative from 04:00 ET, not an instantaneous value; read timestamps differ across names by up to 50 minutes, so cross-sectional comparisons must read notional and read timestamp together.
• Futures/Treasuries/VIX/dollar: CNBC restQuote, read at 07:50–08:01 ET; sovereign bonds read at 08:24 ET. All bond moves are computed in-house by this report as last − previous_close — CNBC's bond change_pct field is self-contradictory across the board today (US 2Y shows −0.0195% while last 4.358% > prev 4.350%; UK 30Y shows −1.105% while last 5.8646% > prev 5.7865%), and is rejected outright.
• The fed funds hike probability is computed in-house by this report; method and assumptions are in the §1 footnote, rather than quoting media figures directly.
• Previous closes were cross-checked against the 8/31 recap reports/us/2026-08-31-recap.md and agree (NVDA 220.78, EIX 53.98, PCG 13.27, SLB 60.10, etc., each matching).

---

0. One-Sentence Summary

Today's war risk is real and escalating (Iran has struck two US military bases in Jordan with missiles, and Trump has threatened to blow up Kharg Island, Iran's main crude export terminal); but that risk is being priced only on the crude curve, with no spillover into any traditional safe-haven asset — gold is falling, defense stocks are flat, tankers are not moving. So the correct framing today is not "this is not a war tape," but "war pricing has run down only one leg: crude." What actually determines index direction is something else: the global sell-off in long-dated sovereign bonds.

⚠️ This section's conclusion was materially revised from the first draft. The first draft said "this is not a war tape" and treated "defense stocks not rallying, gold falling" as evidence that the market was denying the war. That framing was overturned by QC review and external verification: the day did in fact see Iranian missile and drone strikes on US bases in Jordan (the IRGC's "punish the aggressor" operation), plus Trump's public threat against Kharg Island — the war is not only real, it produced the most intense exchange of fire in more than six months. The revised conclusion is about the transmission channel, not about whether the war exists. The revision trail is in the internal record at the end of this report.

1. The strongest catalyst is the global bond sell-off, not Hormuz. US Treasuries moved in sympathy: 10Y 4.792% (+3.4bp), 30Y 5.284% (+3.5bp), 2Y 4.358% (+0.8bp) — another bear steepener. This report's in-house probability of a 25bp hike on 9/16, computed from September fed funds futures, is 64.3% (method in the §1 footnote). The global side, taken in-house by this report (CNBC, read 08:24 ET):

   Sovereign bond	Current	Change	Intraday high	Note
   UK 30Y gilt	5.8646%	+7.8bp	5.9041%	Media call it the highest since May 1998
   UK 10Y gilt	5.2223%	+7.3bp	5.2554%	Leading the global sell-off
   Japan 10Y (JGB)	2.992%	−0.3bp	3.011%	⚠️ Crossed above 3.00% intraday, media call it the highest since September 1996
   Germany 10Y	3.3425%	+1.9bp	3.3642%	Media call it the highest since 2011
   France 10Y	4.1981%	+2.5bp	4.2271%	Now above Italy
   Italy 10Y	4.1856%	+2.2bp	4.2231%	—
   Australia 10Y	5.169%	+0.9bp	5.187%	—

   Table scrolls horizontally

   The single most important cell in this table is the JGB's intraday 3.011%. Japan's 10-year crossing 3% is not just a round-number level — it changes the asset-allocation arithmetic for Japanese institutions: when the domestic risk-free rate pays 3%, continuing to hold hedged foreign bonds yielding less no longer pays, and the resulting repatriation flows in turn intensify selling pressure in US and European bond markets. This is the mechanism that transmits a "domestic Japanese event" into a "global duration event," and it is upstream of today's pressure on US growth stocks.

   ⚠️ CNBC's sovereign-bond change_pct field is broken again today (e.g. UK30Y shows −1.105% while its last 5.8646% > previous close 5.7865%). The change column above is computed in-house by this report as last − previous_close.

2. Three cross-sectional pieces of evidence for "war pricing has not spilled over" (note: the three differ in evidentiary grade, labeled individually):

   • ① Precious metals are being sold. GLD −1.82%, GDX −2.72%, SLV −2.76% (read 08:12–08:28 ET; notional $156.8M / $85.4M / $85.8M, all far above threshold). But it must be stated honestly: this is not independent evidence. Gold already fell more than 3.5% cumulatively over 8/28–8/31 on Warsh's hawkish remarks, and media attributed it precisely to "reviving hike expectations" — that is, gold's decline and this report's rate thesis share the same driver, so using it to prove "the rate thesis holds" is circular. The only thing it can establish is this: in the tug-of-war between rates and war, rates won. One more caveat: gold rose nearly 10% over the full month of August and is now in its third consecutive down day; month-end profit-taking plus a firmer dollar (DXY +0.19%) is by itself sufficient to explain today's −1.8%, with no need for a stronger hypothesis.
   • ② Defense stocks are tracking the index — but the first draft got the direction of this one wrong, corrected here. ITA −0.17%, LMT −0.22%, RTX +0.07%, NOC +0.39% (08:15–08:28 ET). Against S&P futures at −0.62%, these four are actually flat-to-outperforming, not "no bid." More importantly: the first draft read tankers as "already repriced" but read defense stocks as "the market is denying the war" — the same price pattern given two opposite interpretations. The fact is that defense stocks are equally already repriced — 8/31 was day 185 of the US-Iran war, LMT/RTX set all-time highs back in March 2026, and defense names have essentially stopped reacting to Iran escalation headlines since. This item is therefore downgraded to "no information" and is not used as support. (⚠️ The four names' pre-market notionals are only $0.3M–$4.4M, all below threshold, which on its own is insufficient to support any conclusion.)
   • ③ Tanker stocks are not moving: FRO +0.11%, STNG 0.00%. The reading is "already repriced long ago" — FRO is up 103% year to date and rose 6.24% on 8/31 alone. ⚠️ But STNG's quote comes from just 154 shares ($12,000 notional), and this same report flags that group as "ultra-thin (unreadable)" in §7 — having labeled it unreadable, it should not then be read. This item is likewise downgraded to reference only.
   • Net conclusion: of the three, only ① has adequate quote quality, and ① is not independent of the rate thesis. So the judgment "war pricing has not spilled over" rests on weaker evidence than the first draft claimed, and should be treated as an observation to be validated at the open, not an established fact.

3. The only thing actually rallying is the barrel itself. USO +2.21%, Brent November $91.90 (+1.56%), WTI October $87.61 (+2.16%). All 11 energy names in this report's sample are up, median +1.29%, and XLE ($14.8M, the only one clearing the liquidity threshold) already trades above its 52-week high in the pre-market.

   But fundamental verification delivers a counterintuitive conclusion that overturns this report's first-draft ranking: $90 Brent is not bullish, it is a "step-down" — 2026Q2 Brent averaged $102.63, while Q3 to date is only about $87.0. "Oil holding at $90" corresponds to Q3 earnings falling markedly sequentially from Q2. Moreover, the five integrated oil and gas names differ by 2.9x in oil-price elasticity, yet their two-day gains are all crammed into a narrow +3.17%~+4.14% band — the market is doing an undifferentiated sector-beta rally. That makes "which one to buy" far more important than "whether to buy"; see §5②.

4. Flow direction: selling crowded high beta, buying near-dated cash flow. Semiconductors 0/25 up, median −2.15%; software/cybersecurity 0/14 up, median −1.33%; while XLV +0.55%, XLP +0.33%, with staples and pharma green across the board. Nasdaq futures −1.03% are markedly worse than the S&P −0.52% and the Dow −0.50%.

