U.S. Market Recap | 2026-08-11 (ET) Tuesday

Coverage window: 2026-08-11 09:30–16:00 ET regular session, plus 16:00–18:05 ET after hours. Quote conventions:

• Single-stock and ETF previous close / open-high-low-close / volume / after-hours price come from the stockanalysis.com quote API, snapshotted 18:00–18:06 ET; the closing price is the 8/11 16:00 ET regular-session close, and the timestamp of each after-hours price is noted in the tables (mostly 17:4x–18:0x ET).
• Index levels, VIX, VIX9D, Treasury yields, the dollar index and bitcoin come from the CNBC quote API, snapshotted 17:05–18:02 ET; index levels are the 16:00 ET closing marks.
• The "vs pre-market price" column = closing price ÷ this morning's pre-market price − 1, where the pre-market price is taken from the pre-market list being reconciled here (08:09 ET snapshot) or reconstructed as "previous close × (1 + pre-market change)". This column measures the real result of "buying before the open following the pre-market list".
• Earnings figures come from company press releases and reprints by established media, with sources noted in the tables. Data not obtained for this report: ① settlement prices for WTI / Brent / gold were not obtained (see the note in Section 1); the direction of crude is instead expressed via the closing changes of the USO / BNO ETFs, and absolute price levels are quoted only on a media basis and flagged as such; ② the Q3 single-quarter guidance and the FY26 capex ceiling given on the CRWV call were not obtained here, only the full-year revenue / adjusted operating income ranges from the press release and media reprints; ③ advance-decline breadth was obtained only as a third-party percentage measure (40.4% advancing / 56.3% declining), and the raw NYSE and Nasdaq advance-decline counts were not obtained; ④ option implied volatility and the actual magnitude of the IV crush were not obtained; the judgment about IV crush in Section 4 is inferred from the price evidence of "a large earnings beat that did not rise after hours", not from options data.

---

0. One-Line Recap

Today was a shallow risk-off day of "index slightly down, turnover underneath": S&P −0.32%, Nasdaq −0.60%, while the Russell 2000 rose against the tape at +0.32% — what fell was large-cap tech (GOOGL −3.84%, AMZN −2.09%), and what rose was the "power the AI build-out" line (VRT +4.34%, TLN +4.28%, ETN +3.22%, CEG +2.93%, XLU +1.16%) and energy (XLE +1.25%, the strongest SPDR sector of the day). The pre-market list called the strongest theme correctly and picked the right name for the top slot (SE +14.56%), but put the three best performers of the day within that same theme (FRMI +21.09%, QMCO +64.69%, NIQ +41.95%) all into "avoid" — the reason given in each case was "no news found today", and all three in fact had hard company disclosures that day. This is the single most important line in this report: what was wrong was not the criteria, it was the data gathering. After the close, CRWV beat and rose +13.47%, handing tomorrow's tone to AI compute; but the real first-tier event is the 8:30 CPI tomorrow morning, and VIX closed at 15.28 (−1.16%) with VIX9D at 12.52 (−1.96%) — the day before the event, the market once again made insurance a little cheaper.

---

1. Market Overview

Cross-check on an ETF basis: SPY −0.32% (volume 36.73 million shares), QQQ −0.34% (29.15 million shares), DIA −0.32% (2.82 million shares), IWM +0.34% (15.03 million shares). All four ETFs match the indices in both direction and magnitude.

Market breadth: 40.4% advancing / 56.3% declining, the weakest single-day advancing share since 8/7 (chartmill statistics¹). ⚠️ The raw NYSE/Nasdaq advance-decline counts were not obtained for this report; only this one third-party percentage measure is available, so read it as a reference value.

Rates, volatility and FX (16:00–17:05 ET)

Commodities (ETF basis, at the close): USO +1.34%, BNO +1.70%, UNG −0.69%, GLD −0.39%, SLV −1.45%. ⚠️ Futures settlement prices for crude and gold were not obtained for this report (see internal notes). On a media basis: WTI rose for a fourth consecutive session and traded above $83 intraday; Brent was around $89 (Yahoo Finance intraday live blog²). The direction is confirmable (up); read the absolute prices on a media basis.

