U.S. Market Recap | 2026-08-10 (ET) Monday

Coverage window: 2026-08-10 (Monday) 09:30–16:00 ET regular session, including the 16:00–18:05 ET after-hours session. Quote conventions:

• Indices, sector ETFs, VIX, Treasuries, the dollar, crude and gold come from the CNBC quote API, on a regular-session close basis (16:00 ET), with after-hours fields snapshotted at 18:00–18:05 ET.
• Single-stock "today's close / today's change / after-hours price / after-hours change / 52-week high-low" come from the stockanalysis.com quote API, snapshotted at 18:00 ET, and cross-checked name by name against CNBC's previous_day_closing.
• The two sources were reconciled name by name: for the 25 reconciliation names covered here, the "Friday (8/7) close" is identical name by name between CNBC and the pre-market report (SPCX 133.11 / COHR 379.13 / LITE 890.17 / AMAT 539.14 / SLN 11.95 / SION 51.04, etc.), so the baselines for both the "today's change" and the "change vs pre-market price" columns are reliable.
• Earnings figures come from company press releases and third-party reprints and were not checked line by line against the original SEC filings; treat them as "pending verification". Data not obtained for this report: ① NYSE/Nasdaq advance-decline counts (the A/D line) were not obtained; market breadth here is substituted with equal-weight vs cap-weight indices, plus the up/down distribution of the 11 SPDR sector ETFs, with the convention stated in Section 1; ② the after-hours earnings for CRWV / SMCI / LITE have not yet been released (all on 8/11 after the close), so this report can only give the schedule and the verification points, not results; ③ conference-call guidance details for RKLB / PLUG / ACHR (the calls overlapped with the window in which this report was generated), so only press-release-level figures are used here.

---

0. One-Line Recap

Today was a "the index didn't move, but everything underneath was swapped out" rotation day: the pre-market report picked the right catalysts but ranked tradability exactly backwards — the optical modules it ranked No. 1 were the worst performers of the day, and the energy names it tagged "watch only" were the strongest.

1. The calm at the index level is an illusion; underneath it is textbook sector rotation. S&P 7,753.11 (−0.06%), Dow 53,975.98 (−0.11%), Nasdaq 26,605.36 (−0.32%), Russell 2000 3,017.40 (−0.56%) — all four indices were down but only marginally. The real information is at the sector level: among the 11 SPDR sectors, XLE +4.66% led and XLRE −1.29% lagged, a top-to-bottom spread of 5.95pp; if thematic ETFs are included, semiconductors SMH −2.28% were the worst of the whole board, a 6.94pp spread versus XLE. This was not a broad decline; it was a move out of AI/semis and into energy.

2. The pre-market report's single largest misjudgment has to be stated up front: optical modules were not the leader, they were the biggest loser of the day. The pre-market report listed the "optical module China ban" as the No. 1 theme of the day and COHR as the No. 1 watch-closely name, on the grounds that it was "the only one that simultaneously had logical hardness + capital endorsement + remaining room". Actual result: COHR −14.24%, LITE −8.61%, CIEN −5.99%, AAOI −2.08% — all four wiped out, with COHR the single worst decliner on the whole list. More importantly, the verification condition set pre-market was itself correct — Section 9.1 spelled out "hold $393 with LITE/AAOI/CIEN moving in sync; a lone name rising = the last leg". COHR was hammered from a pre-market $393.80 all the way down to a $325.15 close; the verification condition was falsified within the first half hour of trading. The problem was not the absence of a gate, it was putting a "still to be verified" direction in the No. 1 slot.

3. The genuinely strongest theme today was energy — and it was the pre-market report's own hand that downgraded it. WTI closed at $82.13 (about +5.0%), Brent $87.72 (about +5.0%), driven by Iran raising its conditions for reopening Hormuz + a Houthi drone strike on Saudi Aramco's Jizan refinery (the latter was entirely absent pre-market). Oil services OIH +5.94%, E&P XOP +5.73%, XLE +4.66%; among single names HAL +5.49%, SLB +5.28%, OXY +4.90%, COP +4.61%, CVX +4.48%, XOM +4.41%. The pre-market report ranked XOM/OXY 18th/15th and tagged them "watch only" on the basis of "a divergence in which oil was up 1.66% while energy equities were up only 0.64%" — today that divergence was closed to the upside, not confirmed to the downside.

4. One quantitative fact covering the full sample is the most valuable conclusion of the day: of the 25 reconciliation names, 21 closed below their pre-market price this morning, and only 4 closed above (ABCL +7.98%, OXY +3.79%, XOM +3.75%, HZO +0.08%). The pre-market price was, on today's evidence, close to the high of the entire day. This is exactly the warning the pre-market report itself wrote in the first item of Section 9.3 — "the opening price is already someone else's settlement price" — it got the rule right, and still ranked names by strength measured at the pre-market price. See the two-column reconciliation table in Section 2.

5. Two after-hours developments: RKLB posted record revenue but a small EPS miss, −6.42% after hours; PLUG beat on revenue with gross margin turning to roughly breakeven, +7.58% after hours. Note that the "acquisition of Iridium" mentioned in RKLB's press release is a progress restatement of an old deal already announced on June 29, not a new catalyst today (see the dedicated note in Section 4.3).