5. Driver mix: macro (rates) primary × geopolitics (supply) secondary × a ratings day as garnish. Two binary events today: the California SB 492 vote (intraday, affects EIX/PCG) and PANW earnings after the close. Pre-market state: VIX 15.88 (+6.43%), still low; S&P futures 7,659.25 (−0.52%).

📍 Drift recheck (second read at 08:17–08:21 ET, to separate "strengthening on volume" from "fading")

Ticker	08:00–08:08 read	08:17–08:21 read	Notional change	Reading
FRVO	+12.61% / $34.4M	+14.11% / $37.9M	↑	Strengthening
XOM	+1.53% / $13.1M	+1.50% / $68.3M	↑ 5.2x	Flat price on rising volume, the healthiest pattern
HOOD	+1.98% / $94.0M	+1.58% / $97.5M	Volume up slightly	Gain is converging; the ratings premium is being digested
SLB	+1.08% / $1.2M	+0.90% / $1.7M	Slightly up	Still above the 52-week high of $60.46
GPRO	+75.19% / $145.1M	+67.77% / $159.3M	Volume up, price down	Fading, and fading on volume — high-level turnover
MU	−2.15% / $653.1M	−2.32% / $706.0M	↑	Decline continuing
NVDA	−1.58% / $387.9M	−1.46% / $430.6M	↑	Slight convergence
PANW	−1.12% / $14.0M	−1.61% / $17.9M	↑	Weakening into earnings
PCG	+0.60% / $5.8M	+0.15% / $7.1M	↑	The bounce has almost entirely vanished
QQQ / SPY	−1.10% / −0.59%	−1.20% / −0.62%	↑	Index tone unchanged

Table scrolls horizontally

All tables and conclusions in the body use the first read (08:00–08:15 ET); this table is trend corroboration only. The two reads show no directional conflict.

---

1. News Overview

Two convention notes that must be stated:

1. On the ISM release date (verified; the earlier uncertainty is resolved): some search results claimed "September 1 is Labor Day, markets closed, ISM pushed to September 2" — that applies to 2025, not this year (2025-09-01 was indeed a Monday Labor Day). In 2026 Labor Day falls on September 7 (Monday), and today, 9/1, is a Tuesday and the first trading day of September. Confirmed via the Investing.com economic calendar: August ISM manufacturing releases on 2026-09-01, prior 55.6, consensus 55.6. (The "14:00" shown on that site is UTC, corresponding to 10:00 ET, consistent with ISM's standard 10:00 ET release practice.) The ISM website calendar page requires a login, so primary confirmation was not obtained, but the date is doubly corroborated by the third-party calendar and by the next-day schedule in the 8/31 recap.
2. On whether SB 492 has already been voted on: media accounts differ, with some paraphrases saying it "has passed." This report defers to the primary source — the California leginfo bill history page still showed, at this report's read timestamp, a last action of 08/30/26 "Ordered to third reading," with no vote recorded; and Seeking Alpha at 9/1 05:18 ET explicitly writes "vote today." This report therefore treats it as "vote not yet completed, lands intraday today." See §5.

---

2. Strongest Themes, Descending

⚠️ A self-correction that must be stated: the first draft stitched the two rows above into a single theme and named it "the Fed pivots to hiking." But this report's own numbers reject that attribution — 2Y is only +0.8bp while 10Y is +3.4bp and 30Y +3.5bp. If the market were newly pricing a September hike today, the 2Y should have moved most, and the 2Y barely moved. That says the day's marginal driver is the overseas long-end term premium (UK/Japan/Germany), not the Fed path; Fed hike expectations are a pre-existing stock suppressant, not today's incremental catalyst. The trading implications differ: a term-premium shock hurts long-duration assets and valuations, while a repricing of the Fed path would first hurt front-end-sensitive assets and bank net-interest-margin expectations. Also: the first draft's theme name said "hike expectations heating up," but this report gave no in-house probability for the prior day, and without a Δ there is no "heating up"; the two comparable points are Forbes's 66% on 8/31 and this report's in-house 64.3% on 9/1, which point down, not up. | 2 | Hormuz supply shock → crude | Long (energy) | S | US strike on Larak Island on 8/30; Brent $91.90 | Hard: the oil price is itself price evidence | Medium, depends on whether actual shipping disruption occurs | Oil ↑ → upstream cash flow ↑; cracks widen → refiners | XOM/COP/DVN/OXY/SLB/VLO | Already up two days; and Iran-Oman talks on reopening are stuck on a single clause, the transit fee, so the risk premium could vanish overnight in one move (see §5②) | | 3 | California wildfire liability SB 492 (vote today) | Short | A+ | Bill carries an urgency clause and goes to a vote today; BofA double-downgrade with both PTs below the current price | Hard, but outcome unknown — this is a binary event | Event-driven, the main uncertainty resolves today | Liability cap removed → payout exposure unmeasurable → equity value repriced | EIX / PCG | Two-way risk: if the final text is weaker than feared, the short-covering after −23% would be violent | | 4 | AI power (geothermal) | Long | A− | FRVO wins a Google 396MW binding PPA (primary 8-K), contracted capacity 658→about 1,054MW (+60%) | Medium: the contract is hard (Alphabet parent guarantee), but the 3GW framework is March-vintage old news and the 400MW capacity is 8/12 old news; today's only new information is the "framework→contract" step | High (15-year term) | Hyperscalers lock in power → EGS moves from demonstration to scale | FRVO (direct) | Contributes $0 in 2026–27 and cash runs out during 2027; about 205 million shares unlock in November = 2.5x the current float | | 5 | Semiconductor/memory high-beta deleveraging | Short | A | No stock-specific new news; 25/25 down, median −2.15% | Medium: a liquidity and positioning phenomenon, not fundamental deterioration | Short (1–3 days), unless yields keep rising | Crowding + high beta → declines amplified under a real-rate shock | MU/SNDK/MRVL/INTC/AMD | Do not read it as a fundamental negative; an oversold bounce can happen at any time | | 6 | Cybersecurity earnings (PANW after today's close) | Neutral-to-bearish (high bar) | A | Consensus revenue +32%, adj EPS only +3% | Medium | Event-driven | Acquisitions consolidate to lift revenue while compressing margins | PANW / comps: CRWD/ZS/S | The stock is only 4.2% below its 52-week high; expectations are already high | | 7 | Sell-side ratings day | Long (single names) | B+ | DUOL/HOOD/AKAM/TKR/SRE upgraded; EIX/PCG/IBKR downgraded | Weak: ratings are second-hand opinions, not primary facts | Low (mostly one day) | — | HOOD (the only one on volume) / DUOL / AKAM | Pre-market volumes are generally ultra-thin, see the notional column in §3 |

---

3. Overall Single-Stock Strength Ranking (28 names, split by direction then sorted descending)

"Notional" = pre-market price × cumulative pre-market shares, in millions of US dollars. This report uses $10M as a rough threshold for "the pre-market quote is credible" — percentage moves below that value are reference only and are not used as load-bearing points. ⚠️ The threshold answers only "is this quote trustworthy," not "is this news important." A name below the threshold can still be part of the day's strongest theme.

⚠️ Sample boundary (this bears on every "X/Y all up/all down" statement in this report): the breadth statistics here — semiconductors 0/25, software/cybersecurity 0/14, energy 11/11 — are based on representative names selected by this report, not the full industry universe and not the complete index constituent set. They can support the qualitative judgment "direction is highly uniform," but cannot be treated as the universal claim "the entire industry without exception." Name selection skews to large caps and high-recognition tickers; small caps and neglected names are not included. Readers who want a sector-level conclusion should use the ETF reads for SMH / IGV / XLE (−1.76% / −1.15% / +1.19%, notional $47.0M / $8.2M / $14.8M respectively), since the ETF is the full-sample weighted measure.