Sentiment call: shallow risk-off, but not a flight to safety. Three pieces of evidence contradict each other and hold simultaneously: ① indices fell and breadth was worse (40% advancing) — that is risk-off; ② the Russell 2000 closed higher against the tape, XLE/XLU/XLI led, and ITB (homebuilders) +2.14% — money moved outside the index, not into cash; ③ both 10Y and 30Y edged higher and TLT rose only 0.16%; the bond market did not absorb the money leaving equities. The correct characterization is intra-market rotation of "sell large-cap tech, buy real assets and power", layered on top of a general reduction in exposure ahead of CPI.

⚠️ One asymmetry that must be spelled out, and it runs opposite to yesterday's pre-market judgment: yesterday's pre-market report wrote that "VIX is only 15.47; the market has barely bought any insurance for tomorrow's CPI". Today VIX fell further to 15.28 and VIX9D to 12.52 — with only one night to the event, short-dated protection actually got cheaper. This is not "the market thinks CPI does not matter", it is "the volatility sellers are still collecting premium on the last day before the event". This under-hedging asymmetry has not been repaired; it has deepened.

---

2. Pre-Market List Reconciliation

Convention for the two columns: "today's change %" = versus the 8/10 close; "vs pre-market price %" = versus this morning's 08:09 ET pre-market price, i.e. the real result of executing the pre-market list before the open. The lesson from yesterday's recap was that using only the former systematically overstates the hit rate, so both columns are shown side by side.

2.1 Long side (priority deep-dive + watch closely + watch only)

2.2 Avoid / bearish side

2.3 ⚠️ The most important self-criticism in this report: the three "avoid" names were the top three gainers of the day, and all three had news

The pre-market list put FRMI, QMCO and NIQ all into "avoid", with identical stated reasons — "no findable news", "cannot define a falsification condition". That day the three rose 21.09%, 64.69% and 41.95% respectively, ranking No. 3, No. 1 and No. 2 across the entire reconciliation table. Checking each one after the fact, all three had substantive company disclosures either that day or the previous night:

These three errors are not errors of judgment, they are errors of data gathering, and each is of a different nature:

1. FRMI is the most serious of the three, because it hit three things at once. ① Its news was released in the exact middle of the coverage window the pre-market list declared for itself (8/10 after hours), not old news from outside the window; ② its news was structurally identical to RIOT, which was rated the No. 1 theme that day — a binding long-term data center lease, an explicit megawatt figure, an explicit total contract value, an explicit delivery date, right down to the feature of "the lessee is unnamed but points to an AI giant"; ③ the pre-market list devoted an entire section (7.1) to building a tiering table of "has a contract vs has no contract", and placed FRMI squarely in "pure concept, catalyst directness 0/20". The criteria were right, and the market validated them that day — FRMI rose 21% precisely because it had a contract. What was wrong were the facts used when applying the criteria. Using a correct framework on an incorrect list of facts is worse than having no framework: it dresses a wrong conclusion in methodology.

2. QMCO and NIQ are the same category: the second-tier earnings released between the 8/10 close and the 8/11 open were missed. The pre-market list covered 8 companies that had significant pre-market moves on earnings that day, but what it covered was the batch that had already been written up by the media as pre-market movers; small- and mid-cap earnings like QMCO (market cap about $460 million) and NIQ, which do not make the mainstream pre-market movers lists, were missed outright. The cost was writing off two hard earnings reports — one "EPS swinging from loss to profit" and one "full-year guidance raised 13%" — as "noise".

