US Pre-Market Brief | 2026-08-11 (ET) Tuesday

Coverage window: 2026-08-10 (Monday) 16:00 ET regular-session close → 2026-08-11 08:20 ET, covering Monday after-hours earnings (16:00–20:00 ET), the overnight Asia/Europe sessions, and this morning's pre-market. Quote conventions:

• For single stocks and ETFs, "prior close / daily change / pre-market price / pre-market change / pre-market volume / 52-week high-low" come from the stockanalysis.com quote API, snapshot 08:09 ET (a few names at 08:16–08:17 ET, flagged in the tables). Prior close is the 8/10 16:00 ET regular-session closing price.
• Indices, crude oil, gold, Treasury yields, VIX, the dollar index and bitcoin come from the CNBC quote API, snapshot 08:08–08:19 ET.
• Equity index futures come from Yahoo Finance, snapshot roughly 08:00 ET, a different source from the rest of the quotes and about 10–20 minutes earlier — do not mix them with single-stock pre-market data.
• Earnings figures come from company press releases, SEC filings and third-party reproductions; the RIOT / PLUG / FSLR data in the fundamentals subsections was checked item by item by sub-agents against SEC XBRL (companyfacts) and the 10-Qs, with the source tier noted in the text. Data this brief could not obtain: ① the 9-day VIX (VIX9D) is still the 8/10 close of 12.77 and was not updated this morning, so the "short-end volatility premium" item can only carry over yesterday's reading; ② CRWV / SMCI / LITE report tonight (after the 8/11 close), so this brief can only give the calendar, market expectations and verification points, not results; ③ option implied volatility and call/put skew were not obtained, so the options-sentiment part of Section 8 can only cite media-reported implied move magnitudes; ④ pre-market volume in sector ETFs is extremely thin (XLY 9 shares, XLC 11 shares, XLB 271 shares), so pre-market sector direction has no statistical meaning — stated in Section 1.3.

---

0. One-Sentence Summary

Today's hardest catalyst is not in AI chips but in "the land that powers AI" — RIOT converted a bitcoin mine into a 20-year, $9.1 billion AI data-center lease, +13.9% pre-market; but what actually decides this afternoon and tomorrow is tonight's three post-close earnings from CRWV/SMCI/LITE and tomorrow's 8:30 CPI.

1. Strongest catalyst: RIOT's 191MW / $9.1B data-center lease (company 8-K, 8/10 16:05 ET). Initial 20-year term through June 2048; including two 5-year renewal options, total contract value up to $16.1B; the tenant is disclosed only as "one of the world's leading frontier AI labs"; Bloomberg, citing people familiar, reported the counterparty is Anthropic, and neither Riot nor Anthropic has confirmed it. On the same day, Q2 revenue of $174.2M beat expectations. This is today's only S-tier item with "driver type = the company's own major contract disclosure."

2. Strongest theme: AI power and data-center capacity (not AI chips). Besides RIOT, Babcock & Wilcox (BW) is +37.2% pre-market — Q2 revenue $319.7M (+130% YoY, beating consensus by 62%), GAAP net income turning positive at $14.3M (−$58.5M a year ago), a pipeline of over $14 billion including 4–6GW of power-generation opportunity, and full-year adjusted EBITDA guidance raised to $80–105M. ⚠️ One layer has to be peeled off first: the widely quoted "$2.7B of new bookings in the first half, backlog +533%" is an H1 figure, and the overwhelming majority of it was already disclosed in the Q1 results on 5/11 — Q2 standalone bookings were only $151.0M. Today's new information is revenue and guidance, not orders. Two unrelated corporate disclosures on the same day both point to the same industrial line: power and sites are the bottleneck for AI.

3. Flow direction: yesterday semis were dumped and energy was chased; this morning some of that is being taken back. Yesterday SMH −2.28% (worst of the field) and XLE +4.66% (strongest of the field); this morning pre-market SMH +1.02%, XLE −0.22%, with energy names XOM −0.50% / CVX −0.52% / OXY −0.34% all giving back a little, while NVDA +1.28% / AMD +0.86% / MU +0.99% / AMAT +2.28% rebound together. The new fundamental evidence supporting semis is TSMC's July revenue released on 8/10: NT$467.58 billion, +44.7% YoY, an all-time record, +5.6% MoM (first seven months cumulative +37.0%, in New Taiwan dollars) — but that datapoint failed to stop SMH's slide yesterday, which tells you this is not a "fundamentals-priced" stretch of tape. ⚠️ Two things must be stated together: ① this TSMC item was released pre-market ET on 8/10, strictly outside this brief's coverage window; it is listed to explain the fundamental backdrop for semis and should not be treated as a new catalyst for this morning; ② the cumulative +37.0% is in New Taiwan dollars and cannot be compared directly with the company's "USD-denominated growth slightly above 40%" guidance; this brief did not bridge the currencies.

4. Driver-type distribution: today is overwhelmingly an "earnings + guidance" day; macro is tomorrow's business. Overnight, 8 companies showed material pre-market moves on earnings or guidance, and among them all four on the bearish side were driven by guidance rather than the quarter's own numbers: ONON (cut constant-currency revenue growth guidance, −17.0%), UPWK (cut full-year revenue and EBITDA guidance, −18.2%), LIF (full-year guidance left unchanged, −22.5%), HIMS (cut the top end of full-year EBITDA guidance, −6.6%, while the quarter itself was a double beat). The four on the bullish side each had a different driver: RIOT a major-contract 8-K (+13.9%), BW a large revenue beat plus a guidance raise (+37.2%), PLUG a narrowing loss (+12.3%), SE a double beat on quarterly revenue and EPS (+6.8%). This contrast is itself the most important information today: the market is punishing "guidance unchanged" even harder than it punishes "a quarterly miss" — LIF had revenue +38%, adjusted EBITDA +53% and MAU crossing 100 million, and fell a fifth solely because full-year guidance was left unchanged.

5. Pre-market state: indices barely moving, violent dispersion underneath. S&P futures 7,790.00 (+0.17%), Nasdaq futures 29,832.75 (+0.32%), Dow futures 54,110 (+0.09%), Russell futures 3,032.50 (+0.25%); on the ETF read SPY +0.17% / QQQ +0.31% / DIA +0.10% / IWM +0.20%. 10-year Treasury 4.694% (−0.4bp), 2-year 4.226% (−1.3bp), the curve steepening slightly; VIX 15.47, essentially unchanged; dollar index 99.856; WTI $82.17 (+0.05%), Brent $87.58 (−0.16%) — oil stopped after yesterday's +5%. Overnight Asia/Europe: Hang Seng −1.10% (close at 16:08 HKT), the three main European indices up marginally 0.01%–0.27% (European midday); the latest snapshot date for the Nikkei 225 is still 8/10; today is Japan's "Mountain Day" holiday and Tokyo is closed (inferred from the snapshot date and the calendar, not verified line by line against exchange notices).

6. The action requiring the most restraint today is chasing pre-market gaps. The quantitative conclusion from yesterday's recap must be carried over verbatim: of the 25 reconciliation names on 8/10, 21 closed below that morning's pre-market price and only 4 above — the pre-market price was very nearly the high of the day. Today's BW at +37%, ACVA at +17% and FRMI at +15% pre-market all belong to the same category of "prices set on extremely thin pre-market liquidity"; every ranking in Section 3 of this brief is ordered by catalyst quality rather than pre-market gain, for the reasons in Section 8.3.

---

1. News Overview

1.1 Main Items (by impact level and time)

⚠️ Coverage-completeness statement: this brief's sell-side ratings dimension is incompletely covered — the only analyst rating change confirmed today is Baird's FSLR upgrade. Other rating changes could not be obtained due to scraping limits. Therefore wording such as "all" and "every" in Section 6 and Section 9.3 should be read as "within the scope this brief covers," not as an exhaustive sweep of the whole market.

1.2 Macro and Rate Environment (read this first: this is a "hiking" cycle, not a cutting cycle)

If this one is read backwards, the direction of the entire brief goes wrong:

• The July FOMC (7/29) held, but with three hawkish dissents — Cleveland's Hammack, Minneapolis's Kashkari and Dallas's Logan all argued for a 25bp hike. Hammack said again recently that multiple hikes may be needed to contain inflation.
• Last Friday (8/7) July nonfarm payrolls unexpectedly fell by 23,000, with wage growth at a post-pandemic low. The role of that weak jobs print is to reduce the urgency of a September hike, not to open a window for cuts — the market's pricing for this year is currently "one more hike."
• CME FedWatch pricing for the 9/16 FOMC is close to a coin flip (the readings retrieved swung between "unchanged 45.1%" and "hike 61.9%" with inconsistent as-of dates, so this brief does not cite specific probability numbers, only the direction: a September hike is still a live option).
• Tomorrow's 8:30 ET July CPI is the only tier-one event this week. Consensus headline 3.4% / core 2.5%, both slightly lower than June.

Rates and volatility snapshot (08:16–08:19 ET):

How to read it: the 2-year fell more than the 10-year, meaning the rates complex is mildly pricing "CPI won't run too hot"; but the 30-year barely moved and 2s10s kept steepening, so the term premium has not come back down. This is a "wait for the data, take no view" bond market. VIX at 15.47 barely moved either — the market is not buying much insurance for tomorrow's CPI, and that is itself an asymmetry: if CPI comes in hot, existing protection is inadequate.