6. Tone for tomorrow: defensive, with a sharp jump in information density. VIX 15.46 (+3.76%), 9-day VIX 12.77 (+6.77%) — the front end rose more than VIX itself, the classic signature of "buying insurance for the single-day Wednesday CPI event"; 10-year Treasury 4.705% (+4.1bp), 2-year 4.243% (+1.7bp), the curve steepened 2.4bp — the opposite direction to the pre-market call of "flattening"; rising oil pushed the term premium back up. Tomorrow after the close, CRWV + SMCI + LITE all report in the same window; the day after, CPI + COHR.

Today's one-line takeaway: the catalysts were read correctly, the tradability was ranked backwards. Pre-market gave optical modules "watch closely" and energy "watch only", and the day's outcome was precisely those two labels swapped. What today should really teach is not "recommend fewer optical modules", it is that "pre-market gain" cannot be used as a proxy for "intraday strength" — 21 of 25 names broke below their pre-market price, and that ratio says the problem is not stock selection, it is the pricing timestamp.

---

1. Market Overview

1.1 Indices and Volume

All four indices were down, but the ordering itself is the signal: the Russell 2000 (−0.56%) fell nearly 9 times as much as the S&P (−0.06%), small caps clearly weaker than large caps; the Nasdaq was weaker than the S&P, growth weaker than value. The S&P's Friday record close of 7,757.64 was neither rewritten nor broken — today was a sideways session at the highs.

1.2 Market Breadth (substitute measures; A/D line not obtained)

⚠️ The convention has to be stated first: this report did not obtain NYSE/Nasdaq advance-decline (A/D) data; the three sets of self-collected data below are used instead, and the strength of the conclusion is lower than a true A/D line.

Overall read: today was not a day of collapsing breadth, but a dislocation day of "heavyweights dragging the index, no bid in small and mid caps, and money concentrated into energy". Equal-weight outperformance says most S&P constituents were in fact flat; what actually fell were heavyweights like NVDA −2.86%, AAPL −1.53%, AMD −2.86%. And IWM's underperformance says the incremental money did not spill over into small caps — it just moved between sectors.

1.3 Sector Performance (SPDR sector ETFs, close basis)

Thematic ETFs (outside the 11 sectors): OIH (oil services) +5.94%, XOP (E&P) +5.73%, USO (crude) +6.73% all beat XLE itself — the elasticity ordering is "oil services > E&P > integrated energy", the classic signature of an energy move driven by risk premium rather than demand; IGV (software) +2.26%, ARKK +1.27%, XBI +0.42% closed green against the tape; SMH (semiconductors) −2.28% was the worst thematic ETF on the board, 1.4pp weaker than even XLK.

⚠️ One structure that has to be called out: software (IGV +2.26%) rallied hard while semiconductors (SMH −2.28%) sold off hard — both inside the technology sector, moving in exactly opposite directions. This says today's selling was not "selling tech" but "selling the AI hardware capex chain" — software does not consume capex, does not consume power, does not consume debt financing, so it was actually a beneficiary of this rotation. XLU −1.10% and XLRE −1.29% weakened in step, further confirming it: what was sold was the attribute set "asset-heavy + long-duration + financing-dependent", not any particular industry label.

1.4 VIX, Treasuries, Dollar, Commodities

Three readings deserve to be pulled out separately:

1. The curve steepened while the pre-market call was for flattening — the call was wrong, and the reason is explainable. Pre-market saw "2-year +2.2bp vs 10-year +0.6bp" and inferred that "the post-payrolls dovish pricing is being unwound". Today's actual move was the 10-year +4.1bp outpacing the 2-year +1.7bp. A +5% single-day move in oil directly lifts inflation expectations and the term premium, which acts on the long end; the front end is constrained by pricing of the Fed policy path and reacts more slowly. This correction has direct use for tomorrow: as long as oil keeps rising, the long end faces upward pressure, and that is a persistent headwind for long-duration assets (growth stocks, REITs, utilities) — XLRE −1.29% and XLU −1.10% today are already reflecting it.

2. VIX9D (+6.77%) rose nearly twice as much as VIX (+3.76%), the standard shape of "single-day event hedging". The 9-day tenor covers Wednesday's CPI and Thursday's PPI. Note that VIX9D at 12.77 is still below VIX at 15.46, so the term structure remains in contango — the market is buying insurance, not panicking. The first item of pre-market Section 8.4, "the calm at the index level is 'waiting for data', not 'already priced'" — that one was confirmed today, and front-end VIX provided cleaner evidence for it than VIX itself.

3. The dollar (+0.27%) and oil (+5%) rose together, while gold rose only 0.60%. All three moving the same way with gold weaker than oil points to a supply-side geopolitical shock being priced, not a currency debasement or a flight to safety. In a flight to safety, gold should beat oil and the dollar and oil typically move inversely. This combination supports the read that "the energy move is risk-premium driven", and it also means: the moment there is any softening in the Hormuz negotiations, the give-back will be fast.

1.5 Sentiment Read

Qualitatively: rotation-type risk-off, not systemic risk-off.