3. ⚠️ The pre-market list's internal notes had in fact already flagged this, in these words: "the moves in ACVA and FRMI clearly do not match the strength of the news this report can find… if they trade well today, it means this report missed a piece of news and we must check back at recap time." The warning was accurate, but it was written into an internal note and never changed the conclusion labels in the body. The result: the ACVA half of the warning was right (it really was merely in line, and traced a round trip to flat on the day); the FRMI half was right too (it really had missed news), but the state "we may have missed news" was labeled "avoid" rather than "evidence pending, no rating".

From this comes a rule that should take effect immediately: "cannot find news" and "confirmed there is no news" are two different conclusions and cannot share one label. The former is an information gap on our side, and the correct treatment is to tag it "insufficient evidence, no rating" and put it on a follow-up list; only the latter deserves "avoid". Today's cost: a correct tiering framework was turned, by three missing facts, into a demonstrably wrong one.

2.4 Hit rate

The 11 "watch only" names are excluded from the hit rate (it is not a directional conclusion), but the results are worth listing separately: 5 called right (ACVA's round trip to flat, CAH's spike and fade, LITE's after-hours IV crush, NVDA, AMAT), 5 missed (CIFR +5.39%, PLUG +5.21%, WULF +3.40%, APLD +2.17%, plus CRWV at +13.47% after hours), 1 neutral (SMCI).

One line of self-criticism: today was not a problem of judgment, it was a problem of coverage. The top name on the long side (SE +14.56%) was picked correctly, all four verification levels ($21.00 / $11.50 / $13.26 / "thin-volume pre-market prices are not credible") were triggered that day and all four worked, and reclassifying INTC from "dilution negative" to "financing friction" was a precise hit (+0.19%) — the quality at the criteria layer was among the best of any day this quarter. But the top three gainers of the day were all on my "avoid" list, and all three had findable company disclosures. A report's ceiling is set by its criteria and its floor by its data gathering; today the floor dragged down the ceiling: an overall hit rate of 68% (55% on an executable basis), whereas if the FRMI/QMCO/NIQ news had been in hand at the time, they should have ranked No. 2 in "AI power (has a contract)" and near the top of "double beat on earnings" respectively — this is not three labels being wrong, it is three names that should have made the list not making it.

---

3. Theme Verification

3.1 The unexpected theme the pre-market report missed: power equipment and independent power producers (IPPs)

This was the real leading cluster today, and the pre-market list did not contain a single one of them.

The pre-market list's "AI power" universe was made up entirely of miners and data center developers (RIOT, BW, CIFR, WULF, APLD, HUT, CLSK, MARA, FRMI), with not a single generation or power-equipment company. But what the money bought that day was the whole chain: from generation (CEG/TLN), to equipment (GEV/ETN/VRT), to engineering (PWR), to sites (RIOT/FRMI). The reason BW was designated pre-market as "the other end of AI power" was precisely the logic of "supplying generation and thermal equipment" — once that logic holds, GEV, ETN and VRT in the same box must enter the universe alongside it, and their performance that day was far steadier than BW's (BW gave back from +37% to +3.49%, while VRT/ETN climbed steadily all day).

The lesson is the other side of the same coin as Section 2.3: 2.3 was "names that should have entered the universe did not, because news was missed"; here it is "names that should have entered the universe did not, because the theme definition was drawn too narrowly". If "AI power" is defined only as "who converted a mine into a data center", it will systematically miss the highest-quality, lowest-volatility tier of names within the theme.

3.2 How the "has a contract vs has no contract" criterion fared today

The core claim of Section 7.1 of the pre-market list was: names with their own disclosed contracts (RIOT, BW) will outperform those with only a peer-group label (CIFR/WULF/APLD/HUT/MARA/FRMI). The result on the day:

On the surface the criterion failed. But put FRMI back where it truly belongs (it has a binding 222MW / $6.5 billion lease) and the two groups become:

Not only does the criterion hold, its strength on the day exceeded the pre-market expectation — the has-contract group's average is 3.8x the no-contract group's, and the single biggest gainer was the one that had just signed a contract that day. This must be carried into tomorrow unchanged: the methodology of that pre-market tiering table was validated by the market today in the most direct way possible; the only link that failed was the list of facts used to classify.