3.1 Bullish Side

3.2 Bearish Side

---

4. Single-Stock Scoring Model (100 points)

Broken out line by line using FRVO as the example (other names are scored the same way, components omitted). This table has been recomputed against the fundamental verification results; the original score was 81 points, cut to 59, with the reason for each cut marked:

---

5. Detailed Analysis of Top Names

1Fervo EnergyFRVOwatch closely · downgraded from "priority deep-dive"

After verification against primary SEC filings, this item's conclusion is markedly lower than the first draft's. The downgrade process is more valuable than the conclusion, so it is preserved in full.

• Related news (primary): the 8-K was accepted 2026-09-01 08:05:12 ET (event date 8/26), Item 8.01 plus the EX-99.1 press release: subsidiary Cape Generating Station 6 LLC signed a 396MW (not the 400MW claimed by media) enhanced geothermal PPA with Google Energy LLC, at Cape Station in Beaver County, Utah, in four 99MW tranches, with target COD from 2028Q3, over a 15-year delivery term. Alphabet Inc. provides a parent guarantee for Google's obligations. 8-K original⁷

• Catalyst logic: what this affects is revenue visibility, not current EPS. Binding contracted PPAs go from 658MW → about 1,054MW (+60%). Geothermal's distinctive value is a high-capacity-factor baseload — note: the company's own prospectus uses an illustrative capacity factor of about 83%, not the 90%+ often quoted informally.

• 🚩 How much of this catalyst is genuinely new? It must be unpacked in four layers — this is the most important part of this item:

  Layer	Content	New today?
  ① The binding 396MW PPA + Alphabet parent guarantee	Contracted backlog +60%	✅ Yes, a real increment
  ② "Google wants to buy Fervo's power" (the 3GW framework agreement)	Signed 2026-03-19 and covered at length in the IPO prospectus	❌ Five and a half months old
  ③ "That 400MW block of capacity"	The 8/12 Q2 earnings release already said "eight GeoBlocks (400MW) entering Advanced Development"	❌ 20 days old
  ④ The media's "nearly 1GW"	396MW firm + 600MW is only an option, Google has no obligation to purchase	🚩 Overstated by roughly 2.5x

  Table scrolls horizontally

  Conclusion: the only genuinely new information today is the confirmation of the step from "framework → hard contract with an Alphabet guarantee." That step has value (in a field littered with letters of intent, it moves Fervo from story to order book), but it is not "Google suddenly wants to buy Fervo's power." My first draft scored the expectation gap at 9/10, which was wrong — the stock trading near its post-IPO lows does not mean this news was never priced; it means the version that was priced (the 3GW framework) has already fallen from $42.65 to $14.60.

  There is also an asymmetric clause almost no one mentions: if Fervo fails to make that 600MW expansion offer, it must refund to Google amounts previously collected for "deemed delivered energy," an amount the company explicitly calls "not currently estimable." Failing to expand does not just forgo profit, it triggers a clawback.

• Fundamental verification (primary SEC, through the 2026-06-30 10-Q) — financially this company is not a power generator, it is "construction in progress + an order book":

  • Total H1-2026 revenue = $174,000 (one hundred seventy-four thousand US dollars); full-year 2025 revenue was zero. The only asset actually generating power is the 3MW Project Red pilot. So P/E, P/S and gross margin on FRVO are not "high," their denominators are meaningless.
  • The balance sheet is unusually clean: unrestricted cash $2,106.4M, total debt only $228.4M → net cash of about $1.878B. There is no near-term solvency risk — its risk is dilution, not default.
  • The cash burn is accelerating: Q2 capex alone was $226.5M; the company's own H2-2026 capex guidance is $850–900M, 2.1–2.3x the first half ($399.3M). At that pace its own cash is exhausted during 2027, with a gap on the order of $500–900M.
  • The November supply cliff: the IPO lock-up expires around 2026-11-09/10, at which point 205,005,185 Class A shares become freely tradable — 2.5x the current float (the IPO's 80.5M shares). "Refinancing is inevitable" and "a 2.5x-float unlock" collide inside the same 6–12 month window.

• Valuation (a capex adjustment is mandatory, otherwise you reach the opposite conclusion): headline EV per contracted MW is just $3.13M/MW, seemingly half of Ormat's (ORA, the only pure geothermal comparable) $6.77M/MW — but that is a trap, because none of FRVO's 1,054MW has generated a single kilowatt-hour yet; it has to be built first. Adjusting to full load at the company-disclosed actual overnight build cost for Cape Station of $7,000/kW, FRVO is at $8.88M/MW, 31% more expensive than the already-profitable ORA; only at $5,000/kW does it break even, and only at the long-run target of $3,000/kW is it clearly cheap.

  In one line: the market is paying for today's $7,000/kW while pricing $3,000–5,000/kW. What the current share price implies is not "contract value" but "the learning curve will definitely be delivered."

• Revenue scale of the contract (computed in-house, cross-checked against three independent sources): at $100–115/MWh (backed into from the company's own backlog average of $100.3/MWh, the prospectus-disclosed range of $100–130/MWh, and an external binary-cycle LCOE benchmark of $90–110/MWh — three lines converging) and an 83% capacity factor, 396MW at full output implies annual revenue of about $288–331M. For comparison, that is about 24–28% of ORA's TTM revenue of $1.19B. But 2026 and 2027 contribute $0, and 2028 sees only part of the tranches.

• Why the stock previously fell 65.8% (from $42.65 on its second day of trading): the main cause is a sector-wide de-rating across the whole AI power chain, with drawdown size closely tracking "how far you are from cash flow" — SMR −83.9%, OKLO −79.1%, FRVO −63.9%, BE −41.3%, ORA −30.3%, NEE −16.6%. FRVO sits precisely where it should; this decline was not a company-level accident.

• Pre-market and technicals: pre-market $17.32 (+12.61%) / $34.4M (read 08:04 ET). Note: a later read at 08:12 ET showed $17.58 (+14.30%) / $37.1M — both reads are correct; pre-market volume and price accumulate, they are not in conflict.

  ⚠️ One convention correction: the first draft repeatedly used "52-week range $14.60–$42.65," "52-week low," and "the first up candle off the lows" to describe this stock. But FRVO only IPO'd on 2026-05-14 (offer price $27), so it has about 3.5 months of trading history and no 52-week range exists at all. The accurate phrasing is "range since listing," and its true shape is a broken-issue new listing — the previous close of $15.38 is 43% below the $27 offer price, and it is 65.8% off the $42.65 high set on its second day. On a stock with only 3.5 months of history trading below its IPO price, "low level" has no technical meaning, and certainly cannot imply "the news has not been priced."

• Final judgment: watch closely (downgraded from "priority deep-dive"). It remains today's cleanest primary-sourced single-stock catalyst, and its balance sheet is the healthiest in the sector; but the expectation gap is far smaller than the first draft judged, and the risk load far larger. The single largest risk: the build-cost learning curve fails to materialize — if cost stalls at $7,000/kW, every MW signed destroys value, and the bigger the order book the greater the dilution, so the "good news" becomes bad news in proportion to order volume. This risk will not show up as a blow-up; it will emerge slowly, as "capex guidance raised again and again, COD pushed out again and again, one refinancing a year."