• Evidence supporting risk-off: VIX +3.76%, VIX9D +6.77%, small caps −0.56% underperforming, Bitcoin −1.91%, a stronger dollar, real estate and utilities both leading the declines.
• Evidence against systemic risk-off: equal-weight indices (RSP +0.06%, QQQE +0.16%) closed green, sectors at 5 up / 6 down were close to an even split, high-beta names like IGV software +2.26%, XBI biotech +0.42% and ARKK +1.27% rose against the tape, and the S&P is only 0.06% from its record high.
• Conclusion: money did not leave, it switched positions. Out of the "AI hardware capex chain" (semis, optical modules, AI cloud, utilities) and into "energy + health care + software". At the index level this is almost invisible; at the sector level the top-to-bottom spread is 6.94pp. That is exactly why today's recap cannot be done on the indices alone.

---

2. Pre-Market List Reconciliation

2.1 Two-Column Reconciliation Table (core)

Why two change columns: the pre-market report's conclusions were made on the pre-market price at 08:07–08:08 ET, so a reader executing the list as written would have a cost basis at the pre-market price, not Friday's close. Looking only at "today's change %" systematically overstates the pre-market list's performance (because for many names the gain had already happened pre-market). This table therefore gives both:

• Today % = versus Friday's (8/7) close, i.e. the standard daily change;
• vs Pre-Mkt % = versus this morning's pre-market price, i.e. the real P&L of "executing the pre-market list". The gap between the two columns is the part eaten by the pre-market gap.

2.2 Hit Rate

By label (25 names):

By the stricter "executed at the pre-market price" P&L basis:

• Of the 25 names, 21 closed below their pre-market price and only 4 above (ABCL +7.98%, OXY +3.79%, XOM +3.75%, HZO +0.08%).
• The average return of the 5 watch-closely names measured from the pre-market price is −6.75% (COHR −17.43, LITE −12.68, AMAT −6.55, CRWV −3.76, SPCX −0.32).
• In other words: buying the 5 watch-closely names at the pre-market price as listed would have lost about 6.75% equal-weighted today, while the S&P fell only 0.06%.

2.3 Self-Critique (three items, ordered by severity)

First, and most important: the pre-market report used "pre-market gain" as the ranking basis for "intraday strength", and the day's data shows the two are close to negatively correlated.

The core evidence for ranking COHR No. 1 pre-market was "+13.44% Friday, another +3.87% this morning, all four names in the group rising together". But "it has already gone up a lot" is a fact about the past, and it is simultaneously evidence of "capital endorsement" and evidence of "crowded positioning" — pre-market accepted only the first half. Today's full-sample data gives the answer: 21 of 25 names broke below their pre-market price, and the larger the pre-market gain, the more brutal the intraday give-back tended to be (SLN pre-market +45.94% → −11.52% versus pre-market; COHR only +3.87% pre-market but already +13.44% on Friday → −17.43% versus pre-market). The irony is that the pre-market report itself wrote in Section 9.3 that "the opening price is already someone else's settlement price", yet applied that rule only to SLN/HZO/VREX, the three names up more than 40%, without elevating it into a ranking principle.

Second, "pre-market divergence" was read uniformly as bearish, missing that it can also mean "not yet transmitted".

The fourth item of pre-market Section 8.4 listed "crude +1.66% while XOM/CVX were up only +0.64%" as a risk signal, and the third item of Section 9.3 went further, writing "do not express rising oil prices through equities". The result: oil rose 5% on the day and energy equities not only kept up but produced elasticity like OIH +5.94%. A divergence has two readings — either the equity is contradicting the commodity (bearish), or the equity has not caught up yet (a lag). The key to telling them apart is which session the divergence occurs in: pre-market liquidity is thin, and energy names had almost no real volume pre-market (the pre-market report itself wrote that XOM's pre-market volume was just 24,000 shares and STNG had no trades at all) — a "divergence" under those conditions simply does not constitute evidence. The pre-market report did in fact mark it "requires post-open verification" in Section 7.3, but by the Section 9.3 conclusions it had been written as a firm warning. The strength of the evidence was improperly upgraded on its way from the body to the conclusion.

Third, missing ABCL's move exposes a defect in the label system: "the data is already confirmed positive" and "the valuation is too expensive" were merged into the same label.

ABCL was the biggest gainer on the whole list (+34.78%), and the pre-market report's factual judgments about it were all correct — the data readout was confirmed positive, it had already broken its 52-week high pre-market, EV/revenue 33x. But the conclusion label was "watch only", the same label shared with names whose factual judgments were in doubt (SLN) and names whose room had already been eaten (HZO/VREX). These three situations are entirely different in nature: HZO's "watch only" is an arithmetic conclusion (1.77% left), SLN's is missing evidence, and ABCL's is a valuation judgment — and a valuation judgment is the least reliable of the three on a day when a binary event resolves. The suggestion going forward is to split "watch only" into at least three grades — "room exhausted", "insufficient evidence", "valuation too rich" — because they fail in different ways: the first two have a low probability of failing, the third fails very easily on a catalyst day.

⚠️ One counter-fact that must be stated at the same time, or the critique becomes distorted: the 85% hit rate on the defensive side was not luck. NABL (with the actual figures missing, reasoning from the size of the decline to a "guidance cut", closing −35.67%), FRO (going against the intuition of "oil up, buy tankers" to call it a double-edged sword — oil +5% while it fell 3.25%), SLN (refusing to infer the quality of the data from the share price, saving 11.5%), and the arithmetic judgments on SMCI and HZO/VREX — these were all correct calls with methodology behind them, and they should not be dragged down by the failure on the offensive side. Today's problem is ranking and position sizing, not research quality.