⚠️ One limitation that must be stated alongside it: this is a cross-section of 9 names over 1 trading day, and the group averages are dominated by FRMI's single +21%. It is enough to support "keep using this criterion tomorrow"; it is not enough to support "having a contract necessarily outperforms".

---

4. After-Hours Earnings Moves

Quotes are the 18:00–18:06 ET after-hours snapshot; after-hours depth is far thinner than the regular session, and prices will be reset at tomorrow's open.

Point by point:

① CRWV — of the four verification points listed pre-market, tonight answered two, and both skewed positive.

• Verification point ① (whether interest expense breaks the $650–730 million guidance ceiling): it did not; actual was $640 million, below the low end of guidance. A year earlier it was $267 million, so it is still up 1.4x year over year, but for the first time it did not continue to exceed the company's own budget. This is, in this report's view, the most underrated number of the night — the pre-market report defined CRWV's core risk as "interest is eating all of operating income", and this quarter interest growth came in below the company's own guidance for the first time.
• Verification point ④ (whether the near-dated share of backlog holds at 25%): contracted backlog rose from $99.4 billion to $104 billion, ⚠️ but the share recognizable within 24 months was not obtained for this report, so it is impossible to tell whether this is "a bigger number, further out in delivery" or a genuine improvement. This is the first number to fill in once tomorrow's call transcript is out.
• Verification points ② (whether deferred revenue fell sequentially for a second straight quarter) and ③ (whether FY26 capex is revised up above $35 billion): neither was obtained for this report.
• ⚠️ Two numbers that must be read alongside the beat: net loss widened 116% to $626 million; adjusted EBITDA margin fell from 62% to 59%. Revenue beat while unit economics were still deteriorating. Media tallies also put Q2 free cash flow at about −$5.7 billion.
• Characterization: this is an earnings report of "demand side fully confirmed, financing side not yet cleared". The +13.47% is buying the $104 billion backlog and the raised full-year guidance, not profits.

② SMCI — the pre-market breakdown was half right; the half it missed was the magnitude of the FY27 guidance.

• The pre-market judgment that "all that is left tonight is FY2027 guidance and the cash flow statement" was entirely correct — revenue, gross margin and the $60 billion order book had indeed all been disclosed on 7/21.
• But the pre-market report treated "only FY27 guidance is left" as low information content, when in fact it was the single biggest number on the board: FY27 revenue guidance of $65–72 billion against FY26 of about $39 billion implies +66% to +84%. That is not a supplementary detail, it is guidance that redefines the size of the company. "Few pieces of information" does not equal "little information" — this is the methodological hole in this report: I estimated information content by the number of undisclosed items rather than by the variance of the undisclosed items.
• ⚠️ Two negatives that capped the move: ① FY26 full-year operating cash flow of −$6.8 billion (the company is building inventory and capacity); ② the board is conducting an independent review of export-control matters, and these results are preliminary and unaudited (24/7 Wall St.⁹). That most likely explains why revenue guidance implying +66%~+84% bought only +5.82%.
• Also note: revenue of $11.12 billion came in below the $11.56 billion consensus, consistent with the 7/21 pre-announcement of "landing at the low end of the $11.0–12.5 billion guidance".

③ LITE — the cleanest lesson of the night: a large beat + a large guidance raise = +0.05% after hours.

• Revenue hit the top of its own guidance range, adjusted EPS of $3.23 came in nearly 6% above the $3.05 top of guidance, next-quarter EPS guidance of $4.05–4.35 steps up another 25%~35%, and the target operating model was reached more than a quarter early — on earnings quality, this was the best of the three reports tonight.
• After hours: +0.05%.
• The pre-market judgment on LITE was "an extreme combination of bullish operations and bearish valuation; P/S 25.4 and EV/EBITDA 129 demand consecutive quarters of beats plus continuous upward revisions — an extremely high bar". Tonight it delivered the consecutive beat and the raise, and the stock's answer was: it is already in the price.
• ⚠️ The pre-market report gave the "purest IV crush" label to SMCI; the one that actually demonstrated IV crush tonight was LITE. The criterion was right (low information increment → IV crush) but applied to the wrong object: SMCI's information increment was low because the numbers were released early, whereas LITE's was low because the valuation had already discounted the good result — and the latter is the more common cause of IV crush.