1.3 Sectors and Commodities (⚠️ pre-market sector signal unusable)

A convention issue that has to be spelled out: of the 11 SPDR sector ETFs, 3 have pre-market volume in the single digits to the hundreds — XLY has only 9 shares. A "pre-market sector change" computed from volume like that carries no information at all; it is simply the price of the last resting order. This brief therefore does not rank sectors pre-market, and all theme strength in Section 2 is based on the news itself and on single-stock pre-market volume, not on ETFs.

Commodities and crypto (08:08–08:19 ET):

Oil stopped dead after yesterday's +5%. That matters: yesterday's +4.66% in energy was driven by "the expected Hormuz reopening failing to materialize + the Houthi attack on Saudi Aramco's Jizan refinery," and today neither has any incremental news, so oil is flat and energy names are giving back a little pre-market. Energy is probably not today's leading sector but the sector where "what to do with yesterday's position" is the question.

Bitcoin at $64,351, down about 44% from the September 2025 high of roughly $114,100. That backdrop determines that RIOT's +13.9% today is not a crypto move — MSTR $97.33 (52-week high $414.36) and COIN $148.68 (52-week high $402.16) are both still deep in the hole. What RIOT is rallying on is a new identity: "AI data-center developer."

---

2. Strongest Themes, Descending

---

3. Overall Single-Stock Strength Ranking (descending by total score; the bearish side is listed separately in Section 6)

Quote snapshot 08:09 ET (names marked ※ are 08:16–08:17 ET). "52-week position" = (pre-market price − 52-week low) / (52-week high − 52-week low).

3.1 Bullish Side

3.2 The Three Reporting Tonight (a separate block: no scores, only event parameters)

---

4. Single-Stock Scoring Model (100 points)

4.1 Component Detail for the Top 8

Several deliberate judgments, with reasons written down for later review:

• BW's "expectation gap" was cut from 10 in the first draft to 7, the total from 64 to 59, and its rank from 4th to 5th (FSLR moved up). Reason for the correction: the 62% revenue beat really is the largest expectation gap this brief has seen, but the endlessly quoted "H1 new bookings $2.7B / backlog +533%" was overwhelmingly disclosed back in the Q1 results on 5/11, and Q2 standalone bookings were only $151.0M — that part is not new information today and should not count toward today's expectation gap. Earnings elasticity was also cut from 14 to 12: Q2 adjusted EBITDA was only $21.8M against full-year guidance of $80–105M, implying $50–70M in the second half, and that slope is itself a risk. Add in the +37.16% pre-market move, a market cap of only $1.3 billion and a 52-week range of $1.171–$22.03, and tradability claws back almost all of the points the expectation gap earned.
• RIOT's "earnings elasticity" gets only 8 points despite being today's headline. Because adjusted EBITDA this quarter was −$69.7 million and gross margin fell from 40.5% to 23.7%. Almost all of its score comes from "catalyst directness 20" and "persistence 9" — i.e. the market is buying 2028's rent, not 2026's operations.
• MARA/CIFR/WULF/APLD all get 0–5 on "catalyst directness." They have no company news today and are merely RIOT's peers. That is exactly yesterday's lesson — the four optical-module names were pulled up together by a "theme" and then knocked back down together.
• VG's "expectation gap" was cut from a deserved 9 to 5. The $500 million guidance raise is real, but the stock already rallied 7.54% yesterday and is down 4.49% this morning — the market has already traded this report.

---

5. Detailed Analysis of the Top Names

5.1 SE — Sea Limited (total 78, priority deep-dive)

• Related news: 8/11 pre-market (call at 07:30 ET), Q2 revenue $7.8047 billion (consensus $7.2745 billion, a beat of $530 million), EPS $0.86 (consensus $0.79). The company said Shopee and Monee simultaneously strengthened market leadership and raised user penetration, and that it will "invest prudently to serve more users and achieve profitable growth." Source⁴
• Catalyst logic: this is a double beat on revenue and EPS, not a "one-off beat." The item with the biggest valuation impact is EPS — SE has moved from "burning cash to buy share" to "delivering share and profit at the same time," which is a story about an improving valuation denominator, not a sentiment story.
• Theme and stage: SE Asia e-commerce + digital finance, fermenting stage. The 52-week position is only 37.5% (pre-market $122.63, range $77.05–$199.30), leaving 62% of room to the year's high, so it is not a crowded trade.
• Fundamental verification: ⚠️ This brief did not independently verify SE's earnings quality — gross margin, operating cash flow, net cash and similar metrics were not obtained, only press-release-level revenue and EPS. This is the biggest coverage gap in this brief: it ranks #1 overall, yet the strength of the evidence behind that judgment is lower than for RIOT/PLUG/FSLR below, which were checked item by item.
• Pre-market and technicals: +6.82%, pre-market volume 868,788 shares — one of the most liquid bullish names today. Ahead of the report options implied a single-day move of about ±19%, and the current +6.82% has realized only about 1/3 of it, which suggests pre-market pricing is relatively restrained rather than "already fully gapped."
• Final judgment: Priority deep-dive. Three things hold simultaneously: ① the catalyst is its own double-beat earnings, not theme linkage; ② pre-market liquidity is ample, not thin-volume pricing; ③ the 52-week position is low and the implied move is not fully realized, so the expectation gap is not exhausted. The biggest risk is that this brief did not independently verify earnings quality — how much of the EPS beat came from one-offs in the gaming business (Garena) and how much from Monee's credit-provisioning conventions is a question this brief cannot answer.

5.2 VG — Venture Global (total 71, watch closely)

• Related news: 8/11 pre-market, Q2 revenue $4.6 billion (+48%), net income attributable to common shareholders $1.3 billion (+266%), exports of 127 cargoes / 466.4 trillion British thermal units, implying a weighted-average fixed liquefaction fee of $6.45/MMBtu; FY26 adjusted EBITDA guidance raised to $8.7–9.1 billion (from $8.2–8.5 billion), cargo guidance narrowed and raised to 500–518; new or expanded long-term contracts with TotalEnergies (0.85 MTPA / 5 years), EnBW (about 0.82 MTPA / 5 years), Vitol (increased to 1.7 MTPA) and Atlantic-SEE (20-year SPA increased to 1.0 MTPA). Source¹¹
• Catalyst logic: the guidance midpoint was raised by about $550 million (roughly +6.7%), and the long-term contracts lock in visibility on future cash flow. On report quality alone, this is one of the best of the day.
• But the market's reaction is the opposite, and that is the point: VG already rallied 7.54% yesterday (against a backdrop of Hormuz tension + energy sector +4.66%), and after the report this morning it is down 4.49% pre-market (462,782 shares, not thin). This is a clean sell-the-news: the good news was already traded yesterday.
• Theme and stage: LNG/energy, crowded stage. 52-week position 66.4%.
• Fundamental verification: this brief did not run a sub-agent; only press-release numbers are available. How much of the +266% net income came from higher liquefaction fees, how much from cargo volume and how much from one-off items was not obtained.
• Pre-market and technicals: −4.49%; if it keeps probing lower at the open, the combination of "guidance raised by $500 million yet the stock falls" would itself create a second opportunity — but only if it drops back to where it was before yesterday's rally (around $13.26).
• Final judgment: Watch closely, but don't chase pre-market and don't chase the drop at the open. This is a name with a very high fundamental score and a very low timing score. Verification points: if the decline narrows within the first 30 minutes on rising volume, yesterday's gain was front-running rather than over-reaction; if it breaks $13.26 on volume, the market believes the guidance raise is not enough to support yesterday's move.

5.3 RIOT — Riot Platforms (total 68, watch closely)

• Related news: 8/10 16:05 ET, company press release + SEC 8-K. A data-center lease for 191MW of critical IT capacity, with the tenant described as "one of the world's leading frontier AI labs" (unnamed); initial term of 20 years through June 2048, initial contract revenue of about $9.1 billion, total contract value up to about $16.1 billion including two 5-year renewal options; cumulative NOI guidance $7.3–8.2 billion, averaging $365–411 million per year; 96MW delivered in December 2027 and the full 191MW at run-rate in June 2028; Morgan Stanley providing $573 million of bridge financing. CEO Jason Les said the company has "signed 241 megawatts in a little over six months, representing approximately $9.8 billion of long-term contracted revenue." Bloomberg, citing people familiar, reported the tenant is Anthropic, but neither Riot nor Anthropic has confirmed it, and this brief treats it as "unnamed."
• Catalyst logic: this is an identity re-rating from "coin-price beta" to "infrastructure cash flow." The item with the biggest impact is not current revenue but the valuation method — a company with 20 years and $9.1 billion of contracted rent should in theory be priced on discounted REIT/infrastructure cash flows, not on its bitcoin holdings.
• Fundamental verification (SEC XBRL companyfacts CIK 1167419 + 10-Q, checked item by item by a sub-agent):