---

3. Theme Verification

3.1 Did Pre-Market Identify the Strongest Theme? — No, and the ranking was almost exactly inverted

• The optical modules ranked 1st pre-market → the worst today.
• The energy theme ranked 4th pre-market → the strongest today.
• The semiconductor equipment theme ranked 6th pre-market → second from the bottom today.
• The only correctly ranked theme was the top-tier M&A theme (S grade), but that is a theme with no room left to trade.

This is not a luck problem, it is a ranking-criteria problem. The pre-market ranking criteria are "logical hardness × capital endorsement × remaining room", and within those three, "capital endorsement" is measured by the gain over the past 1–3 trading sessions. On a rotation day, that term systematically ranks "yesterday's winners" at the top, and the very definition of a rotation day is "yesterday's winners become today's losers". Energy was ranked 4th precisely because it had not gone up yet pre-market — and "hasn't gone up yet" was, today, a reason to buy, not a reason to avoid.

3.2 Why Optical Modules Collapsed: Three Reasons, None of Which Is "the Ban News Was Falsified"

⚠️ First, to be clear: there was no reverse development at all on the FCC ban today. The theme's decline has nothing to do with the news itself; the reasons are:

1. Pre-earnings de-risking (the main cause). LITE reports FY26 Q4 tomorrow (8/11) after the close, and COHR reports FY26 Q4 the day after (8/12) after the close — the pre-market report wrote only "LITE reports this week", never pinned down the specific date, and never registered that both companies' earnings fall on precisely the two days when the ban theme was hottest. The market cut positions two days ahead.
2. The sheer height of valuation and prior gains. COHR is up about 233% over the past year, with a TTM P/E of roughly 160x; LITE is up about +141% year to date. At that level, any profit-taking — which needs no reason at all — gets amplified.
3. The "optical vs storage" fight over capital allocation. Weekend social-media discussion of "whether the bottleneck in AI data centers is optical interconnect or storage" triggered a reallocation of money within the theme. ⚠️ This one sits at the level of market narrative, has the weakest evidence, and storage names themselves were divergent today (SNDK +2.12%, WDC +0.93%, but MU −1.89%), so it cannot be said that money moved directly from optical modules into storage. It is listed here only for completeness and is not used as a basis for judgment.

Direct implication for tomorrow: the optical module decline is event-driven de-risking, not a break in the logic. This means LITE's earnings tomorrow after the close are the first real adjudication point for this theme — if LITE's guidance confirms the share shift created by the ban, today's decline was an overreaction; if the guidance is bland, then "policy tailwind ≠ near-term results" gets confirmed, and COHR takes another hit the day after.

3.3 Themes Missed Pre-Market: Software and Health Care

IGV (software) +2.26%, XLV (health care) +1.67%, XBI +0.42% — none of the three was among the 8 pre-market themes.

The pre-market theme list was driven entirely by "news added that day" (M&A, policy, geopolitics, clinical, earnings), and not one theme was driven by "where is money coming from and where is it going". Yet the two most profitable directions today (energy +4.66%, software +2.26%) were both on the receiving end of the rotation.

⚠️ This is a structural methodology gap, not a one-off today: a news-driven theme list can by construction only see "places with news", while on a rotation day money tends to flow to places that have no news but are cheap or under-owned. Software's rise today cannot be explained by any single piece of news (NABL was down 35.67%); it rose because it does not consume capex, does not consume power, and does not consume long-end rates — that is an attribute, not a headline. The suggestion is to add an "attribute dimension that does not depend on same-day news" to the pre-market list (duration, capex intensity, financing dependence); otherwise rotation days will be systematically missed.

---

4. After-Hours Earnings Moves

⚠️ The figures in this section come from company press releases and third-party reprints and were not checked line by line against the original SEC filings. After-hours quote snapshot 18:00–18:05 ET.

4.1 RKLB — Rocket Lab | After hours −6.42% ($80.04 → $74.90)

Reading the 6.42% after-hours decline: a big revenue beat but an EPS miss, and the market chose to price off EPS — the classic signal that "highly valued growth stocks no longer pay up for revenue growth". Combined with the pre-market report's flag of "earnings-implied move of about ±9.9%", the actual −6.42% after hours falls inside the implied range, making it a normal reaction rather than a collapse.

⚠️ One attribution correction that has to be spelled out: in relaying RKLB's press release, some financial media listed the "acquisition of Iridium (IRDM)" as a "strategic development" of this quarter's report. That transaction is an old event already announced on 2026-06-29 ($54/share in cash and stock, enterprise value about $8 billion, expected to close mid-2027, with a $3.6 billion bridge loan from Deutsche Bank and Wells Fargo), not a new catalyst today. IRDM closed at $49.32 today (−1.22%), still a 9.5% discount to the $54 deal price, reflecting the long closing timeline (about a year) and exchange-ratio volatility risk — nothing to do with today's earnings. Treating an old announcement as today's good news is the single easiest error to make in a recap, and the hardest to catch yourself.