② The energy complex (XOM / CVX / COP / DVN / OXY) — watch closely

• Related news (completed via QC review; the first draft badly understated the intensity of the escalation): ① US forces struck two rocket launchers on Iran's Larak Island on Sunday 8/30, the first since late July; ② Iran's Revolutionary Guard immediately struck the King Hussein and Al Azraq US bases inside Jordan with missiles and drones (the "punish the aggressor" operation); ③ Trump posted an AI-generated bombing video on Truth Social on 8/31, threatening to blow Kharg Island, Iran's main crude export terminal, "to pieces." Hormuz carries about 20% of global oil shipments, and Kharg Island carries the overwhelming majority of Iran's crude exports. CNBC² · Al Jazeera³

  The first draft wrote only "strike on Larak Island + a drone intercepted by the UAE," substituting "one drone intercepted" for "Iran striking US bases + the US threatening to destroy the other side's main oil terminal," systematically understating the day's supply-side risk. Corrected here after QC review.

• Catalyst logic: pure supply-side risk premium. Note this is day two — the group was already up 1.64%–2.71% on 8/31.

• 🚨 Fundamental verification delivers a counterintuitive conclusion that overturns this report's first-draft ranking — this passage must be read in full:

  ① $90 Brent is not bullish, it is a "step-down." EIA official quarterly average Brent spot: 2026Q2 = $102.63; while Q3 to date (7/1–8/25) is only about $87.0. In other words, the just-reported Q2 earnings were earned in a $102.63 environment, and "oil holding at $90" corresponds to Q3/Q4 earnings falling clearly and sequentially from Q2, not rising. Any valuation anchored on "annualized Q2" will systematically overestimate.

  ② So the real expectation gap is not "is $90 high enough," but "will the sequential decline in Q3 earnings be larger than consensus." All five names currently trade near 13x forward P/E, which implies exactly that "pullback from the Q2 peak."

  ③ For pure upstream names, $90 is largely already in the price. Backing out the implied oil price from forward P/E: COP ~$90, DVN ~$93, essentially overlapping current spot at $92.10. If Brent merely "holds" at $90, there is no additional repricing room — any increment must come from prices rising further or the conflict escalating further (the Kharg Island threat is precisely that variable).

  ④ The oil-price elasticity ranking is nearly the reverse of this report's first-draft recommendation ranking. By "after-tax profit from each +$10 of Brent ÷ market cap":

  	After-tax profit per +$10 Brent	% of market cap	Rank
  OXY	$2.03B (company 10-K disclosed sensitivity)	3.37%	1
  DVN (gross basis)	$1.72B	3.08%	—
  DVN (excluding capped positions)	$1.11B	1.99%	2
  COP	$3.02B	1.90%	3
  CVX	$5.83B	1.44%	4
  XOM	$7.77B	1.17%	5 (lowest)

  Table scrolls horizontally

  Only OXY gives an explicit per-barrel sensitivity in 10-K Item 7A ($265M pre-tax per $1 barrel, WTI + Brent combined); XOM/CVX/COP/DVN disclose none, so the remaining figures above are estimated from production and pass-through coefficients (backtested on OXY, that method has an 11% error, biased optimistic by roughly a tenth).

  ⑤ The two-day price reaction shows almost no differentiation, and this is the most tradable information in this item. The five names' fundamental elasticities differ by 2.9x (1.17% vs 3.37%), yet their two-day gains from the 8/29 close to the 9/1 pre-market are all crammed into a narrow +3.17% ~ +4.14% band (XLE itself +3.28%). That says this repricing round is a sector-beta / ETF-flow rally, not differentiated pricing by each company's elasticity. Converted into "implied years capitalized": XOM about 8.9 years (relatively overshot), OXY only 2.3 years (the largest elasticity but the smallest gain).

  ⑥ Two structural changes that will mislead you, both recent, and analyzing from stale impressions invalidates the whole passage:

  • OXY sold OxyChem to Berkshire for $9.7 billion in all cash on 2026-01-02, and is now near-pure upstream. On the surface its debt improved dramatically (long-term liabilities $20.6 billion → $13.5 billion), but Berkshire's $8.287 billion of preferred stock remains on the balance sheet, with an 8% rigid dividend ($680 million a year) not yet redeemed. Including it, OXY's adjusted net debt/EBITDA is 1.37x, the highest of the five (above DVN's 1.17x). "OXY now has the cleanest balance sheet" only holds if you ignore the preferred.
  • DVN completed its merger of equals with Coterra on 2026-05-07 (issuing 531.6 million shares). Three consequences: post-merger, only 37% of barrels-of-oil-equivalent is crude (it has become more "gassy"); about 207,500 barrels/day of crude in H2 2026 is hedged-capped at WTI $66–73, while WTI at $87.86 is already far above the cap, so the incremental elasticity on those barrels is zero; and Q2 has already measured it — unhedged realized price $95.10 vs $88.09 including hedges, giving up $7.01 per barrel. DVN's true oil-price elasticity in H2 2026 ranks 4th, not 1st, and will recover noticeably in 2027 once the capped volume falls to 95,863 barrels/day.

  ⑦ "Pure upstream has the greatest elasticity" does not hold in a supply-shock scenario. XOM's refining segment after-tax profit rose from $1.366 billion in 2025Q2 to $5.465 billion in 2026Q2 (+300%), and CVX's downstream from $737 million to $4.868 billion (+560%), both above the growth in their own upstream. A supply shock lifts crude and crack spreads simultaneously, and integrated players earn on both ends. The cost is that the most expensive part of XOM's and CVX's current share prices is refining, not upstream, and the key variable is whether crack spreads can hold Q2's extreme levels.

  ⑧ Natural gas is the disaster zone this round, and it is widely overlooked: Q2 Henry Hub averaged $2.90 (Q1 was $5.05); OXY's US realized gas price was −$1.48/Mcf (negative, on negative pricing at the Permian's Waha hub); DVN's unhedged gas price was just $0.35/Mcf, only 12% of Henry Hub. Post-merger, 40% of DVN's volumes are gas — that is both its current drag and its biggest option relative to the other four, but it is a natural gas bet, not an oil-price bet.

• Pre-market and technicals: 11/11 up, median +1.29%. What matters is position: XLE ($64.72 vs 52-week high $64.70), SLB ($60.75 vs $60.46), VLO ($363.00 vs $362.79), MPC ($378.12 vs $376.56) all trade above their respective 52-week highs in the pre-market. The sector is breaking out rather than bouncing, and this is the hardest technical evidence in this item.

  ⚠️ A necessary precision note: those 52-week highs are historical intraday highs excluding today, and pre-market trades do not formally constitute a 52-week high — only regular-session trades count. So the accurate phrasing is "the pre-market price has exceeded the prior high," and a genuine breakout must be confirmed after 09:30, which is exactly why §8 lists "XLE > $64.70" as an opening verification signal.

• 🚩 The biggest downside risk to this long thesis is not "oil falling back" but a deal that could be announced at any moment — its specific shape must be spelled out:

  • The deadline for ending the US-Iran war expired on 8/17, with the two sides deadlocked over control of Hormuz and Iran's frozen assets; there are currently no direct US-Iran negotiations, with Qatar and Pakistan passing messages.
  • But there is a separate and more nearly ripe channel: Iran and Oman are negotiating over reopening Hormuz. The core dispute is whether Iran can charge transit fees on passing merchant ships — Oman opposes any fee, and the US says any agreement containing a transit fee is unacceptable. Iran's foreign minister has also stated explicitly that "a deal with Oman does not mean Hormuz will reopen."
  • Reading: this is a situation of "negotiations advancing but stuck on one specific clause," not "no progress at all." That means the disappearance of the risk premium could be a single, overnight move that offers no chance to cover — once the transit-fee issue loosens, energy longs bought today get liquidated inside a gap. This is the core reason this report rates energy "watch closely" rather than "priority deep-dive."