• ⚠️ One accounting fact that has to be written down: RIOT discloses bitcoin fair-value gains and losses separately in XBRL. Across the 6 quarters the sub-agent checked one by one (Q1'25, Q2'25, Q3'25, FY2025 full year, Q1'26, Q2'26), every GAAP-profitable quarter's profit was driven by the bitcoin remeasurement, and the remeasurement gain exceeded net income itself — i.e. stripping out the remeasurement, all 6 observation periods were losses. ⚠️ This is a statement about the checked periods, not a universal claim about the company's entire history. Behind Q2 2025's +$219 million of net income sits a +$471 million coin-price remeasurement masking roughly −$250 million of actual operating loss; this quarter's −$74.61 million remeasurement loss corresponds to −$237 million of net income, or about −$163 million excluding it. For any future quarter that "swings to a profit," the first thing to do is check this line.
• Where the cash came from: Q2 cash rose from $205.7 million to $471.4 million, while operating cash flow was −$90.10 million and capex −$85.92 million — the difference was entirely filled by $443 million of coin sales. Coin holdings fell from 19,273 to 11,380 in a year (−41%). Mining cost: $49,912/coin excluding depreciation, $90,631/coin including depreciation; against the current coin price of $64,352, it still makes money on a cash basis but loses about $26,000 on every coin mined on an all-in basis.
• Share dilution (an important positive correction): over five years the share count rose from 84.10 million to 375.26 million (+346%), but from 369.62 million on 2025-07-29 to 375.26 million on 2026-08-07 is only +1.53% YoY, and February to August 2026 was a net decrease. This historical dilution burden really has been held down in 2026, at the cost of switching to coin sales and project debt. Potential dilution remains: $594.4 million of converts maturing in 2030, with a fair value of about $1.2 billion at 6/30 (deep in the money), convertible from Q3 2026.
• Valuation: pre-market market cap ≈ $22.09 × 375.26 million = $8.29 billion; TTM revenue $674.5 million → P/S 12.3x; P/B 3.79x; P/E not applicable.
• 【 Sub-agent discount calculation, not company disclosure 】: average annual NOI of $385 million, 20 years starting in 2028, discounted back at 8% gives a present value of about $3.2 billion; less $2.1–2.3 billion of capex, net value added is about $1.0 billion; at a 10% discount rate it is about $500 million. And this morning's pre-market market cap increase is about $1.04 billion. In other words, this one gap roughly prices in the entire 191MW lease at the most optimistic discount rate in a single move.
• Pre-market and technicals: +13.87%, pre-market volume 1,850,576 shares — the second most liquid bullish name today. 52-week position 57.4% (pre-market $22.09, range $11.04–$30.32).
• Final judgment: Watch closely, not priority deep-dive. The news is S-tier and fully verifiable, which is not in dispute; the only reason it sits behind SE and VG is that the valuation has already discounted the best case, and there are still about 6 quarters of pure cash burn in between. The biggest risk is not "the contract isn't real" but "the financing chain between signing and collecting rent": $2.1–2.3 billion of capex is required while operating cash flow has been negative for at least 10 consecutive quarters, cash is $471 million, and bitcoin holdings are down to $666 million after selling 41%. $1.7–2.1 billion of project debt + $573 million of bridge + continued coin sales — if any link breaks, or the coin price falls another leg (shrinking coin-sale proceeds directly), the company is forced back to equity financing — and equity financing is exactly the old road that diluted this company 4.5x over the past five years.

5.4 BW — Babcock & Wilcox (total 64, watch closely)

• Related news: 8/10 after hours. Q2 revenue $319.7 million (consensus $196.97 million, a 62% beat; +130% YoY), GAAP EPS $0.07 (est. $0.03), GAAP net income turning positive at $14.3 million (−$58.5 million a year ago); FY26 adjusted EBITDA guidance raised to $80–105 million; total pipeline >$14 billion including 4–6GW of generation opportunity, driven mainly by the "Base Electron" project.
• ⚠️ On the orders, a date check is mandatory: first-half new bookings of $2.7 billion (+1,058% YoY) and end-of-Q2 backlog of $2.6 billion (+533%) are first-half cumulative figures, the overwhelming majority of which were already disclosed in the Q1 results on 5/11; Q2 standalone bookings were only $151 million (+38%). These numbers therefore do not constitute today's expectation gap, and this brief has accordingly cut BW's expectation-gap score from 10 to 7 (see Section 4.1).
• ⚠️ Another number that must be read alongside: Q2 adjusted EBITDA was only $21.8 million, against full-year guidance of $80–105 million — implying $50–70 million in the second half, several times the first half. That slope is itself the biggest risk in this guidance. Source²¹
• Catalyst logic: it is the other end of the same industrial line as RIOT — RIOT provides the site and power interconnection, BW provides the generation and thermal equipment. Two independent disclosures landing on the same day are what makes today's "AI power" theme hold. For BW the item with the biggest impact is backlog +533%: it stretches a cyclical equipment maker's revenue visibility from one quarter to several years.
• Theme and stage: AI power, launch stage (this is the first time the theme has been priced into BW specifically).
• Fundamental verification: ⚠️ This brief did not run a sub-agent, and BW's balance sheet was not obtained. That is a major gap — BW has been highly levered for years and restructured multiple times, and a $2.6 billion backlog at a company with a fragile balance sheet is both an opportunity and a risk (executing large orders requires working capital). Market cap is only about $1.3 billion against a backlog of $2.6 billion, an orders/market-cap ratio of 2x — a ratio that itself says the market previously did not believe these orders could convert into profit.
• Pre-market and technicals: +37.16%, pre-market volume 607,940 shares (not thin for a $1.3 billion company). 52-week position 52.7%, but the range is $1.171–$22.03 — an 18-fold rally and a round trip within a year; this is a structurally high-volatility name.
• Final judgment: Watch closely, but explicitly do not chase the first print. The expectation gap is the largest on the board today (a 62% revenue beat), the catalyst is fully verifiable and the theme is right; every deduction is on tradability: already +37% pre-market, small cap, unverified balance sheet, orders concentrated in a single project. Verification points: if it can hold above $11.50 in the first 30 minutes on volume, institutions are absorbing; if it falls quickly below $10, that is a textbook case of thin-volume pre-market pricing being broken by opening liquidity.

5.5 FSLR — First Solar (total 63, watch closely)

• Related news: 8/11 pre-market, Baird upgraded from Neutral to Outperform with the price target raised from $205 to $318 (implying +32.9% versus yesterday's close). Reasoning: the utility-scale solar market has tailwinds at both the fundamental and share-price level; Section 232 polysilicon trade policy could remove the order overhang from 2029/30 onward. Also in the period, HSBC upgraded to Buy on 8/7 and Mizuho set a $324 target.
• One thing must be cleared up first: today's +3.62% has nothing to do with earnings. FSLR's Q2 was released in late July and already digested, and yesterday (8/10) the stock actually closed down 4.29%. This morning is purely a sell-side re-rating, containing no new fundamental information. So the right question is not "do the fundamentals support today's move" but "do the fundamentals support $318."
• Fundamental verification (SEC XBRL CIK 1274494 + 10-Q, checked by a sub-agent):

Revenue falling while profit surges is a story about price and mix, not volume: the +11.7pp in gross margin comes mainly from a higher share of US domestic capacity, which brings 45X advanced-manufacturing tax credits into gross profit. First-half operating cash flow of −$359.8 million looks alarming, but it is seasonally normal (−$458.4 million in the same period last year, +$2.0571 billion for the full year), and it is $98.6 million better than a year ago, so it is not a risk signal.

• ⚠️ The real risk is in the order book: contracted backlog fell from 50.1GW at the end of 2025 to 45.1GW (−5.0GW), worth $13.6 billion, implying an average price of $0.302/W. Sub-agent estimate: after deducting roughly 10GW of shipments in the second half of 2026, only about 8.8GW per year remains for 2027–2030, roughly half of the 17.0–18.2GW shipment guidance for 2026 (half of which is 8.5–9.1GW, and 8.8 falls inside that range). ⚠️ This is an estimate reverse-engineered from "H1 revenue = 42% of full-year guidance," not a company-disclosed year-by-year delivery schedule. The CFO has explicitly acknowledged that customers are waiting for the Section 232 outcome before ordering.
• What $318 implies in multiples: at the $318 target price the implied market cap is $34.26 billion (elsewhere in this brief "market cap" is current price × share count; the convention differs here and is flagged accordingly); FY2026E P/E of about 17.9–19.8x, EV/adjusted EBITDA of about 11.7–12.6x, P/B 3.32x. The absolute levels are not aggressive (for a manufacturer with a 57% gross margin and $1.5 billion of net cash), and $318 is only 0.9% below the 52-week high of $320.95, so it is essentially "back to the year's high" rather than a breakout beyond the historical valuation band. But on 2027E EPS of $27.64 (Zacks) it is only 11.5x — the gap between those two numbers shows that the entire tension in this target rests on "can 2027 EPS jump from about $17.7 to about $27.6 (+56%)," and the current order book does not support that slope.
• Pre-market and technicals: +3.62%, but pre-market volume is only 22,484 shares — extremely thin, so this price is unreliable. 52-week position 49.5% (pre-market $248.00,52-week range $176.47–$320.95).
• Final judgment: Watch closely. The fundamentals are the most solid of the three names verified in this brief (the only one with real profits, the only one with net cash, the only one with zero dilution), but today's catalyst is a sell-side rating (authority 10 points rather than 15), pre-market volume is extremely thin, and the load-bearing point of the target price is a Washington policy event rather than something the company controls. Verification points: the timetable for Section 232 polysilicon policy, and whether next quarter's backlog stops falling and turns back up.