4.2 PLUG — Plug Power | After hours +7.58% ($2.11 → $2.27)

Reading the 7.58% after-hours gain: the market completely ignored the large EPS miss and bought one thing only — gross margin repairing from −31% to 0%. For a company that has long survived on share issuance, gross margin turning positive is the first threshold where a "story" becomes a "business", and it genuinely carries more information than a single quarter's EPS. Free cash flow narrowing from −$230 million to −$100 million is a second piece of evidence in the same direction.

⚠️ But the pre-market "avoid" label should not be judged wrong because of this 7.58%. The pre-market reasoning was "share price $2.30, −49.8% below the 52-week high, a long history of dilution, earnings elasticity only 2 pts", and not one of those was overturned by tonight's report — gross margin has only just reached breakeven, the company is still lossmaking, still burning cash. And the arithmetic has to be done properly: pre-market price $2.30 → after hours $2.27, so anyone executing the pre-market list is still down 1.30% as of now. The "+7.58% after hours" is measured against today's $2.11 close, not against the pre-market price. This is exactly why this report insists on the two-column convention.

4.3 ACHR — Archer Aviation | Today +11.99%, after hours −1.44% ($6.26 → $6.17)

⚠️ A difference in convention has to be flagged: the consensus the pre-market report cited was "EPS −$0.25, revenue $1.95 million", whereas the FactSet convention is "EPS −$0.34, revenue $1.94 million". The actual EPS of −$0.34 is exactly in line versus FactSet, but a 36% miss versus the −$0.25 cited pre-market. This report adopts the FactSet convention (it has a named source), and flags that the source of that −$0.25 pre-market figure was not labeled and may be a different adjustment basis.

The real driver of today's +11.99% is not earnings, it is M&A: ACHR reached a definitive agreement with Boeing to acquire three Boeing subsidiaries — Wisk Aero, Insitu and SkyGrid; Boeing will receive newly issued Archer Class A shares and warrants, ending up with close to a 20% stake, and has committed to invest up to $55 million in Archer's next financing round, while retaining a cross-license to Wisk's core autonomous flight systems. The transaction brings Archer a profitable defense business with over $200 million in annual revenue across 35 countries.

⚠️ This is the second attribution correction to the pre-market report in this recap: pre-market read ACHR's +13.77% pre-market move as "a low-priced-stock speculative gap ahead of earnings with no anchor for expectations". In fact the M&A news was already embedded in that pre-market gap. What is interesting is that the conclusion label ("watch only") still turned out correct on results — after hours $6.17 is below the pre-market price of $6.36, so a buyer at the pre-market price loses 1.57%. But "the label was right" and "the reasoning was right" are two different things, and only the latter determines whether the method can be repeated. For a company with just $1.94 million of revenue, a $200 million revenue defense asset is a change in order of magnitude; that fact deserves a standalone re-rating rather than being waved away as "low-priced-stock speculation".

4.4 Other After-Hours Moves (not on the pre-market list; for reference tomorrow)

A pattern worth noting: UPWK and LIF were both up in today's regular session (+1.65% / +2.24%), and both fell about 20% after hours. Just like RKLB, this says the common adjudication point for tonight's batch of earnings is "guidance", not "the quarter's numbers". For CRWV / SMCI / LITE tomorrow after the close, guidance will likewise carry more weight than the quarter's beat/miss.

---

5. Flows and Sentiment

5.1 Sector Rotation: an "Attribute Rotation", Not an "Industry Rotation"

Inflow side: energy (XLE +4.66%, with oil services OIH +5.94% showing the most elasticity), health care (XLV +1.67%), software (IGV +2.26%), materials (XLB +0.61%). Outflow side: semiconductors (SMH −2.28%), technology (XLK −0.88%), real estate (XLRE −1.29%), utilities (XLU −1.10%), optical communications (COHR −14.24%).

Put the outflow list side by side and what they share is not an industry, it is a set of attributes:

And today the long end rose 4.1bp while oil rose 5% — the two things that simultaneously hit exactly the attribute set "long duration + high capex + high financing dependence". Sector performance is the result; the attributes are the cause. XLU −1.10% is the best evidence for this logic: utilities are supposed to be a beneficiary of the AI power theme, yet today they fell alongside semiconductors, because they are simultaneously the longest-duration and most financing-dependent sector.

5.2 NVDA's $500 Billion Financing Platform, and the Contradiction with Its 2.86% Decline That Day

On the most important piece of news in the AI chain today, the market gave a negative answer. Nvidia signed a memorandum of understanding (MOU) with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR — six of the world's largest private capital pools — to set up a standalone compute financing platform that would mobilize more than $500 billion of third-party capital for its customers (frontier AI labs, enterprises, AI clouds).

⚠️ Two conventions must be stated clearly: ① the $500 billion is a "long-term mobilization target", not a committed capital pool; ② at present this is only an MOU, with definitive agreements still to be signed.

NVDA closed at $217.55 (−2.86%) on the day. A positive of that magnitude producing a nearly 3% decline means the reaction itself carries more information than the news:

The market's reading is "Nvidia has to find money for its customers, which means its customers don't have money themselves". When a chip supplier starts arranging financing for downstream demand, that can mean either "demand is so strong it is worth unlocking with financial engineering" or "demand has already outgrown what customers' own balance sheets can carry" — today the market chose the latter reading.