• Final judgment (materially revised after fundamental verification): narrowed from "watch the whole sector closely" to "only OXY merits watching closely, the rest downgraded."

  • This remains the day's clearest direction with the best breadth (11/11) on the long side, but breadth is not odds: elasticities differ by 2.9x while two-day gains differ by only 1pp, which says the market is not pricing differentially — and that is exactly where differentiation can be made.
  • OXY is the only "highest elasticity + relatively smallest gain" combination (2.3 years capitalized vs XOM's 8.9 years). But its symmetry must be stated at the same time: an 8% rigid preferred, adjusted leverage of 1.37x (highest of the five), the highest breakeven oil price at ~$67, and almost no buyback cushion — the elasticity is symmetric, so it falls hardest too. That is neither a virtue nor a flaw, it is its attribute.
  • XOM is cut from the first draft's top recommendation to watch only: it has the lowest elasticity (1.17%) but has risen the most (8.9 years capitalized), and the most expensive part of the stock is refining.
  • Three hard constraints: ① the 8/31 lesson — the group's direction was entirely right that day, yet 6 names closed below their own pre-market prices, so wait for a post-open pullback rather than chasing at the pre-market price; ② size positions on the assumption that an overnight Hormuz deal gap could happen at any time; ③ remember that $90 is a step-down from Q2's $102.63 — longs in this direction are buying "further escalation," not "oil holding where it is."

⚠️ A convention inconsistency requiring self-correction (found in QC review): the first draft called "XLE/SLB/VLO/MPC all making 52-week highs in the pre-market" "the hardest technical evidence in this item" in this section, while writing in §9 that the same SLB had "pre-market volume of only $1.2M, cannot serve as evidence" — the same number is the hardest evidence in one section and inadmissible in another, which is selective enforcement of the threshold. Now applied uniformly by the threshold: of the four, only XLE ($14.8M) clears the $10M threshold; SLB ($1.2M), VLO ($2.4M) and MPC ($0.4M) do not, so this item's technical evidence reduces to XLE alone, with strength downgraded accordingly. By the same token, in the breadth evidence "energy 11/11 up," only XOM ($13.1M) and XLE clear the threshold, with the rest mostly at a tenth of the threshold in notional — that breadth should be treated as "reference," not "evidence."

③ EIX / PCG — California wildfire liability (avoid, binary event)

• Related news: ① SB 492 carries an urgency clause, so it can be voted on after the formal session ends; Seeking Alpha at 9/1 05:18 ET says it goes to a vote today. ② BofA double-downgraded this morning: EIX buy→neutral PT $81→$51; PCG buy→neutral PT $24→$13.
• Catalyst logic: what draws the most attention in the bill is that it does not include the $6 billion per-incident cap on draws from the state wildfire fund, and does not include provisions limiting insurers' subrogation against utilities — the upper bound of payout exposure is therefore unmeasurable, and that is what was being priced on 8/31 in EIX −23.07% and PCG −20.06%.
• Key reading: both of BofA's new PTs sit below the current share price (EIX pre-market $54.25 vs PT $51; PCG pre-market $13.35 vs PT $13). This is the sell side "catching up" with capitulation only after −20%; the information increment is low, but it confirms that no institution is catching the falling knife here. Pre-market EIX +0.50%, PCG +0.60%, almost no bounce.
• Final judgment: avoid. This is two-way risk, not a one-sided short opportunity: −23% has already priced "the text is very bad," and if today's vote or the final amendment is weaker than the market expects, short-covering will be violent. Until the vote result lands, any direction is a gamble.
• ⚠️ Status conflict disclosed: some media paraphrases say the bill "has passed," but the primary California leginfo bill history page still showed, at this report's read timestamp, a last action of 08/30/26 "Ordered to third reading," with no vote recorded. This report defers to the primary source.

4Palo Alto NetworksPANWearnings after today's close · watch only

• Consensus: adj EPS $0.98 (+3% YoY), revenue $3.35B (+32% YoY). Yahoo Finance¹⁰
• Three things that must be spelled out:
  1. "Revenue +32% was bought, not organic acceleration" — this first-draft judgment is directly contradicted by public data and is corrected here. PANW's F4Q ARR guidance of $8.90–8.95 billion implies about 28% organic growth, a clear acceleration from F3Q's 17%, and on a harder year-ago base. So the +32% revenue does contain substantive organic acceleration, not pure M&A stacking. What survives is: the +3% EPS scissor gap does come from integration and acquisition-related costs (that line reached $113M in FY26 Q3) — that is, "margins diluted by M&A" is true, "the growth was bought" is false.
     • A genuine bear data point the first draft lacked: on Jefferies's numbers, CyberArk's performance-vs-plan has fallen from 3.8% to 0.9%. That is a more specific and harder-to-rebut concern than "bought revenue."
  2. The consensus is the midpoint of company guidance, to the dollar. The Q4 guidance PANW gave on 6/2 (FY26 Q3 results) was revenue of $3.345–3.355 billion (+32% YoY), midpoint $3.350 billion; and the current so-called "Wall Street consensus" is $3.35 billion. They are the same number. In other words, the bar for "beating expectations" is a test the company wrote itself — if there is a "beat" tonight, the first question is by how many basis points, not whether. Company Q3 earnings release¹³
  3. The stock is only 4.2% below its 52-week high ($398.88); expectations are fully reflected.
• Final judgment: watch only. Holding overnight is betting on a quarter whose bar has been raised and whose margins are being diluted by M&A. But note the direction: if 28% organic ARR acceleration is delivered, this is a bullish quarter, and the bear case should rest on margins and CyberArk on a standalone basis, not on "the growth is fake."

5GoProGPROavoid / short watch

• Related news (including a timing correction that must be stated): YouTuber Mark Fischbach (Markiplier) holds 8.5% and is the largest single shareholder. But that stake was filed via a 13G on 8/20 with a 7/13 holdings date — meaning the fact had been public for 11 days before 8/31, during which the stock did not move. The real catalyst on 8/31 was the secondary spread from Bloomberg's coverage and related video content, not the ownership disclosure itself. And a 13G is a passive filing, with the filer explicitly disclaiming any intent to seek control — which if anything reinforces the "avoid" conclusion. The 8/31 regular session was +46.01%, with roughly 55% more after hours.
• Facts on the other side that must be stated (verified, and more serious than the phrase "going-concern doubt" suggests):
  1. PwC issued a revised audit opinion with a going-concern explanatory paragraph, and GoPro refiled its financial statements on 2026-06-01, formally acknowledging "substantial doubt." Note the sequence: the original 10-K's audit opinion contained no such item; the triggering event occurred after the 10-K was filed, and the revised opinion was added later via an 8-K.
  2. The most critical point: GoPro itself expects to breach the minimum liquidity, EBITDA and asset coverage covenants of its credit facility at the next test date. This is not "there may be difficulties ahead," it is an already pre-announced default.
  3. Cash has fallen from $102.8M to $49.7M; full-year revenue is $652M (−19% to −35% versus the 2023 peak of $1 billion), full-year loss $93.5M; the most recent quarter's revenue fell −31.3% YoY to about $105M, with camera units −38%.
• 🔗 One cross-connection worth noting: GoPro attributes its gross-margin deterioration explicitly to the memory cost crisis — saying memory prices rose 80%–115% in a single week during the last week of March 2026. In other words, the same memory price surge that took SNDK from $50 to $2,354 is precisely what crushed GoPro's gross margin. Today these two names are −2.79% and +75% respectively, opposite in direction but standing at two ends of the same supply chain.
• Pre-market: $1.535 (+75.19%), notional $145.1M, on pre-market volume of 94.53 million shares — the liquidity is real, the narrative is not. Market cap is only about $160 million. (The 08:20 ET recheck had already faded to +67.77%, and faded on rising volume.)
• Final judgment: avoid. One individual investor buying 8.5% does not change cash flow, does not change the audit opinion, does not change camera units at −38%, and certainly does not change the covenant breach at the next test date. This is a pure retail squeeze occurring in a company that has already pre-announced a financial covenant breach. Listed as a short watch, but with the caveat: the timing of such squeezes is unpredictable, borrow is scarce and extremely expensive, and entering a short mid-squeeze is not advised.