5.6 AMAT — Applied Materials (total 58, watch only)

• Related news: none of its own. The indirect catalyst is TSMC's July revenue released 8/10: NT$467.58 billion, +44.7% YoY, +5.6% MoM, a single-month record, with January–July cumulative +37.0%, faster than the company's full-year guidance of "USD revenue growth slightly above 40% in 2026."
• Catalyst logic: TSMC's monthly revenue is actually realized revenue, not a forecast, and is one of the highest-quality high-frequency datapoints in the semiconductor chain. Rising capex intensity → equipment orders. But the transmission chain is long, so catalyst directness gets only 10 points.
• Theme and stage: semis, divergence stage — fundamentals strengthening, price weakening. Yesterday SMH −2.28% was the worst of the field and AMAT itself was −3.16%.
• Pre-market and technicals: +2.28% (08:16 ET), pre-market volume 33,950 shares (on the thin side). 52-week position 64.7%.
• Final judgment: Watch only. The reason is not that the logic is wrong but that the timing is — AMAT reports after the close on 8/13. Buying AMAT today on the strength of an industry datapoint means taking on the risk of an earnings report two days early. This is a name you can wait on: see the 8/13 results and guidance first, then decide whether to play the TSMC line.

5.7 PLUG — Plug Power (total 48, watch only)

• Related news: 8/10 after hours, a Q2 loss smaller than expected. Revenue $178.29 million (only +2.5% YoY), gross margin improved from −30.7% to −0.9%, operating loss narrowed from −$176.95 million to −$64.11 million, operating cash flow improved from −$191.81 million to −$94.06 million; full-year revenue guidance raised to +15%~16%.
• ⚠️ The sub-agent's key finding: about 1/3 of the narrowing loss is one-off. The operating loss improved by $112.84 million YoY, which breaks down as:
  • Gross profit improvement +$51.78 million (structural, credible) — a continuous slope of Q4'25 +2.4% → Q1'26 −13.2% → Q2'26 −0.9%, not a single-point jump.
  • Opex decline +$61.06 million (mostly one-off) — the CFO explicitly explained that $39.7 million of it is the recovery of previously impaired assets, chiefly a $37.0 million gain from a customer contract dispute settled in June, and said clearly it is not repeatable. Excluding it, real Q2 opex was about $102 million and the YoY decline narrows from −49.4% to about −17%; management's normalized opex run-rate of about $75 million per quarter is also above the reported $62.4 million.
  • Conclusion: on a comparable basis the Q2 operating loss should be read as about −$104 million, not −$64.11 million.
• The −$188.2 million net loss should be ignored: the operating loss was only −$64.11 million, and the $124.1 million difference is fair-value movement on converts and warrants — purely non-cash, and moving opposite to the share price (stock up → GAAP loss widens). This is the same class of problem as RIOT's bitcoin remeasurement, only in the opposite direction.
• The biggest risk is runway and the H2 assumption being bound together: unrestricted cash is only $161.89 million, which on Q2's net burn of $61 million is about 2.7 quarters; and the premise of "positive EBITDAS in Q4 2026" is that second-half revenue jumps about 40% above the first half (full-year guidance $817–824 million, H1 actual $341.8 million, so H2 needs $475–482 million) — while first-half revenue grew only +2.5% YoY. Shareholders' equity fell from $1.681 billion to $582 million in a year (−65%).
• A positive correction on dilution: over three years the share count rose from 593.4 million to 1.394 billion (+135%), but from February to August 2026 it rose only 0.23%, share-issuance proceeds in Q1 2026 were $0, and H1 financing activities were a net cash outflow of $67.92 million — there was no meaningful ATM issuance in the first half of 2026. Potential dilution: $375 million of converts issued in November 2025 and maturing in 2033, with a conversion price of about $3.00 and full conversion of about 125 million shares (+9%); the current price of $2.37 has not reached it, but it constitutes a technical valuation ceiling.
• Valuation: pre-market market cap $3.31 billion, TTM revenue $744.1 million → P/S 4.45x, P/B 5.69x. A company with a −0.9% gross margin burning $94 million of operating cash per quarter, at 4.45x sales.
• Pre-market and technicals: +12.32%, pre-market volume 6,516,165 shares (among the highest on the board). 52-week position 30.6%.
• Final judgment: Watch only. The improvement is real but roughly half of it is amplified by one-off items, while runway is only 2.7 quarters and the story requires a +40% second-half jump to deliver. This is not "it can't go up," it's "the payoff structure is not suited to entering after a +12% pre-market move." Verification points: whether second-half revenue really starts to jump in the Q3 report, and whether the $509.6 million of restricted cash can be released.

5.8 CRWV — CoreWeave (reports tonight, watch only)

• Related event: Q2 after 16:00 ET tonight. Consensus revenue about $2.55 billion (+111% YoY), adjusted loss per share about $1.22; the focus is the credibility and conversion pace of the $99.4 billion backlog. The options market implies a post-earnings single-day move of about ±15.5% (about ±$13.20).
• Fundamental verification (10-Q + official results materials, as of 2026-03-31, checked by a sub-agent) — in one sentence: interest is eating all of operating profit.

This is the most important comparison in this section: the company's own Q2 guidance is revenue $2.45–2.60 billion, adjusted operating income $30–90 million, and interest expense $650–730 million — interest is 7 to 24 times adjusted operating income. The quarterly interest-expense sequence is $264 million → $267 million → $311 million → $388 million → $536 million (Q1'25→Q1'26), more than doubling in a year.

• Debt structure (10-Q detail): total principal $25.149 billion, with effective rates in tranches as high as DDTL 1.0 at 15%, Magnetar at 12%, DDTL 2.0 at 11%, and two senior notes at 10% each; $6.066 billion of principal matures within 2026 against only $2.244 billion of cash + $777 million restricted. Q1 capex alone was $6.786 billion, with FY26 guidance of $31.0–35.0 billion (about 2.6x revenue guidance); TTM free cash flow −$10.62 billion; Altman Z-Score 0.36.
• ⚠️ The term structure of the backlog is the real point: of the $99.4 billion backlog (+284% YoY), only 25% is recognized within 24 months (about $24.9 billion, roughly $12.4 billion annualized), 36% in months 25–48, and 39% beyond 48 months. And FY26 revenue guidance is itself $12.0–13.0 billion — the near-end backlog barely covers current guidance, and the overwhelming majority of the $99.4 billion is realized more than two years out.
• Customer concentration: in Q1, customer A was 45% of revenue and customer B 20%, the top two together 65% (a year ago a single customer A was 72%). Deferred revenue fell from $8.2 billion at year-end to $7.5 billion at 3/31 (−$700 million QoQ), which is worth watching.
• Valuation: EV $80.99 billion / EV-Sales 13.0; P/S 7.73, forward P/S 3.08.
• Pre-market: +0.96%, pre-market volume 301,464 shares. 52-week position 30.8% (pre-market $89.04, range $60.55–$153.20) — already 42% off the high.
• Final judgment: Watch only, and it is the one of tonight's three with the largest information increment (both the numbers and the guidance are still unknown). A low 52-week position means expectations have been beaten down, which helps; but a ±15.5% implied move means taking a position before the report is mathematically a coin flip, and the leverage structure (debt of $25.1 billion, Z-Score 0.36) means directional risk may outweigh the IV crush — the shape of "even if volatility collapses you may not make money." Verification points (tonight): ① whether actual Q2 interest expense breaches the $730 million top of guidance; ② whether deferred revenue falls QoQ for a second consecutive quarter; ③ whether FY26 capex is revised up above $35 billion alongside another round of high-cost borrowing; ④ whether the within-24-months share of backlog can hold 25% — if backlog hits a new high but the near-end share falls below 20%, that means "bigger numbers, further out."

5.9 SMCI — Super Micro (reports tonight, watch only)

• ⚠️ The single most important correction in this brief: the three core numbers for SMCI tonight were already pre-announced by the company in an 8-K on July 21. A sub-agent went back to the SEC original (EX-99.1, 2026-07-21) and verified it word for word; the text reads:

  "Revenues for the fourth quarter of fiscal year 2026 are estimated to be near the low end of our guidance of $11.0 billion to $12.5 billion." "GAAP and non-GAAP gross margins are estimated to be in the range of 15% to 17% which is significantly higher than our guidance of 8.2% to 8.4%, primarily due to a favorable customer and product mix." "Backlog rose to record levels at the end of fiscal 2026 with total new orders in excess of $60 billion received during the fourth quarter."

  Three corrections to the circulating version: ① the gross-margin raise is real, but it is not "adjusted gross margin" — the company explicitly says both GAAP and non-GAAP are 15%–17%, a harder statement than the rumor; ② the $60 billion of new orders is real ($60B), but the forward-looking-statements paragraph of the same release carries its own disclaimer: "some of which may not constitute firm commitments and may be subject to cancellation or delays" — the company wrote that itself; ③ revenue landing at the low end of the $11.0–12.5 billion guidance is a negative item, which was overshadowed by the positive on gross margin.

• The key inference that follows: all of this was published on 7/21, and the stock rose 17% after hours that day and another 24% the next day. The only genuinely new information tonight is FY2027 guidance and the cash-flow statement.