---

5. Flows and Sentiment

Sector rotation (the 11 SPDR sector ETFs, in descending order of closing change)

Non-SPDR references: TAN (solar) +1.62%, ITB (homebuilders) +2.14%, SMH (semiconductors) +0.62%, JETS (airlines) +0.83%, XOP +1.39%, OIH +0.96%, IBIT −0.80%, ARKK +0.20%, XBI +0.03%.

There is only one way to read this table: today "real assets + power + industrials" outperformed "platform tech".

• The three leaders XLE / XLU / XLI are one line — rising oil prices, AI electricity demand, and the industrial companies that build the infrastructure for both. ITB +2.14% moving in the same direction as XLU +1.16% shows this was not purely defensive buying (in a defensive move utilities rise but homebuilders usually do not).
• The laggards XLC / XLY / XLK are also one line — GOOGL −3.84% alone is enough to explain XLC's −0.50%, AMZN −2.09% explains XLY, and AAPL −1.09% with MSFT −0.44% explains XLK. The attribution offered by the media is "the succession of new financing moves by Intel, Nvidia and others has renewed scrutiny of the intensity of AI capex" (Yahoo Finance²). ⚠️ This is a media attribution, not a verifiable fact; this report found no company-level negative specific to Alphabet on 8/11.
• A contrast worth recording: on the same day, "the people spending money on AI" (the capex spenders GOOGL/AMZN/MSFT/META) fell, and "the people collecting money from AI" (CEG/TLN/VRT/ETN/RIOT/FRMI) rose. This is not "AI has peaked", this is the expected profit split of the AI value chain shifting upstream toward power and infrastructure. It is the most informative flow signal of the day and the one most worth tracking tomorrow.

VIX and Treasuries: two signals pointing in opposite directions, both saying "nobody is paying for CPI"

• VIX 15.28 (−1.16%), VIX9D 12.52 (−1.96%). With only one night to CPI, short-dated and standard volatility fell simultaneously. VIX9D at 12.52 means the market is pricing annualized volatility of under 13% for the next 9 days (which include tomorrow morning's CPI).
• 10Y 4.694% (+1.0bp), 30Y 5.247% (+1.2bp), 2Y 4.222% (+0.4bp). The long end rose more than the short end, the curve continued to steepen slightly, and the term premium did not come back down. TLT rose only 0.16%; the bond market was not going long.
• Putting it together: equities are cutting exposure, the bond market is waiting, and the volatility market is collecting premium. Three markets show three different levels of preparation for the same event, and the volatility market is the least prepared of them. This is consistent with yesterday's pre-market conclusion, and the asymmetry has become more extreme.

Risk-on / risk-off call: shallow risk-off, but money did not leave — it just changed boxes. Criteria: ① indices fell but the Russell rose, and while breadth was poor there was no panic selling; ② VIX fell rather than rose; ③ high-beta ARKK +0.20% and XBI +0.03% did not fall with the market; ④ what actually got sold were the few largest-weight, most crowded large-cap tech names. This is "positioning shifting from crowded to uncrowded ahead of an event", not de-risking.