6RobinhoodHOODwatch closely · but with one internal contradiction

• Related news: Morgan Stanley upgrade to overweight, PT $124→$150.
• Pre-market: +1.98%, notional $94.0M — the only ratings-driven name today trading on real volume; DUOL ($7.1M) and AKAM ($0.5M) are both far below the credible threshold.
• ⚠️ A contradiction that must be flagged: HOOD's revenue is highly correlated with crypto trading, and the entire crypto chain is down today: MSTR −3.32%, COIN −2.66%, IBIT −1.16%, BITO −1.41%. HOOD is rising against its own fundamental driver, supported only by a rating. That weakens this item's logic hardness — if COIN keeps weakening after the open and HOOD follows it lower, treat that as the ratings premium being fully digested.
• Final judgment: watch closely, using COIN as the concurrent verification indicator.

7DuolingoDUOLwatch closely

• Related news: Evercore ISI's Mark Mahaney upgraded to outperform (from in line), PT $105→$210, anchored on 20× 2028 EBITDA, and raised 2027/2028 EPS estimates to 10%/25% above consensus respectively.
• Reading: the valuation anchor is set on 2028, and today is precisely the day when far-forward cash flows are punished hardest by rising rates — that is this item's internal tension. The stock is about 58% off its 52-week high of $353, so the expectation gap is genuinely there.
• Pre-market: +6.50%, but notional is only $7.1M, below this report's credible threshold, so gap-fill probability is high.
• Final judgment: watch closely, do not chase at the pre-market price.

⑧–⑩ Semiconductors/memory (MU / SNDK / MRVL) — watch only

• Related news: this report could not find any stock-specific negative news dated today. Reports found on SK Hynix, the CXMT listing, and a Broadcom-driven sell-off are all old news from late July to early August, and this report does not record them as today's catalysts.
• Reading: this is a positioning phenomenon, not a fundamental event. 25/25 down with a median of −2.15% is a highly uniform distribution — the uniformity itself says the driver is top-down (real rates + crowding), not bottom-up.
• A self-refutation (though the argument itself is flawed, stated here as well): the first draft reasoned that "if this were purely duration de-rating, low-P/E memory stocks should be the most resilient, but they fell as hard as software, so it is crowding deleveraging rather than duration." That premise does not hold: a NAND stock that went from $50 to $1,566 within a year has a low P/E that comes from peak-cycle earnings, not from valuation protection — this is the textbook shape of a low-P/E trap, and such names falling harder than software on a deleveraging day is the norm, not the anomaly. The conclusion (crowding deleveraging) may still hold, but it cannot rest on that argument. The cleaner support is the curve shape in §0①: the 2Y barely moved while the long end sold off, which says the shock comes from term premium rather than the policy path, and a term-premium shock hits precisely the most expensively valued, most crowded positions.
• Final judgment: watch only. Do not chase shorts (no fundamental deterioration), do not bottom-fish (yields have not peaked).

---

6. Bearish / Avoid List

---

7. Intra-Theme Ranking

Theme A: Crude supply shock (11/11 up)

This table has been reordered against the fundamental verification results, and the ranking is nearly the reverse of the first draft's. The "fundamental support" column is after-tax profit per +$10 of Brent as a share of market cap.

The tanker row needs a separate note: pre-market FRO +0.11%, STNG 0.00% (on just 154 shares) looks like "no reaction," but that is a misreading. FRO is already up 103% year to date and rose 6.24% on 8/31 alone; war risk premiums have gone from 0.25% of hull value pre-war to 3%–10%. The money on this leg was already made over the past eight months; today's flatness is "already repriced," not "no reaction."

Theme B: AI power (severe internal divergence, the most informative contrast today)

The first draft wrote here that "the ones with contracts rose, the ones without all fell; the market is buying contracts, not the theme." On recheck that statement does not hold and must be corrected — and the corrected fact is more interesting:

There were actually two large contracts today: Fervo's 396MW Google PPA, and Anthropic's roughly $35 billion cloud computing agreement with Lambda (with the compute sitting in Hut 8's 350MW Texas campus, NVIDIA holding the lease). The latter is far larger in dollar terms, yet the entire AI data center chain still fell across the board — CRWV −2.82%, NBIS −2.65%, IREN −2.61%, CIFR −3.03%, and even the directly benefiting HUT rose only 1.41% (on volume below threshold).

So the correct conclusion is not "the market buys contracts, not themes," but "the force of macro deleveraging overwhelmed micro contracts": a $35 billion contract could not hold this sector up. The one name that could rally, FRVO, is a small cap + jumping off 52-week lows combination, and its rise owes more to position than to "contracts can beat rates." This correction also lowers the degree of optimism in §5① on FRVO.

⚠️ Two more universal claims corrected: XLU at +0.05% is positive, so "everything without a contract fell" does not hold; and the first draft's line in §9② that "this theme had only this one name up today" also does not hold (both HUT and XLU are positive).

---

8. Opening Verification Signals

Pre-market (before 09:30)

• Can energy hold its 52-week highs: watch XLE > $64.70 and SLB > $60.46. If they fall back into the range right at the open, the breakout was manufactured on thin pre-market volume and should be downgraded.
• Can FRVO's volume keep expanding: the $34.4M notional should keep growing into the open; if turnover in the first half hour after the open cannot clearly exceed the pre-market cumulative, it was a news pulse rather than accumulation.
• Can the 10Y Treasury hold above 4.80%: this is the master switch for every asset today (4.792% pre-market, already at the door).
• Can the 10Y JGB hold 3.00% (intraday high 3.011%): this is the dividing line for whether this global duration sell-off is short or long, to be rechecked in the next Asian session.

Intraday (first 30 minutes)

• Do semis gap-fill: SMH is −1.76% pre-market. If more than half the decline is recovered within 30 minutes of the open, the "crowding deleveraging" judgment holds (rather than fundamental deterioration), and NOW on 8/31 was exactly that script (+4.16% rally off the open).
• 10:00 ET August ISM manufacturing (prior 55.6 / consensus 55.6): this is today's only macro data point, and it is the most direct test of this report's central thesis.
  • Beat (>55.6) → reinforces "the economy is hot enough, the Fed should hike" → yields up further → bearish growth, bullish energy, and this report's thesis is confirmed.
  • Miss (<55.6) → hike expectations retreat → relief for growth stocks, a demand-side negative for energy. Note that this is not "bearish for everything," it flips today's entire long/short direction.
• The subcomponent to watch most is Prices Paid rather than the headline index: it drives both hike expectations and the inflation trade, and is the common upstream of today's two main threads.
• The SB 492 vote: EIX/PCG can jump one way at any moment intraday.

Options sentiment

• VIX 15.88 (+6.43%) is still below 16 — under a combination of "war escalation + global bond sell-off," that level says the market is not pricing tail risk. That is both evidence for "no need to panic" and a hint that "protection is cheap."
• PANW earnings after the close: buying its options today means absorbing an overnight IV crush.