• Fundamental verification (last four quarters' trajectory; the fiscal year runs July to June):

Gross margin jumping from 6.30% to 15–17% is attributed by the company to "a favorable customer and product mix" — but that 6.30% in Q2'26 was attributed to mix at the time as well. This is the point to listen to most skeptically on tonight's call.

• ⚠️ The genuinely overlooked landmine is on the balance sheet (Q3'26 vs FY25 year-end): inventory rose from $4.680 billion to $11.103 billion (+137%, inventory turns of about 108 days); receivables rose from $2.204 billion to $8.413 billion (+282%, DSO about 74 days); cash fell from $5.170 billion to $1.290 billion (a net decline of $3.88 billion over nine months) while the company remained accounting-profitable throughout — conversion of net income into cash is very poor; total debt rose from $4.758 billion to $8.773 billion (+84%).
• Convert dilution: actually not a problem. The 2028 notes convert at about $61.06 and the 2030 notes at about $55.20 (capped call cap $81.78), and at the current price of $31.46 all are deep out of the money with no dilution pressure right now; the problem runs the other way — refinancing at maturity.
• Valuation: P/E 16.49, forward P/E 9.25, P/S only 0.60, EV $28.20 billion; short interest 15.02%.
• Pre-market: +2.51%, pre-market volume 1,043,665 shares (the most liquid of tonight's three). 52-week position 32.5% (pre-market $32.25, range $19.48–$58.78).
• Final judgment: Watch only, and it is the purest IV-crush risk of tonight's three — the good news was released early on 7/21, the residual information content is the smallest, and the probability is relatively high that implied volatility collapses after the event while the stock lacks a fresh catalyst. Verification points (tonight): ① FY2027 full-year revenue/gross-margin guidance (the only unknown), and whether it matches the delivery pace implied by $60 billion of orders; ② whether Q4 operating cash flow and period-end inventory and receivables keep swelling; ③ whether management characterizes the 15–17% gross margin as a one-off customer mix; ④ when pressed on how much of the $60 billion of orders is firm commitment, whether they give a number. Note one asymmetry: revenue at the low end of guidance + 108 days of inventory suggests the "high margin" may come from cherry-picking orders rather than pricing power — if FY27 guidance does not match, the market will re-price it back on the low-valuation P/S 0.6 logic.

5.10 LITE — Lumentum (reports tonight, watch only)

• Related event: FY2026 Q4 is released tonight. It just fell 8.61% yesterday to close at $813.51, and is +0.48% pre-market this morning ($817.45, 111,176 shares). 52-week position 72.5% (pre-market $817.45, range $111.20–$1,085.68) — up more than 6x over 52 weeks, the highest-positioned and steepest-rising name on today's entire board.
• Fundamental verification (Q3 FY26, as of 2026-03-28, from the 10-Q original) — operating quality is the highest of the three: revenue $808.4 million (+90.1% YoY), GAAP gross margin 46.55% / non-GAAP gross margin 47.9% (35.2% a year ago), GAAP net income $142.5 million / EPS $1.50, non-GAAP operating margin 32.2% / diluted EPS $2.37. By segment: Components $533.3 million (+77.3%), Systems $275.1 million (+121%). The company says it is still "significantly under-shipping demand" (capacity-constrained, unable to ship). Four-quarter revenue/gross-margin trajectory: $480.7 million/37.28% → $533.8 million/37.65% → $665.5 million/39.02% → $808.4 million/46.55% — accelerating with gross margin expanding in step; the quality is real.
• Q4 guidance (given in May): revenue $960–1,010 million, operating margin 35–36%, non-GAAP diluted EPS $2.85–3.05. Consensus is about $989 million of revenue (+105% YoY) and about $2.99 of EPS, sitting at the guidance midpoint with no buffer.
• ⚠️ On the FCC's proposed ban on Chinese optical modules, the sub-agent pulled the decisive numbers from the 10-Q's own table:

But the 10-Q's own caveat must be read alongside it: revenue is attributed by country of initial product shipment, and "certain customers may request shipment of our product to a contract manufacturer in one country, which may differ from the location of their end customers." Thailand's 21.9% and Mexico's 12.8% are highly likely to be contract-manufacturing hubs (Thailand is where Fabrinet is), and do not represent end demand there; by the same token "China 8.7%" is a shipment-based figure and does not mean Chinese end demand is only 8.7%. China + Hong Kong at 25.5% combined is the reference upper bound. Also: shipments to Huawei have been fully halted since the beginning of 2024.

The conclusion that follows is entirely consistent with the theme downgrade in Section 7: LITE's demand-side exposure is limited (direct China 8.7%) and it is nominally a beneficiary of the ban; but the 10-Q explicitly flags the reverse risk — China's export controls on rare earths and critical minerals have "affected the availability and price…adversely affecting our operations, margins and sales," and indium phosphide (InP) is an essential material for high-speed lasers. The real China exposure is on the supply-chain input side, not the revenue side.

• ⚠️ The most underappreciated item: the fair value of the converts. The four series (2026/2028/2029/2032) total $3.183 billion on the books, while their fair value totals $19.075 billion — about 30% of the current $63.3 billion market cap, and that economic value belongs to convert holders, not common shareholders. The 2032 notes' capped call cap is $268.24, and at the current price of $813 that has been far exceeded and upside protection is entirely exhausted.
• NVIDIA's strategic stake: on 2026-03-02 it subscribed for about 2.9 million Class A convertible preferred shares at $695.31/share, $2.0 billion in cash in total — this is the main reason Q3 cash jumped from $657.7 million to $2.618 billion and shareholders' equity from $846.6 million to $2.973 billion. Note: LITE's cash growth comes mainly from that financing, not from operations.
• Valuation: P/E 154.65, forward P/E 50.94, P/S 25.43, EV/EBITDA 129.03, P/B 19.62; short interest 11.29% of float.
• Characterizing yesterday's −8.61%: COHR was −14.24% the same day, so it was collective de-risking ahead of earnings, not stock-specific bad news.
• Final judgment: Watch only — an extreme combination of bullish operations and bearish valuation. Revenue +90%, gross margin expanding 12pp, under-shipping under capacity constraints, an NVIDIA stake: operating quality is the highest of the three. But P/S 25.4 and EV/EBITDA 129 demand "consecutive quarters of beats plus continuous upward revisions," and that bar is extremely high. A 52-week position of 72% + an 8.6% drop just yesterday + earnings tonight + a theme with reverse supply-chain dependence: four risk factors stacked. Verification points (tonight): ① whether FY27 Q1 guidance can maintain "further sequential acceleration"; ② whether gross margin peaks on InP/rare-earth costs or product mix; ③ the number management gives when quantifying the benefit from the FCC ban — the market may already be pricing in "taking over Chinese vendors' share," and anything below imagination is a negative; ④ the fully diluted share count disclosed for the first time. There is also a convention trap: GAAP EPS of $1.50 versus non-GAAP of $2.37 differs by 58%, and guidance is given on a non-GAAP basis — if the media compare GAAP against the non-GAAP consensus, a fake "big miss" will appear.

---

6. Bearish / Avoid List

6.1 A Deep Check on the Four Decliners: the reasons they are falling are not the same class of thing at all

After a sub-agent went back to the SEC originals (8-K Ex-99.1, 10-Q, 6-K, 424B5) and verified HIMS / ONON / UPWK / INTC item by item, the conclusion is that treating "guidance cut" as one homogeneous event would produce three wrong conclusions across these four companies:

Point by point:

① HIMS — what fell is "the quality of growth," not "growth." Revenue +38%, subscribers +19%, monthly revenue per subscriber +21%, US revenue growth recovering from 3% on an H1 basis to 16% in Q2, and marketing spend of $262 million up only 20% (slower than revenue growth — this is not burning cash to buy customers). What it traded away is gross margin: branded GLP-1 booked on a gross basis + acquired overseas businesses (overseas revenue exploding from $7.5 million to $131.4 million) knocked 12pp off gross margin. That is a completely different animal from "subsidized customer acquisition" — the latter is reversible, the former is a re-pricing of the business model. ⚠️ The revenue-quality clue that most needs watching: net receivables exploded from $32.1 million at the end of 2025 to $375.3 million, of which $347.4 million is "vendor discounts and rebates receivable" (explicitly disclosed in the 10-Q, arising from supply agreements, with the year-end 2025 balance described as "not material"); correspondingly accounts payable rose from $143.3 million to $501.3 million. And 10-Q Note 20 discloses that on July 1, 2026, two wholly owned subsidiaries entered into a receivables sale agreement with JPMorgan for up to $400 million. This means Q3 operating cash flow will be materially flattered by a financing action, and when reading the Q3 statements the size of those sales must be stripped out first. Within half a year the company has gone from net cash of $154 million to net debt of about $1.39 billion.