This maps directly onto CRWV tomorrow after the close: the pre-market report already worked out CRWV's $8.06 billion of debt maturing within 12 months vs $2.27 billion of cash on hand, D/E of 7.39, and an interest coverage ratio of −0.08. With the market having just expressed its skepticism about "financialized AI supply chains" via −2.86%, the "refinancing arrangements" line item will carry more weight in CRWV's report tomorrow night than pre-market estimated. On the same day, INTC announced a $15 billion common stock offering (underwriters may add another $2.25 billion, for a possible total of $17.25 billion) and closed down 4.06% — the two events point to the same judgment: the market is currently extremely sensitive to how AI capex is financed, and its tolerance for dilution and leverage has dropped markedly.

5.3 VIX, Treasuries and the risk-on/off Read

• VIX 15.46 (+3.76%), VIX9D 12.77 (+6.77%): the front-end gain is about 1.8x that of VIX, with demand for insurance heavily concentrated in the next 9 days (covering Wednesday's CPI + Thursday's PPI). But VIX9D is still below VIX, the term structure is not inverted, which says this is "event hedging" and not "panic".
• 10Y 4.705% (+4.1bp) > 2Y 4.243% (+1.7bp), the curve steepened 2.4bp: oil-driven inflation expectations acting on the long end. This is the most direct source of pressure on long-duration assets today.
• Dollar +0.27%, gold only +0.60%, Bitcoin −1.91%: safe-haven money did not mobilize at scale, and gold being weaker than oil further says this is a supply shock rather than a flight to safety.

Qualitative conclusion: rotation-driven risk-off. Fear gauges up moderately, equal-weight indices closing green, high-beta ARKK/XBI rising against the tape — none of that supports "money leaving". Meanwhile small-cap underperformance, Bitcoin falling, a stronger dollar and a higher long end — those support "risk appetite declining at the margin". Put together, the accurate description is: money stayed in the market, but moved out of the most aggressive, most financing-dependent positions and into ones with more certain cash flow.

---

6. Outlook for Tomorrow (2026-08-11, Tuesday)

6.1 Theme Continuity

6.2 Tomorrow's Earnings and Macro Calendar

Macro (8/11, Tuesday):

• NFIB Small Business Optimism Index (typically 06:00 ET) — small in magnitude, not a market event.
• ⚠️ There is no heavyweight macro data tomorrow. This week's macro adjudication points are all after Wednesday: Wednesday 08:30 ET July CPI, Thursday PPI, Friday retail sales + the preliminary University of Michigan consumer sentiment reading.

Earnings (8/11 after the close, three reports in the same window):

⚠️ Scheduling risk (continuing the flag from pre-market Section 5.4b, and more serious today): three earnings reports land in the same after-hours window, and they belong to two themes currently moving in opposite directions (AI compute falling, optical modules falling but with the logic intact). The three sets of guidance will contaminate each other's after-hours price discovery, especially since SMCI and CRWV are both in the AI hardware chain — either one's guidance will spill over into the other. Single-name price reactions tomorrow after the close are less trustworthy than usual.

6.3 Key Watch List for Tomorrow (Ticker + verification point)

6.4 What to Avoid

1. Chasing energy higher. For anyone not already on board, buying tomorrow after a +5% move is paying for someone else's risk premium. This is a geopolitical-news-driven theme, and news can reverse overnight (recall that last Friday the market was still trading "a deal is coming").
2. Bottom-fishing optical modules (before LITE reports). Today's decline is pre-earnings de-risking, and the earnings have not arrived. Buying the dip ahead of the adjudication point is betting on a binary event that resolves tomorrow night.
3. Chasing the immediate price reaction in CRWV / SMCI / LITE tomorrow after the close. Three reports in the same window means the first 30 minutes of price discovery will be cross-contaminated; and tonight RKLB / UPWK / LIF have already demonstrated the "good quarter, bad guidance → down anyway" pattern.
4. The combination of high duration + high capex + high financing dependence (REITs, utilities, capital-heavy AI cloud). As long as the long end keeps rising on oil, the headwind for that attribute set remains.
5. The two M&A arbitrage names (HZO / VREX). The room has gone to zero: HZO has 1.69% left, VREX about 2.4%, and today fully confirmed that "a pre-market buyer made 0.08% on the day".
6. ⚠️ Do not read PLUG's +7.58% after hours as a turnaround. Anyone who executed at the pre-market price of $2.30 is still down as of now (after hours $2.27). Gross margin has only just reached breakeven; the company is still lossmaking, still burning cash ($100 million per quarter).

6.5 Inputs for Tomorrow's Pre-Market List

Five concrete suggestions for tomorrow morning's pre-market slot, all derived from today's reconciliation:

1. ⚠️ Highest priority: the ranking criteria must change. Today proves that "pre-market gain / trailing 1–3 day gain" as a proxy for "capital endorsement" systematically ranks yesterday's winners at the top on a rotation day. Suggestion: change "trailing 3-day gain" from a positive factor to a neutral one, and make it a negative factor when the gain is excessive. Supporting data: 21 of the 25 reconciliation names closed below their pre-market price; the 5 "watch closely" names lost 6.75% equal-weighted when executed at the pre-market price, versus the S&P at only −0.06%.
2. Add an attribute dimension that "does not depend on same-day news". Of the two strongest directions today (energy +4.66%, software +2.26%), software had no same-day news at all. Suggestion: add a column to the theme table for duration / capex intensity / financing dependence, to capture the receiving end of rotations.
3. "Pre-market divergence" must not be written directly into a conclusion. Pre-market liquidity is thin (XOM's pre-market volume was just 24,000 shares), so price and volume in that session do not constitute evidence. Suggestion: all sector-level divergence judgments based on pre-market changes should be downgraded uniformly to "pending post-open verification" and must not enter the Section 9.x conclusions.
4. Split the "watch only" label. At minimum into "room exhausted" (HZO/VREX, low probability of failing), "insufficient evidence" (SLN, low probability of failing), and "valuation too rich" (ABCL, fails very easily on a day when the catalyst resolves). Missing ABCL's +34.78% today is exactly the third category.
5. Earnings dates must be pinned down to "which day + before/after the close". Today LITE was written as "reports this week" without pinning it to tomorrow after the close, which directly caused the failure to anticipate the pre-earnings de-risking in optical modules. Suggestion: for any name entering the top three of a theme, the earnings date must be verified name by name down to the specific date and session.

Names to pre-load for tomorrow's pre-market: LITE (earnings after the close, the optical module adjudication), CRWV + SMCI (earnings after the close, AI compute financing sensitivity), XLE/OIH (whether energy can put together a second day), COHR (whether there is a second round of de-risking ahead of its report the day after), and the 10-year Treasury and WTI (the two master switches of today's rotation).

---

⚠️ Internal Data-Collection Log (not for clients)

⚠️ yfinance was again unusable this time (YFRateLimitError), the fourth consecutive trading day. All data in this report was obtained through the dual CNBC + stockanalysis channels, with no fields left blank due to collection failure; the only thing not obtained was the A/D advance-decline count.

1. yfinance rate-limit status: yf.Ticker('SPY').fast_info throws YFRateLimitError: Too Many Requests outright. Consistent with the second root cause (pure IP ban) in the "yfinance-rate-limit-root-cause" memory; no curl_cffi spoofing was attempted this time (already proven ineffective on prior days), going straight to the alternate channel saved about 3 minutes. Recommend the U.S. slot demote yfinance from default channel to backup channel.

2. The two channels that actually worked (recommend making them standard):

   • CNBC quote API: quote.cnbc.com/quote-html-webservice/restQuote/symbolType/symbol?symbols=A|B|C&requestMethod=itv&exthrs=1&output=json. Field behavior in the post-close session is completely different from pre-market, which is an important new finding: after the close, last = that day's closing price (correct), change_pct = that day's change (correct), previous_day_closing = the prior trading day's close (correct and extremely useful), and ExtendedMktQuote.last / .change_pct return the after-hours price and after-hours change normally in the post-close session. ⚠️ This contrasts with the pre-market slot's record that "extendedLast returns None across the board pre-market" — CNBC's extended-hours fields are available after the close and unavailable pre-market. This asymmetry must be written into the data-collection notes of both slots.
   • stockanalysis.com /api/quotes/s/<TICKER>: in the post-close session p = today's close, cp = today's change, ep/ecp = after-hours price / after-hours change, u/eu = timestamps for the two moments. The eu timestamp is critical; this time it was confirmed name by name to fall between 17:39–18:00 ET, with no lag.

3. Cross-validation method (recommend making this a standard step for the recap slot): this time CNBC's previous_day_closing was reconciled name by name against the "Friday close" in the pre-market report, all 25 matched (SPCX 133.11 / COHR 379.13 / LITE 890.17 / AMAT 539.14 / SLN 11.95 / SION 51.04 / NABL 4.99 / AIOT 4.41, etc.). This step is the precondition for the "vs Pre-Mkt %" column to hold — as long as the baseline price checks out, back-solving the pre-market price from the pre-market report's Pre% is reliable. This technique upgrades "reconciliation" from qualitative to quantitative and is the highest-information-density column in this report; recommend writing it into SKILL.md.

4. ⚠️ One instance of "silent one-day lag" contamination that was intercepted and must be recorded: searching "stock market close August 10 2026 breadth" returned Zacks/Yahoo articles stating "S&P 500 gained 0.6% to end at 7,757.64, Nasdaq advanced 1.3% to 26,690.62". Those two numbers are the August 7 (Friday) close, not August 10's. My own CNBC data was SPX 7,753.11 (−0.06%) and IXIC 26,605.36 (−0.32%), and 26,690.62 is exactly today's IXIC prior close — checking the prior close exposed it immediately. Exactly the same shape as the "stale-source-silent-lag" memory: the article doesn't error, it just gives you yesterday's numbers. The "8 sectors advanced" line in the same article was therefore unusable, and this report substituted the self-collected 11 SPDR ETFs (5 up, 6 down). Lesson: any index level obtained from search must have its publication date checked by back-solving from a self-collected prior close.

5. A second piece of bad data in the same search: one source gave market breadth as "49.2% advancing / 57.3% declining" — the two sum to 106.5%, internally contradictory, discarded outright. This report therefore states honestly that "the A/D line was not obtained" and substitutes equal-weight vs cap-weight (RSP/SPY, QQQE/QQQ) + the sector ETF distribution. That is better than inventing a plausible-looking number.