---

6. Next-Day Outlook (2026-08-12, Wednesday)

6.1 Theme continuation

6.2 Tomorrow's earnings and macro calendar

6.3 Key names to watch (ticker + verification point)

6.4 What to avoid

1. Chasing today's leading cluster (XLE / XLU / CEG / TLN / VRT / ETN) before CPI is released. This group has already run up several days, and utilities and power equipment are highly sensitive to long-end rates — if CPI runs hot they get hit twice (rates up + prior gains).
2. Chasing CRWV / SMCI's after-hours gains by buying at tomorrow's open. After-hours depth is extremely thin and prices will be reset tomorrow morning; RIOT (the open was the high of the day) and BW (open $11.86 → low $9.00) just demonstrated the price structure after a gap today.
3. Treating tonight's three earnings reports as one category of event. CRWV is demand confirmation (+13.47%), SMCI is "big guidance + big blemish" (+5.82%), LITE is "perfect report + zero reaction" (+0.05%). What each implies for tomorrow is completely different, and in particular CRWV's strength cannot be used to infer the direction of LITE / COHR — LITE tonight already proved that spillover does not hold.
4. Shorting bearish names at 52-week lows (UPWK actually closed 3.98% above its pre-market price today, HIMS 2.86% above, RKLB 4.29% above). Today the pre-market gaps on the bearish side were systematically bought back in part, which is a new piece of evidence from this reconciliation.
5. Chasing QMCO (+64.69%) and NIQ (+41.95%) the day after today's spike. What they rose on was already-published earnings, so the information is fully out; the reason this report names them is to record the error, not to chase the move belatedly.
6. Building positions on a single script of "mild CPI → broad risk-on". Today's flow signal is rotation between sectors rather than a change in aggregate exposure, and the most likely outcome after CPI lands is likewise accelerated rotation, not everything rising or falling together.

6.5 Input notes for tomorrow's pre-market list

1. [Highest priority] Decouple "news scanning" from "conclusion labels". Today's lesson is the three names in Section 2.3. Tomorrow morning's pre-market list must add a new label tier: "insufficient evidence, no rating", specifically for names that moved pre-market but for which we cannot find news. It is not the same as "avoid". Today FRMI/QMCO/NIQ were all killed off by the implicit equation "cannot find = avoid", and all three were the top three gainers of the day.
2. [Highest priority] Expand the earnings scan for the window from the 8/10 close to the 8/11 open. Neither QMCO nor NIQ, both missed today, was on a mainstream "pre-market movers" list. Tomorrow morning's data gathering should independently run through a complete list of "every company reporting after the previous close plus before the current open", rather than relying only on media-curated movers lists.
3. Expand the "AI power" universe from miners to the full chain. Permanent members must include: generation (CEG / TLN / VST), equipment (GEV / ETN / VRT), engineering (PWR), sites (RIOT / FRMI / CIFR / WULF / APLD). Today the best names in this theme (other than FRMI) were all outside the pre-market list's field of view.
4. Inherit the "has a contract vs has no contract" criterion; it was strongly validated today (corrected has-contract group +9.64% vs no-contract group +2.54%). But pair it with a pre-check: before tagging a name "no contract", we must first confirm that its company filings over the prior 24 hours have been searched, not just media coverage.
5. Keep the "vs pre-market price" column, but yesterday's pattern did not repeat today. Of the 19 names that gapped up pre-market today, 9 closed below their pre-market price and 10 closed above; on 8/10 it was 21 out of 25 below. The two days point in opposite directions, which shows that "pre-market price = high of the day" is not a stable rule but a feature of that day's market structure. Keep recording both columns side by side tomorrow, but stop using it as a prior discipline to push rankings down — it was precisely that discipline today that ranked BW, up +37% pre-market, behind SE at +6.82% (this time that was right), but the same logic will misfire tomorrow on a genuinely strong gap. The correct use is "a pre-market gain earns no bonus", not "a pre-market gain incurs a penalty".
6. Three numbers from the CRWV call must be filled in tomorrow morning: the share of backlog within 24 months, the sequential change in deferred revenue, and the FY26 capex ceiling. This report could not obtain them given its timing, and they set the pricing baseline for the AI compute chain tomorrow.