② ONON — the "proactive volume control" claim has one counter-argument coming from the inventory statement. The evidence supporting management's narrative is strong: DTC +34.3% in constant currency, its share rising to a quarterly record 45.7%, apparel cc +56.2%, Asia-Pacific cc +54.7%, a record gross margin of 65.4% (+3.9pp) that already fully absorbs US import tariffs and includes no tariff refunds, with full-year gross-margin guidance actually raised to ≥65.0% and adjusted EBITDA margin guidance maintained at 19.5–20.0%. A company with broadly weakening demand can rarely lift gross margin by 3.9pp while holding pricing discipline — that is the main reason this brief does not accept the "demand collapse" explanation, but it is a probabilistic argument, not an exclusive proof. But the counter-evidence is also hard: inventory rose from CHF 360.4 million at 2025-06-30 to CHF 472.9 million (+31.2%), while reported net sales over the same period grew only 13.5%; implied days inventory outstanding rose from about 114 days to about 146 days (+32 days). Proactive volume control and passive inventory build look identical on the income statement; only the inventory statement can tell them apart, and this time the inventory statement leans toward the latter. ⚠️ Channel inventory (stock in wholesalers' hands) is not disclosed by the company and was not obtained by this brief — which is precisely the link the "proactive volume control" narrative most needs verified and which is least visible. Another convention reminder: what was cut is constant-currency growth ("at least +23%" → "low-20%"), while the absolute CHF guidance barely moved (new CHF 3.47–3.56 billion, midpoint 3.515 billion ≈ the old low end of 3.51 billion). The new guidance implies a full-year FX drag of about 4–5pp, less than Q2's actual 8.1pp — i.e. the company assumes the currency headwind narrows in the second half. That is an assumption, not a fact.

③ UPWK — this is the only one of the four where the company itself revised future numbers down, and what it revised is the slope.

Active clients 763,000 (−4.1% YoY, down another 2.7% QoQ). GSV and revenue are heading down at the same time, and the revenue decline is pushed by the company's own guidance into quarter-by-quarter acceleration (−1.7% → −10.8% → −13.0%). ⚠️ Convention reminder: GSV and revenue are two different series and have been split into separate columns above — do not stitch them across columns into a single trend line; Q1'26 revenue YoY was not obtained by this brief, so "quarter-by-quarter acceleration" currently rests on three points (Q2 actual + Q3/Q4 guidance), not four. Even so, this remains the only one of today's four decliners where the company itself revised the slope of its future numbers down. But the same dataset also shows this is "structural substitution" rather than "volume disappearing": take rate rose 39bp against the trend to 19.83%, GSV per active client grew QoQ for the 8th consecutive quarter to $5,230 (+5%), AI-related work GSV +22%, the AI strategy-and-consulting subcategory +50%+, and Business Plus (SMB) GSV +24% QoQ. Low-complexity work is being eaten by automation (clients leave), and the remaining clients spend more per head — the platform is shrinking, but concentrating. ⚠️ One convention that has to be called out: adjusted EBITDA margin of 33% (+415bp) is a record, but free cash flow is −45% YoY and −49% in the first half. The EBITDA was manufactured by "shrinking to improve efficiency" (opex −1%), while FCF reflects the real cost of severance (a 24% workforce cut announced in Q1), taxes and working capital. Looking only at EBITDA overstates this company's cash generation.

④ INTC — this time it is not a dilution negative but "financing friction"; the reason it fell only 1.2% is in the operating data.

• Dilution math: on the 6/27 base of 5.043 billion shares, post-offering it is 5.2535 billion (+4.17%), or 5.2851 billion with the greenshoe fully exercised (+4.80%).
• Book value per share actually increases: equity attributable to the parent of $87.542 billion ÷ 5.043 billion = $17.36 → adding the $19.7 billion of net proceeds gives $20.41 (+17.6%). Issuing at 5.47x book value necessarily accretes book value rather than diluting it.
• Net debt is close to zero: cash $12.874 billion + short-term investments $16.853 billion = $29.727 billion, total debt $50.54 billion → net debt $20.81 billion → about $1.11 billion after the offering. The offering documents state a commitment to "maintaining a strong balance sheet and investment-grade ratings" — the rating is a hard constraint, and using equity rather than debt was forced by the rating while also being the financially superior choice.
• What "general corporate purposes" actually refers to: on 2026-04-08 the company bought back Apollo's 49% minority stake in the Irish SCIP (Fab 34) for $14.2 billion in cash; Q2 adjusted free cash flow was −$8.419 billion; in H1 it issued $13.0 billion of new debt and repaid $9.0 billion of old debt; FY26 gross capex guidance was raised to >$20 billion, with 2027 "significantly higher." This is not opportunistic financing; it is filling a hole punched jointly by "buying back the JV stake + an upswing in the capex cycle."
• ⚠️ That $11.0 billion GAAP net loss in Q2 must be broken apart: operating income was a positive $1.796 billion and non-GAAP diluted EPS was a positive $0.42; −$12.529 billion came entirely from fair-value movement on the "Escrowed Shares" — stock delivered in tranches to the US government according to milestone progress under the Secure Enclave agreement signed with the Department of Commerce on 2025-08-22, accounted for as a derivative liability and remeasured at fair value. Intel's share price rises → this liability grows → the GAAP loss grows. Causally, this "huge loss" is an accounting consequence of the share price rising, not of deteriorating operations. In the same quarter revenue was $16.128 billion, +25% (the fastest in fifteen years), DCAI +59%, the Foundry loss rate narrowed from −71.7% to −36.2%, and Q3 guidance has GAAP EPS turning positive at $0.31. That is most likely why it fell only 1.2% — though it should be said that inferring what the market thinks from the price reaction is ex-post attribution, not independent evidence.
• ⚠️ But there are three further layers of contingent dilution beyond the $20 billion, and those are what really deserve monitoring: ① 143 million undelivered escrowed shares (71 million of which are already in basic EPS, 71 million contingent), whose derivative-liability fair value has risen from $2.65 billion at the end of 2025 to $15.57 billion; ② up to 241 million warrants issued to the Department of Commerce with a strike of only $20.00, with intrinsic value of about $18.3 billion at a $96 share price, and exercisable only if Intel directly or indirectly holds less than 51% of the foundry business — any plan to spin off the foundry or give up control has to pay this bill first; ③ 70 million equity-incentive shares excluded as anti-dilutive because of the loss.

---

7. Within-Theme Rankings

7.1 AI Power and Data-Center Capacity (today's #1 theme)

This table is the most important one in this brief. Yesterday's pre-market brief listed the four optical-module names (COHR/LITE/CIEN/AAOI) indiscriminately as beneficiaries of the same theme, and all four were wiped out. Today's grading principle is: does it have a contract/disclosure of its own. RIOT and BW do; the other seven do not.

7.2 Tonight's Three Reports (ordered as events, not by strength)

This table is reordered relative to Section 3.2, on the basis of "information increment" rather than "importance." SMCI drops from an intuitive 2nd to 3rd for one reason only: it can deliver the fewest new facts tonight — yet its IV is priced as if for a "full earnings report," which is the standard recipe for an IV crush.

7.3 Energy / Hormuz (yesterday's theme, digesting today)

The only difference between energy today and yesterday: yesterday there was incremental news (the Houthi attack on the Jizan refinery), today there is none. Oil is flat and the names are all slightly red pre-market — this is "positions hesitating," not "new money arriving."

---

8. Opening Verification Signals

8.1 Pre-Market (09:00–09:30 ET)

• The gap direction and volume of the theme leaders: watch closely whether pre-market volume in RIOT ($22.09, +13.87%) and BW ($12.18, +37.16%) keeps expanding. RIOT's pre-market volume has already reached 1.85 million shares and is credible; BW's 610,000 shares is also enough for a $1.3 billion market cap. If both fade on shrinking volume between 09:00 and 09:30, this is overnight sentiment rather than institutional accumulation.
• Gap-and-go or gap-fill: BW is today's most typical gap-risk name (+37%). The test: whether it can hold $11.50 after the open (about 2/3 of the pre-market gain). RIOT's corresponding level is $21.00 (about 3/4 of its gain).
• Never look at the pre-market price on thin-volume names: FSLR (22,484 shares), CAH (12,233 shares), LIF (15,328 shares), NIQ (8,153 shares), ESLT (2,803 shares), OIH (379 shares), TAN (201 shares), XLY (9 shares), XLC (11 shares) — these names' pre-market prices can re-price sharply the instant the market opens, so the pre-market reading cannot be a basis for decisions.

8.2 Intraday (09:30–10:00 ET)

• The first-30-minutes trend: today's key contrast is RIOT/BW (AI power, with contracts) versus MARA/CIFR/WULF/APLD/FRMI (same theme, no contracts). If the latter group rallies along and does not fade, it is a theme move with poor persistence; if the latter group fades quickly while RIOT/BW hold up independently, the market is discriminating "contract vs concept" and the theme is of higher quality. This is today's most valuable observation.
• Sector ETF linkage: only after the open do we get a valid sector signal for the first time (pre-market ETF volume is unusable). Watch whether SMH can extend its +1.02% pre-market bounce to recover part of yesterday's −2.28%, and whether XLE can hold yesterday's +4.66% now that oil has stalled.
• Volume confirmation: PLUG's 6.51 million shares pre-market is among the highest on the board, but it is a $2.37 stock — a large share count is not a large dollar amount. Confirm absorption in dollars, not shares.

8.3 ⚠️ One Discipline That Must Be Carried Over: the pre-market price is not intraday strength

Yesterday's (8/10) full-sample quantitative conclusion: of the 25 reconciliation names, 21 closed below that morning's pre-market price and only 4 above (ABCL +7.98%, OXY +3.79%, XOM +3.75%, HZO +0.08%). On that day the pre-market price was very nearly the high of the whole session.