Risks

• Reversal after the gap: the core lesson of the 8/31 recap — of 15 names on the long side, 8 closed below their pre-market prices. Right direction ≠ making money.
• A lone riser: if FRVO has no peer follow-through (BE/OKLO/SMR are all down today), it is a single-stock event rather than a sector launch, and position size should be set for a single-stock event, not for a theme.
• Index futures diverging from Treasuries: currently the two are moving together (stocks down, bonds down); if stocks fall while bonds rally, the narrative has switched from "inflation/hikes" to "recession/safe haven," at which point the energy long thesis fails immediately.
• ⚠️ This report's only "unhedgeable" risk: Hormuz deal headlines. The Iran-Oman talks are stuck on the transit fee alone, and a deal could land at any moment (including while US markets are closed). It would hit both this report's energy longs and the "barrel" expression (USO) simultaneously, and would do so as a gap — this is a risk that cannot be managed with intraday stops, only with position size.

---

9. Final Conclusions

① The 5 names most worth watching today

② Today's 3 strongest themes

1. Overseas long-end term-premium shock (S) — core catalyst: UK 30Y at its highest since 1998, Japan 10Y crossing 3.00% intraday, German 10Y at its highest since 2011. Note the distinction from "the Fed hiking": in this report the 2Y is only +0.8bp while 10Y/30Y are +3.4/+3.5bp, so the day's marginal driver is term premium, not the policy path. Durability medium. Expression: short MU/MRVL/NOW, long XLV/XLP. The single variable most worth tracking is whether the JGB can hold 3% — it determines whether Japanese money keeps repatriating, and therefore whether this global duration sell-off lasts days or months.
2. Hormuz supply shock (S) — core catalyst: the US strike on Larak Island on 8/30, Brent $91.90. Durability medium, and it is "step-function" rather than "decaying" — the Iran-Oman reopening talks are stuck on the single transit-fee clause, and once a deal is struck the premium vanishes overnight in one move. Representatives: XOM/SLB/COP/DVN, with USO for pure β.
3. AI infrastructure/power (A−, but the direction is "contracts cannot hold the sector up") — there were in fact two large contracts today: FRVO's Google 396MW PPA (primary 8-K, contracted backlog +60%), and Anthropic–Lambda's roughly $35 billion cloud agreement (landing in Hut 8's 350MW Texas campus, with NVIDIA holding the lease). Yet the entire AI data center chain still fell across the board (CRWV −2.82%, NBIS −2.65%, CIFR −3.03%), and HUT rose only 1.41%. Representative: FRVO; the lesson this theme delivers today is that macro deleveraging overwhelmed micro contracts, not that "buying the contract is right."

③ Directions to avoid today

• Avoid "using the war narrative to buy safe havens": gold −1.82%, gold miners −2.72%. But note that this report has already downgraded this evidence — gold's decline shares a source with the rate thesis, it has fallen three days running, and it rose nearly 10% over the full month of August, so profit-taking is sufficient explanation. The conclusion is still "do not buy gold on war," but the reason is "gold is currently tracking real rates," not "the market is denying the war."
• Avoid DVN: on the surface "small-cap pure upstream = high elasticity," but post-merger only 37% is crude, and about 207,500 barrels/day in H2 2026 is capped at WTI $66–73 (current WTI $87.86 is far above the cap). Q2 already measured $7.01 per barrel given up. This is the easiest name to get wrong today.
• Avoid the crypto chain: MSTR −3.32%, COIN −2.66%. Note: IBIT at −1.16% is actually close to the index, so "crypto collapsing across the board" is an overstatement; what is genuinely weak is the leveraged proxy (MSTR).
• Avoid GPRO-style retail squeezes: a company that has already pre-announced it will breach three credit covenants, up about 156% in two days.
• Avoid chasing energy at pre-market prices: the 8/31 lesson is that this group was "entirely right on direction, flat on execution." Wait for the pullback, do not chase the gap.
• Do not short semiconductors: the decline is a positioning phenomenon rather than fundamental deterioration, and the bounce will come quickly.

④ Final one-sentence judgment

What is actually moving today is global long-dated sovereign debt — the UK 30Y at its highest since 1998, Japan's 10Y above 3% intraday — while the US 2Y barely moved, which says this is a term-premium shock rather than a repricing of the Fed path, and it has killed precisely the most expensively valued, most crowded assets. The war is real and escalating (Iran striking US bases in Jordan, Trump threatening to blow up Kharg Island), but its pricing has run down only the crude leg, with no spillover into gold, defense or freight. So today's opportunity is not "whether to buy energy" — $90 Brent is actually a step-down from Q2's $102.63, and is already in the price for pure upstream — but "which one to buy": the five integrated oil and gas names differ by 2.9x in oil-price elasticity, yet their two-day gains are all crammed into a one-percentage-point band, and that mispricing is the only executable thing here.

---

Internal Record (not sent to clients)

This report did not use yfinance. All single-stock/ETF pre-market data came through the stockanalysis.com quote API (api.stockanalysis.com/api/quotes/s/<ticker>), which had zero failures across 70+ tickers this run, and whose returned fields include ep/epd/ecp/epv/eu (pre-market price / previous close / percent change / pre-market volume / read timestamp), allowing direct reverse-computation checks; it is currently the most reliable channel for the pre-market slot. The memory entry "CNBC extended-hours fields are unusable pre-market" was confirmed again — this run did not attempt to pull single-stock pre-market data from CNBC and went straight to stockanalysis.

CNBC's bond change_pct field is again broken as a single field today, and over a wider range than the memory entry recorded:

• US2Y returns change_pct = -0.0195%, while the same record has last=4.358% > previous_close=4.350%, the opposite direction.
• New finding: the fault is not limited to Treasuries, sovereign bonds are affected too — UK30Y returns -1.105% while last=5.8646% > prev=5.7865%; DE10Y returns -0.1614% while last > prev; FR10Y, IT10Y and AU10Y are all the same type. That is, change_pct is untrustworthy for all of CNBC's bond symbols, not just Treasuries. The memory entry "CNBC Treasury changePct field breaks as a single field" should be expanded to "CNBC bond-class."
• All values were computed in-house as last − previous_close. Note that the bonds' change field itself is correct (it agrees with the in-house computation); only change_pct is broken.

Two traps encountered on the search side (both disclosed in the body as conventions, but the process is not sent to clients):

1. Year contamination (resolved): searching "9/1 economic calendar" returned "September 1 is Labor Day, markets closed, ISM pushed back" — that conclusion belongs to 2025 (2025-09-01 was indeed a Monday Labor Day). In 2026 Labor Day is 9/7. The ISM website's rob-report-calendar page 302-redirects to an SSO login page, so primary confirmation was not obtained; the Investing.com economic calendar was used instead and confirms a 9/1 release (prior/consensus both 55.6), noting that the site displays the time in UTC as "14:00," which must be converted to 10:00 ET — copying it directly would put the release 4 hours late. The body has been updated to verified status.
2. Old news as new news: searching "9/1 semiconductors fall" returned reports on an SK Hynix earnings miss, CXMT's Shanghai listing, and a Broadcom-driven sector sell-off, all of which check out as late July to early August 2026. The search engine itself flagged that "the results contain no specific information for 9/1." §5 explicitly states "this report could not find any stock-specific negative news dated today," and old news was not recorded as today's catalyst.

One primary-vs-secondary source conflict (SB 492): multiple secondary paraphrases say the bill "has passed," but the leginfo bill history page (primary) has a last action of 08/30/26 "Ordered to third reading," and Seeking Alpha at 9/1 05:18 ET writes "vote today." Handled per the primary source and disclosed both ways in the body. Note that leginfo may lag in updating, so this needs rechecking at recap time.