---

Internal Operations Notes (not sent to clients)

Data-gathering failures and workarounds:

• yfinance was rate-limited by Yahoo on this machine again: YFRateLimitError: Too Many Requests, failing at the cookie/crumb stage, unable even to fetch 5 days of SPY history. All single-stock and ETF data in this report was switched to stockanalysis.com's /api/quotes/s/<TICKER> endpoint, with zero failures (3 batches, about 80 names in total). This matches the memory yfinance-rate-limit-root-cause; stockanalysis remains the most reliable substitute.
• stockanalysis after-hours fields were fully available today: ep (after-hours price), ecp (after-hours change), eu (after-hours timestamp), cl (previous close), o/h/l/p (open-high-low-close) all had values. This is a positive instance of the same pattern recorded in the memory cnbc-extended-hours-asymmetry — "after hours works, pre-market does not" — so this data path is safe for the after-hours slot. Also, o/h/l allowed this report to compute intraday structure for the first time — BW ($11.86 → $9.00), ACVA ($8.27 → $6.67 → flat close), CAH ($258.30 → $240.26) — and the method in the memory recap-needs-ohlc-not-just-close produced substantive conclusions in three places here.
• CNBC quote API: indices/VIX/VIX9D/Treasuries/DXY/BTC all worked, but no commodity futures could be fetched at all. Tried @CL.1, @BZ.1, CL.1, BZ.1, @CL.26Q, @GC.26Z, @GC.1 — either None was returned, or stale overnight values (UNCH) timestamped 2026-08-11T01:00:00. This is the same class of failure as yesterday's pre-market inability to fetch .SPXFUT / @ES.1; CNBC's futures symbol system remains unusable on this machine. Switched to expressing direction via the closing changes of the USO / BNO / UNG / GLD / SLV ETFs, quoting absolute prices only on a media basis and flagging them in the body. Recommendation: futures data needs a separate stable path; this is now two consecutive slots blocked by the same problem.
• WebFetch returned HTTP 403 from CNBC (the CoreWeave earnings piece), 404 from a guessed investors.coreweave.com URL, and a DNS timeout from stocktitan.net. CRWV's financial detail was ultimately scraped from the company press release as reprinted by Yahoo Finance (the key $640 million interest expense number came from that source), and the full-year guidance came from aggregated WebSearch results. CRWV's Q3 single-quarter guidance and capex ceiling were therefore not obtained for this report, as stated in the header and in 4.1.
• GOOGL's closing change came in two versions: the Yahoo intraday live blog said −2.87%, while stockanalysis and CNBC, two independent sources, both gave −3.84% (close $343.80, previous close $357.52). −3.84% was used; the media figure is presumably an intraday reading. This kind of cross-check is worth keeping as a habit — a percentage change from a single media source is often an intraday screenshot.

Most important finding of this round (methodological, must go into memory):