⚠️ But the sample size must be stated: this is a cross-section of 1 trading day and 25 names, not a multi-day statistic. It is enough to support "this needs re-verifying today," not enough to support "this happens every day." Tonight's recap should compute the same two columns again and see whether the ratio repeats.

The direct implications for today:

• Section 3's rankings are deliberately not ordered by pre-market gain. BW is +37.16% pre-market, the largest on the board, but ranks 4th; SE is up only 6.82% pre-market and ranks 1st. The reason: a pre-market gain measures "how many people wanted to buy overnight," not "how strong this name is today."
• For every name, the correct reconciliation baseline is "relative to the pre-market price," not "relative to yesterday's close." Tonight's recap should compute both columns: the change relative to the 8/10 close, and the change relative to this morning's pre-market price. Using only the former systematically overstates the hit rate.

8.4 Options Sentiment

• ⚠️ This brief did not obtain real-time IV or call/put skew data. The only options information available is media-reported implied move magnitudes: CRWV's post-earnings single-day implied move is ±15.5% (about ±$13.20); SE implied about ±19% before its report, about 1/3 of which has been realized this morning.
• Post-earnings IV crush risk: options on tonight's CRWV/SMCI/LITE will all experience an IV collapse by tomorrow morning. Buying options on these three before the reports can lose money even if the direction is right.

8.5 Risk Signals (if they appear, lower position expectations)

1. Reversal after a gap: if BW, RIOT, ACVA or FRMI break below half of their pre-market gain within the first 30 minutes, treat the pre-market pricing as broken.
2. A lone name rising with no sector follow-through: if RIOT rallies hard while CIFR/WULF/APLD all turn negative, this is a single-stock event rather than a theme — which is not necessarily bad, but it means you cannot use the "theme" as a reason to buy the other names.
3. Divergence between index futures and the 10-year: right now futures are slightly up and the 10-year slightly down, which is consistent. If an intraday combination of "futures up, 10-year rising fast" appears, it usually means CPI expectations are being re-priced intraday, which is negative for high-multiple growth.
4. Waiting on the sidelines ahead of CPI: tomorrow morning's 8:30 CPI is the only tier-one event this week, so the market will most likely go quiet on shrinking volume after 14:00 this afternoon. VIX at only 15.47 means the market has bought almost no insurance for this event — the biggest asymmetric risk right now.
5. Tonight's three 16:00 reports: any one of them blowing up would hit three lines at once tomorrow morning — AI compute (NVDA/AMD/AVGO), AI servers (SMCI's peers) and optical modules (COHR/AAOI). Positions this afternoon need to leave room for that event.

---

9. Final Conclusions

9.1 The 5 Stocks Most Worth Watching Today

Note on ranking convention: this table is ordered by today's attention priority, which differs from Section 3.1's ordering by total score. There is only one difference — VG ranks 2nd by total score in 3.1 but 4th here — because its score comes almost entirely from earnings quality, while it already rallied 7.54% yesterday and is down 4.49% this morning, making the timing dimension terrible. The total score measures "how good this catalyst is"; this table measures "what you can do with it today."

9.2 Today's 3 Strongest Themes

9.3 What to Avoid Today, and Why

1. Within this brief's coverage, names that are "on the same theme but have no company news of their own" and whose liquidity or position also fails to support them — MARA, FRMI, QMCO, NIQ, HUT and CLSK are marked "avoid"; CIFR, WULF and APLD likewise have no company news but do have AI hosting businesses in operation, so they are marked one notch lower as "watch only" rather than "avoid." Yesterday's wipeout of all four optical-module names is the price of this mistake: the market first pulls them up together on the theme, then knocks them back down one by one on fundamentals. Today's grading criterion is a single question — does it have a contract or disclosure of its own.
2. Treating "FCC ban = optical-module beneficiary" as a reason to buy (COHR / LITE / AAOI / CIEN) — what is being avoided is this thematic logic, not the names themselves (LITE is still marked "watch only" rather than "avoid" in Sections 3.2/5.10, because its operating quality really is the highest of tonight's three). That "beneficiary" logic contains a reverse supply-chain dependence: the designated substitutes themselves rely on indium phosphide exported from China, and COHR's CEO has already traveled to Beijing over InP license delays. A structure of "benefiting on the concept, constrained on the supply chain" is not a tradable positive.
3. Any name with extremely thin pre-market volume (thousands to tens of thousands of shares) — ESLT (2,803 shares), NIQ (8,153 shares), CAH (12,233 shares), LIF (15,328 shares), FSLR (22,484 shares); plus those already down to three digits or fewer with no price meaning at all: OIH (379 shares), TAN (201 shares), XLC (11 shares), XLY (9 shares). These prices are not prices, they are the last resting order. ⚠️ Note that FSLR also appears on the watch list in 9.1 — there is no contradiction: what should be avoided is this morning's pre-market price propped up by 22,484 shares, not the name itself.
4. Buying CRWV / SMCI / LITE before the reports — the implied single-day move is ±15.5% (CRWV), so taking a position before the report is mathematically a coin flip, and tomorrow morning adds an IV crush on top. SMCI is especially poor value: revenue, gross margin and the $60 billion of orders were all published on 7/21 and already traded, leaving only FY27 guidance and the cash-flow statement tonight, yet you pay implied volatility as if for a "full earnings report."
5. Treating today's four decliners as the same class of event — HIMS revised margins rather than growth (Q2 in fact double-beat its own guidance), ONON revised the measure rather than the amount (the absolute CHF barely moved), INTC revised its capital structure rather than its earnings (book value per share actually +17.6%), and only UPWK revised the slope of its future numbers down. Using the single label "guidance cut" to understand them would produce three wrong conclusions out of four.
6. Shorting bearish names that have already fallen to the bottom of their 52-week range — UPWK (52-week position 3.9%), ONON (3.7%), GETY (0.3%), OPFI (0.0%). The logic being right does not make the position right; at the very bottom of the range the payoff on a short is already poor.
7. New positions opened after 14:00 this afternoon — tomorrow morning's CPI plus tonight's three reports are two stacked events.

9.4 The Final One-Sentence Judgment

Today's information is concentrated on the "powering AI" line, and unusually there are two unrelated corporate disclosures corroborating each other (RIOT's 191MW lease, BW's 62% revenue beat and guidance raise) — this is today's only theme with both logical hardness and verifiable terms; but its cash flow does not begin until the end of 2027, and this morning's gap has already booked, in one move, the entire net value added of that lease computed at an 8% discount rate (at a 10% discount rate the net value added is only about $500 million, i.e. it has already discounted it twice over; this conclusion is highly sensitive to the discount-rate assumption, and the rigorous way to put it is "no cushion is left," not "it is definitely expensive"). What actually decides today's close and tomorrow's open is not this theme but tonight's three 16:00 reports and tomorrow's 8:30 CPI — and with VIX at only 15.47, the market has paid almost no insurance premium for either. Under that combination, the thing most worth doing today is not stock picking but controlling exposure ahead of the events.

---

Internal Operations Log (not sent to clients)

Data-retrieval failures and workarounds:

• The CNBC quote API could not retrieve index futures: @ES.1 / @NQ.1 / @YM.1 / @RTY.1 / ES=F / .SPXFUT all returned empty rows (API code:1). Futures data was switched to the Yahoo Finance quote page (snapshot around 08:00 ET), a different source from the rest of the quotes and about 20 minutes apart, which is stated in the header conventions. The spot indices (.SPX/.DJI/.IXIC/.RUT) worked normally but returned only the 8/10 closes, as expected.
• WebFetch returned HTTP 403 for CNBC and TheStreet, and HTTP 503 for the Yahoo Finance analyst-ratings article; switched to curl + a custom UA to fetch the HTML and parse it locally — CNBC succeeded, while the Yahoo article's body still could not be obtained due to front-end rendering. Today's sell-side ratings in this brief are therefore limited to the single Baird/FSLR item explicitly noted in the CNBC pre-market roundup.
• Old news mixed into search results, intercepted once: while searching for "today's analyst ratings," the results included several downgrades related to the Roku/Fox acquisition; verification confirmed that deal was announced on 2026-06-15, not today; the quotes corroborated it too — ROKU was +0.03% and FOXA +0.60% pre-market, showing it was not today's catalyst. Removed, not written into the report. This is exactly the same class of trap as ai-attribution-stitches-old-news in memory.
• stockanalysis's screener JSON API returned 404 (both /api/screener/s/f and /api/screener/a/f were tried); switched to scraping and parsing the HTML tables at /markets/premarket/gainers/ and /losers/, which produced a complete gainers/losers list. But the quotes API /api/quotes/s/<TICKER> works perfectly, with the pre-market ep/ecp/epv/eu fields usable, and all single-stock pre-market data in this brief goes through that path.
• MU returned an ERR once in the final batch retrieval (a single-name timeout) while the other 29 names in that batch were fine; MU's reading carries over the +0.99% from the prior batch at 08:06 ET, and is used at that timestamp in the text.
• One fundamentals-analyst sub-agent (CRWV/SMCI/LITE) failed on its first run because the API connection dropped (error: Connection closed mid-response). It was re-run with a narrowed scope (400–600 words per company, skipping options data that would certainly be rate-limited), and the second attempt returned successfully and with high quality. Lesson: when this kind of sub-agent fails, re-running with a narrower scope works better than retrying as-is. All three fundamentals-analysts returned in the end.
• A sub-agent reported being rate-limited by Yahoo/yfinance mid-way and switched on its own to the stockanalysis quote API + SEC EDGAR originals, without affecting the conclusions. This matches the judgment in memory's yfinance-rate-limit-root-cause: the rate limiting is intermittent, and stockanalysis is a reliable fallback channel.
• VIX9D was not updated this morning (still the 8/10 close of 12.77); the CNBC API returns only a row dated 2026-08-10 for that symbol. Yesterday's recap used it to judge "buying short-dated insurance for CPI"; that thread is broken today, and it is stated in the header.