The hike probability is computed in-house, with method and assumptions as follows (the body gives only the conclusion; the derivation is preserved here for recap cross-checking):

• EFFR = 3.63% (New York Fed API, 2026-08-28 read, target range 3.50–3.75%)
• September fed funds futures @FF.1 = 96.295 → implied September average of 3.705%
• The September 2026 FOMC is 9/15–16, with the decision announced 9/16 (Wednesday) at 14:00 ET including the SEP dot plot — verified online, assumption holds.
• Assuming the decision on 9/16 takes effect 9/17 → 16 of 30 days at the old rate, 14 at the new; under a 25bp hike the monthly average is 3.74667%
• Two known weaknesses of this method: ① it assumes EFFR stays at 3.63% ahead of the meeting (mid-month reserve fluctuations would add a few bp of noise); ② it considers only "hike 25bp or hold," so if the market assigns nonzero probability to 50bp, this formula overstates the probability of 25bp. Both are within a few percentage points and do not change the conclusion that "a majority probability of a hike is already priced."
• P = (3.705 − 3.63) / (3.74667 − 3.63) = 64.3%
• Cross-check: the October contract @FF.2 = 96.20 → implied 3.80% → P ≈ 68% (an upper bound, not stripping out the October meeting); November @FF.3 = 96.14 → 0.92 hikes cumulatively priced through November. All three are mutually consistent.
• Media figures were 57% (KuCoin, earlier), 66% (Forbes, 8/31) and about 70% (Bloomberg, 9/1), trending up and in the same order of magnitude as the in-house 64.3%.

yfinance was rate-limited again on this machine (Too Many Requests, measured by a sub-agent at 12:04 UTC). The sub-agent used ps aux to confirm there were no leftover retry processes → this is an IP/CDN-level block, curl_cffi spoofing does not work, so no time was wasted retrying and it went straight to primary SEC EDGAR + stockanalysis. This matches the memory entry "two root causes of yfinance rate limiting," and this instance is the pure IP-block type. All of FRVO's financial figures therefore come from primary SEC filings (10-Q / 424B4 / 8-K / XBRL companyfacts), which is actually a higher evidence grade than yfinance.

One major self-correction on FRVO (must be recorded for the recap): the first draft scored FRVO at 81 points / priority deep-dive / expectation gap 9-10, on the basis of "52-week lows + the contract is unpriced." After fundamental verification it was cut to 59 points / watch closely / expectation gap 5-10. Where it went wrong: I treated "the stock is at lows" as evidence that "the news had never been priced." In fact the 3GW Google framework agreement was signed on 2026-03-19 and written into the IPO prospectus, and the 400MW capacity was named in the 8/12 earnings release — the version that was priced has already fallen from $42.65 to $14.60. This is the mirror image of "pre-disclosed ≠ repriced": being at lows does not mean unpriced; being at lows may be the result of the pricing. The first draft also took the media's "400MW" and "nearly 1GW" at face value, whereas the primary 8-K says 396MW firm + a 600MW non-binding option. Lesson: order figures paraphrased by media must be run through the primary 8-K before they go into a scoring table.

Sub-agent status: all three agents (FRVO fundamentals, energy complex fundamentals, risk QC) returned and were fully merged into the body. This report therefore went through a large revision, and the following three records of conclusions being overturned should be reviewed carefully at recap time:

① The energy ranking was entirely reversed by fundamental verification (largest impact). The first draft made XOM the top recommendation (76 points), on the grounds of "sector bellwether + best liquidity." After verification: XOM has the lowest oil-price elasticity of the five (every +$10 of Brent equals just 1.17% of market cap), and over two days it has already capitalized about 8.9 years of this oil move, making it the biggest relative gainer; OXY has the highest elasticity (3.37%, and it is a company-disclosed 10-K sensitivity) yet has capitalized only 2.3 years. Where it went wrong: I substituted "liquidity and market cap" for "earnings elasticity" in ranking longs, and those two things are nearly uncorrelated in an oil shock. There were also two structural changes that are purely stale knowledge and would inevitably be written wrong from memory: OXY sold OxyChem to Berkshire for $9.7 billion on 2026-01-02 (now near-pure upstream, but the $8.287 billion 8% preferred remains on the balance sheet, so adjusted leverage of 1.37x is actually the highest of the five); DVN merged with Coterra on 2026-05-07 (post-merger only 37% is crude, and about 207,500 barrels/day in H2 is capped at WTI $66–73, cutting elasticity from 3.08% to 1.99%). Lesson: any judgment involving "what business structure this company has" must be checked against the last two quarters' 10-Qs, never from memory.

② The implicit premise "$90 Brent is bullish" is wrong. 2026Q2 Brent averaged $102.63, while Q3 to date is only about $87.0. "Oil holding at $90" corresponds to Q3 earnings falling sequentially from Q2. The first draft treated rising oil as simply bullish throughout, without realizing the base period itself was higher. This is the same-type error as "percentage moves must be reverse-computed against their base," applied to a macro variable.

③ The intensity of the war was badly understated, and the central thesis was rewritten as a result. The first draft's headline sentence was "this is not a war tape," with the evidence "defense stocks not rallying, gold falling, tankers not moving." QC review found that the day actually saw Iran striking two US bases in Jordan with missiles and drones and Trump threatening to blow up Kharg Island, Iran's main crude export terminal — while the first draft wrote only "strike on Larak Island + a drone intercepted by the UAE." More seriously, there was a logical double standard: I read tankers as "already repriced" and defense stocks as "the market denying the war," choosing two opposite interpretations for the same price pattern based on whether it supported my conclusion (in fact defense stocks are equally already repriced — the war is on day 185, and LMT/RTX set all-time highs back in March). And on a relative basis defense stocks actually outperformed the index (ITA −0.17% vs SPY −0.62%), so the first draft read them on an absolute basis while reading XLV/XLP on a relative basis — the ruler was not consistent either. The conclusion has been rewritten from "this is not a war tape" to "war pricing has run down only the crude leg" — that is a conclusion about the transmission channel, not about whether the war exists.

④ Four other specific errors already corrected: the $10M notional threshold was selectively enforced (SLB's $1.2M was called "the hardest evidence" in §5 and "cannot serve as evidence" in §9); PANW's "revenue was bought, not organic acceleration" was directly overturned by the 28% organic growth implied by F4Q ARR guidance (an acceleration from F3Q's 17%); FRVO was described with a "52-week range" despite being a broken-issue stock that only IPO'd on 2026-05-14 with just 3.5 months of trading history; and the AI power theme's "everything without a contract fell" was falsified by two positive numbers, XLU (+0.05%) and HUT (+1.41%), while the larger Anthropic–Lambda $35 billion agreement that day was missed entirely — it was same-day and also WSJ-originated alongside Fervo's PPA, and only the latter was picked up.

⑤ Gaps still open (must be checked at recap): the per-barrel sensitivities for XOM/CVX/COP/DVN are estimates rather than company disclosures (backtested on OXY, the method has an 11% error, biased optimistic by roughly a tenth), and each company's investor-day materials need checking; OXY's implied Brent of only ~$73 is markedly below the other pure upstream names COP ($90) / DVN ($93), and it cannot be determined whether the market is genuinely more pessimistic or the forward EPS data itself is flawed; and there are two conventions for the ISM consensus, 55.0 and 55.6, with this report using the latter. Also: the data.sec.gov companyfacts API silently lags for XOM (updated only through 2026-05-04, while the Q2 10-Q was filed on 8/3; the API does not error, it simply withholds the new data) — a bulk pull would silently be one quarter short, so this was switched to reading the filings directly.

---

⚠️ Risk disclosure: this list is pre-market information gathering and observation 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; automatically generated content may contain timeliness gaps or factual errors, so company disclosures/SEC filings prevail and this must not be used directly as a basis for trading.