• "Cannot find news" was treated as "there is no news", and the cost was that the top three gainers of the day were all on the avoid list. FRMI (+21.09%), QMCO (+64.69%), NIQ (+41.95%) — the stated reason for all three was "no findable news". Checking after the fact, all three had substantive disclosures, and all of them fell inside the coverage window the pre-market list declared for itself (8/10 16:00 → 8/11 08:20) — not old news from outside the window, but a detection miss inside it.
• The most glaring case is FRMI: its news (a binding 222MW / $6.5 billion data center lease) was structurally identical to RIOT, which was rated the No. 1 theme that day, and the pre-market list had specifically built a "has a contract vs has no contract" tiering table that classified it as "pure concept / catalyst directness 0". The framework was right and was strongly validated by the market that day (corrected has-contract group +9.64% vs no-contract group +2.54%), but executing a correct framework on incorrect facts is worse than having no framework — it dresses a wrong conclusion in methodology.
• The pre-market list's internal notes had in fact warned about this: "the moves in ACVA and FRMI clearly do not match the strength of the news this report can find… if they trade well today, it means this report missed a piece of news". The warning was accurate, but it was only written into an internal note and never changed the labels in the body. This is a process defect: an "I may have missed something" in the internal notes must have a channel that carries it through to the conclusion labels in the body, otherwise it is just an after-the-fact disclaimer.
• A separate, independent coverage gap: the theme definition was drawn too narrowly. The real leading cluster in "AI power" that day was VRT / TLN / ETN / CEG / GEV / PWR + XLU, and the pre-market list included none of them; the universe was entirely miners and data center developers. Once the reason BW was included ("supplies generation and thermal equipment") holds, GEV/ETN/VRT must enter the universe alongside it.
• Information content is not the number of pieces of information (SMCI). The pre-market report correctly judged that "all that is left tonight is FY27 guidance and the cash flow statement", but inferred from that "the lowest information increment". In fact FY27 guidance of $65–72 billion versus FY26 of about $39 billion was the highest-variance number on the board. Information content should be estimated by the variance of undisclosed items, not by their count.
• The IV crush criterion was applied correctly to the wrong object (LITE vs SMCI). The pre-market report gave "purest IV crush" to SMCI (because the numbers were released early), while the one that actually demonstrated textbook IV crush was LITE: revenue above the top of guidance, EPS 6% above the top, next-quarter guidance up another 25–35%, and +0.05% after hours. IV crush caused by a valuation that has already discounted the result is more common than the kind caused by early disclosure, and also more insidious.

Reconciliation conventions on record (to be inherited by the next report):

• This report reconciled 33 names in total, of which 31 had a recoverable pre-market price (the pre-market list gave no gap% for HUT/CLSK).
• Yesterday's pattern of "21 of 25 closed below their pre-market price" did not repeat today: of the 19 names that gapped up pre-market, 9 closed below their pre-market price and 10 closed above. The two days point in opposite directions. Conclusion: this is not a rule, it is that day's market structure. The correction "a pre-market gain earns no bonus, but should incur no penalty" is written into item 5 of Section 6.5.
• A new observation on the bearish side: of the 11 names that gapped down pre-market, 4 (UPWK +3.98%, RKLB +4.29%, HIMS +2.86%, INTC +1.37%) closed above their pre-market price, i.e. that pre-market gap was bought back. This supports the existing discipline of "do not chase shorts at 52-week lows".
• The hit rate is given on both measures (68% on a closing basis, 55% on an executable basis), and the next report should keep this dual measure; a single measure systematically flatters the result (memory recap-vs-premarket-price-column).

Open gaps to be inherited by tomorrow's pre-market list:

1. Three numbers from the CRWV call: the share of backlog within 24 months, the sequential change in deferred revenue, and the FY26 capex ceiling. These are the pricing baseline for the AI compute chain tomorrow; this report could not obtain them because of its publication time (18:0x ET).
2. Raw NYSE/Nasdaq advance-decline counts were not obtained; this report has only a third-party percentage.
3. Futures settlement prices for crude and gold — CNBC futures symbols have been unusable on this machine for two consecutive slots, and another data path is needed.
4. Details of FRMI's TensorWave lease: the identity of the guarantor ("one of the world's leading AI companies"), the project financing arrangements, and the year-by-year recognition schedule of the $6.5 billion. This is the same class of risk as RIOT's "financing chain between signing and collecting rent", and FRMI's free cash flow last quarter was about −$448.5 million with a current ratio of 0.5. The 8/13 earnings report must be revisited.
5. The scope and timetable of SMCI's independent export-control review, and when these "preliminary and unaudited" results become final.
6. The five items left unclosed by the pre-market list (SE and BW balance sheets, HIMS's $347.4 million vendor rebate receivable, ONON channel inventory, the composition of UPWK's active-client attrition) were not advanced here — SE is the most urgent of them: it rose 14.56% today to lead the entire board, and the quality of its earnings still has no independent verification whatsoever.

---

⚠️ Risk disclaimer: This recap is an after-hours review of information and observations only and does not constitute investment advice. Data may differ in timeliness or definition; please rely on company disclosures / SEC filings, and do not use this directly as a basis for trading.