Sub-agent results and coverage gaps (these affect the strength of the judgments and should be noted at recap time):

• SE ranks #1 overall but no fundamentals sub-agent was run on it — the three agent slots went to RIOT/PLUG/FSLR and the three reporting tonight. SE has only press-release-level revenue and EPS, with earnings quality unverified.
• BW ranked 4th in the first draft with a full 10 on expectation gap, and QC found that "H1 new bookings $2.7B / backlog +533%" was overwhelmingly Q1 content disclosed on 5/11 (Q2 standalone bookings only $151.0M), i.e. an old disclosure was being credited to today. Expectation gap was cut to 7, earnings elasticity to 12, total 64→59, rank 4→5, with the date check added in four places in the text. This is the most valuable correction of this round: in the same report I got "already pre-announced on 7/21" right for SMCI, yet applied a different standard to BW. The balance sheet remains entirely unverified.
• The moves in ACVA (+17.08%, 857,000 shares) and FRMI (+15.48%, 1.278 million shares) clearly do not match the strength of any news this brief could find. ACVA's EPS and full-year guidance are both merely "in line with consensus," and FRMI has no news today at all. Both were handled as "unexplained = avoid," but if they trade well today, this brief missed a story and it must be re-checked at recap. Four judgments the sub-agents overturned that I was originally going to get wrong (this is the most valuable part of this run):

1. SMCI's "gross margin 8.2–8.4% → 15–17%" I had originally flagged as "to be verified, possibly a change of convention." The sub-agent went back to the SEC EX-99.1 (2026-07-21) and verified it word for word: it is true, and harder than the rumor — the company said both GAAP and non-GAAP are 15–17%. More importantly, the very existence of that pre-announcement: revenue, gross margin and the $60 billion of orders were all published back on 7/21, and the stock rose 17% after hours that day and 24% the next. I had originally written SMCI up as "a full earnings report tonight," and that was wrong — tonight leaves only FY27 guidance and the cash-flow statement.
2. INTC I had originally written as "bearish (dilution), A+ tier." The sub-agent calculated: dilution is only 4.17%, while the offering price of $95 = 5.47x book value, so book value per share actually rises +17.6% and net debt goes from $20.8 billion to $1.1 billion. And that $11.0 billion GAAP loss in Q2 is the escrowed-share fair-value remeasurement (stock up → liability up), with operating income a positive $1.796 billion. Re-rated to "neutral-to-bearish / financing friction." Had the original draft gone out, that would have been a directional error.
3. HIMS I had originally written the "$62.5 million FTC litigation accrual" as a Q2 standalone figure — it is in fact an H1 cumulative figure, and Q2 standalone was $47.5 million. Corrected. More importantly, the sub-agent pointed out that Q2 was in fact a double beat against the company's own guidance, plus the $347.4 million of "vendor discounts and rebates receivable" and the $400 million receivables sale agreement signed with JPMorgan on 7/1 — the latter will flatter Q3 operating cash flow with a financing action, an angle I did not have at all.
4. ONON I had originally written only the two-sidedness of "constant-currency growth cut vs CHF absolute raised." The sub-agent added the decisive counter-evidence: inventory +31.2% YoY while revenue grew only 13.5%, with DIO rising from about 114 days to about 146 days. "Proactive volume control" and "passive inventory build" look identical on the income statement, and only the inventory statement can arbitrate — this directly changed how much credence to give management's narrative.

Gaps still open (by priority, to be carried into tonight's recap and tomorrow's pre-market):

1. The balance sheets of SE and BW — ranked 1st and 4th overall, with only press-release-level numbers for both. BW especially matters: the ratio of "backlog $2.6 billion / market cap $1.3 billion" itself says the market did not believe the orders would convert, and this company has a history of high leverage and multiple restructurings.
2. The moves in ACVA and FRMI cannot be explained — if they trade well today, this brief missed a story and it must be re-checked.
3. The counterparty and payment terms of HIMS's $347.4 million of "vendor discounts and rebates receivable", and the actual drawn size in Q3 of the $400 million JPMorgan receivables sale agreement signed on 7/1. This is the accounting counterparty to the 76%→64% gross margin.
4. ONON's channel inventory (sell-in vs sell-through) — not disclosed by the company. It is the only key that arbitrates "proactive volume control vs passive inventory build."
5. The composition of UPWK's active-client attrition (whether those leaving are purely low-complexity small clients). The implied Q4 growth range of −5.9% to −20.0% is 14pp wide, and that width itself says the company has no conviction about the inflection.

risk-auditor QC round log (one round was run before finalizing; the results are worth recording):

• QC caught 1 order-of-magnitude unit error: the first draft wrote RIOT's "241MW corresponding to roughly $9.8 billion (≈$9.8B) of total long-term contract revenue" as "$980 million of annualized NOI" in both the overall ranking and the Top 5 — wrong by an order of magnitude and also swapping "20-year cumulative contract revenue" for "annualized NOI." This is exactly the same type of error as english-summary-billion-yi-tenfold-error in memory, and it landed in the two most conspicuous places. Fixed.
• QC caught 1 directional characterization error: the first draft labeled INTC's $20 billion offering in the theme table as "pure dilution event / bearish / B+," while the body text (after adopting the sub-agent's conclusion) said "book value per share +17.6%" — two opposite characterizations of the same name in the same report. Unified to "capital-structure event / neutral-to-bearish / B."
• QC caught 1 methodological double standard (the most valuable one): I correctly identified for SMCI that "it was pre-announced on 7/21 → only FY27 guidance is left tonight," yet for BW I treated Q1 orders disclosed on 5/11 as "the largest expectation gap on the board" and gave it full marks. The same criterion applied two different ways in the same report. The score was cut and the date check added in four places.
• QC also produced a false positive of its own, which was verified and rejected: it claimed this brief had missed HPE (+5%, upgraded by Morgan Stanley) and VRSK (−6.5%, a Delaware judge ruling that it must complete the AccuLynx acquisition) from the CNBC pre-market roundup. I re-fetched the body of that URL, and neither is in the 8/11 piece; the quotes falsify it too: HPE closed +2.74% on 8/10 but is only +0.13% pre-market this morning; VRSK closed −5.55% on 8/10 but is +0.40% pre-market this morning (on only 546 shares), and the AccuLynx ruling was on 8/7 (Friday). Both are yesterday's tape, not today's catalyst. Interestingly, this is exactly the same class of error as the Roku/Fox item (6/15 old news) that I intercepted earlier — which shows the "stitching in old announcements" trap works on QC agents too, and being QC is no reason to accept its findings by default.
• One QC suggestion that was not adopted this time: it proposed renaming the "short-watch candidate?" column in Section 6. Not changed, because that is a fixed output structure mandated by SKILL.md and column names cannot be altered at will. The wording in that column was tightened instead to reduce its directiveness.
• Other adopted items: 52-week positions were uniformly recomputed off pre-market prices (ACVA 39.9%→45.1%, MARA 15.2%→19.4%, QMCO 51.4%→65.8%, ONON 5.2%→3.7%, UPWK 3.2%→3.9%, etc.); the UPWK table split GSV and revenue into two columns (the first draft had stitched two series into one trend line); TSMC's cumulative +37.0% got the note "in New Taiwan dollars, no currency bridge applied, not directly comparable with USD guidance"; FSLR's "less than half" was changed to "roughly half" and flagged as an estimate; RIOT's discount-rate sensitivity was written into the conclusion; universal quantifiers were narrowed in several places ("all," "every," "any one"); and the header's account of the futures-retrieval failure was changed back to a pure statement of conventions.

Methodology self-check (against yesterday's recap lessons):

• Yesterday's biggest failures were "putting a to-be-verified direction in the #1 slot" and "using the pre-market gain as intraday strength." Today's responses: ① Section 3's rankings are ordered by catalyst quality rather than pre-market gain, with BW at +37% pre-market ranked 4th and SE at +6.8% ranked 1st, and the component-level reasoning for that trade-off written out in Section 4.1; ② Section 7.1 has a dedicated "with contract vs without contract" grading table, explicitly marking MARA/FRMI as pure concept; ③ Section 8.3 writes the quantitative conclusion "21 of 25 fell below their pre-market price" directly into the verification discipline.
• The US-market analogue of memory's single-day-fund-flow-mean-reverts (a single day's flow extreme is a contrarian indicator) is "the pre-market gain extreme." Readings like BW +37% and FRMI +15% are treated throughout this brief as risk deductions, not strength credits.

---

⚠️ Risk disclaimer: 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 have information lag or factual errors, so company disclosures / SEC filings prevail and this must not be used directly as a basis for trading.