6. Another piece of bad data not accepted: one source claimed TPC (Tutor Perini) was −82.8% after hours. The stock does not appear in stockanalysis's top 10 after-hours decliners (the leader was ZONE −23.50%), and the magnitude also contradicts the description of "just raised 2026 EPS guidance", so it was judged a data error and not written into the body.

7. ⚠️ The most important attribution interception in this report (same shape as the "ai-attribution-stitches-old-news" memory, and it nearly recurred today): on the first search for RKLB earnings, tradingkey's summary listed "announced an agreement to acquire Iridium" under "strategic developments", which reads as a new catalyst today. Follow-up confirmed the transaction was announced on 2026-06-29 (RKLB +16% and IRDM +25% that day), an old event being restated as progress in the earnings release. Taking it at face value would have produced the completely wrong causal claim that "RKLB announced the acquisition of Iridium today yet fell hard after hours". Section 4.3 of the body now carries this as a standalone correction paragraph. The discriminating technique matches that memory: always check the announcement date before citing; another corroborating clue is that IRDM was only −1.22% today, which would be impossible for a genuinely new announcement.

8. Two substantive attribution corrections to the pre-market report, worth recording separately:

   • ACHR: pre-market attributed the +13.77% gap to "pre-earnings speculation in a low-priced stock". The actual driver is the acquisition of Boeing's Wisk Aero / Insitu / SkyGrid subsidiaries (Boeing takes close to a 20% stake + invests up to $55 million), bringing a profitable defense business with $200 million+ in annual revenue. For a company with $1.94 million of quarterly revenue, that is a change in order of magnitude. The pre-market conclusion label happened to be right (after hours $6.17 < pre-market $6.36), but the reasoning was entirely wrong. "Right label" does not equal "right method", and this situation is the most dangerous kind because hit-rate statistics will never expose it.
   • Conflicting consensus conventions on ACHR: pre-market cited EPS −$0.25 / revenue $1.95 million (source not labeled); the FactSet convention is −$0.34 / $1.94 million. The actual −$0.34 is in line versus FactSet but a 36% miss versus the pre-market figure. The body has adopted the attributed FactSet convention and stated the difference. Lesson: consensus figures must carry a source, or the beat/miss conclusion drifts with the convention.

9. SMCI consensus was cut ahead of the print and the pre-market report did not catch it: pre-market back-solved a Q4 threshold of $11.56 billion from the FY26 consensus of $39.5 billion; the convention found today is about $11 billion (the low end of management's $11–12.5 billion guidance range). The two differ by 5%, in the direction of a cut. Section 6.2 of the body lists both figures and explains what it means (lower probability of a miss, and lower value in a beat). Lesson: consensus must be re-pulled the day before earnings; the consensus in the pre-market report may already be stale.

10. WebFetch 403 sites (consistent with the pre-market slot's record, confirmed again): thestreet.com 403. stockanalysis.com and fool.com fetch normally; this time the fool.com piece supplied two key additions (INTC's $15 billion offering, ACHR's Boeing acquisition), making it the highest-information-density WebFetch of the day; recommend adding it to the U.S. slot's standard sources.

11. Methodology self-assessment for this report (directly related to several memories):

    • The "vs pre-market price" column is this report's core addition; it upgrades "pre-market list reconciliation" from "was the direction right" to "would following it have made money". The result is stark: 21 of 25 names broke below their pre-market price. That number is more persuasive than any qualitative critique; recommend making it a required column for the recap slot.
    • Same origin as the "recap-needs-ohlc-not-just-close" memory, but going a step further: that one is about "using only the close records an overnight gap as a successful same-day recommendation" (the DOCS case); this report is the mirror image of the same problem in the U.S. pre-market slot — the cost basis of a pre-market recommendation is the pre-market price, and reconciling with the closing change systematically overstates the hit rate. Looking only at "today %", ABCL +34.78%, HZO +46.08%, SLN +29.12% and ACHR +11.99% would all be recorded as "big gainers on the list", when in fact, executed at the pre-market price, the last three were all losses.
    • Echoes the "premarket-breadth-before-depth" memory: the pre-market report did extremely deep financial-statement digging on the four AI compute names (P/S percentile, accounts payable tides, prepayment share), and all of that analysis is correct today but none of it is the reason for today's moves — today's reason was sector rotation. The depth was not wrong, but it was applied to the wrong question: it answers "should this be held long term", while the list has to answer "what to buy today". This is the third appearance of the same lesson as "financial-quality-is-not-a-short-term-ranker", and it should be formally written into the us-premarket SKILL.md as an ironclad rule rather than recalled from memory each time.

12. Two new facts recommended for saving to memory (newly discovered this time, not covered by existing memories):

    • The pre-market/post-close asymmetry of CNBC's extended-hours fields: ExtendedMktQuote is available after the close and returns None pre-market. This dictates that the pre-market slot must use stockanalysis for pre-market prices, while the recap slot can use CNBC alone.
    • The "vs pre-market price" two-column reconciliation method and today's empirical result (21/25 broke below the pre-market price), plus its dependence on cross-validation via previous_day_closing.

---

⚠️ Risk disclaimer: this recap is a post-close information review and observation only and does not constitute investment advice. Data may differ in timeliness or convention; please defer to company disclosures / SEC filings, and do not use this directly as a basis for trading.