US Market Recap | 2026-07-28 (ET) Tuesday

Coverage window: 2026-07-28 09:30–16:00 ET regular session + 16:00–18:05 ET after-hours Data pull time: 2026-07-28, approx. 22:05 UTC (= 18:05 ET, 2 hours into after-hours) ⚠️ Change of reconciliation baseline (important): reports/us/2026-07-28.md (today's pre-market list) does not exist — this machine did not generate a US pre-market report today. To keep the feedback loop intact, this recap instead uses Section 6 "Next-Day Outlook" of reports/us/2026-07-27-recap.md as the reconciliation baseline — that section gave, for today (7/28), 5 theme-continuation calls, falsifiable validation points for 6 watch-list names, and 6 avoid items, so it is no less reconcilable than a pre-market list. The hit-rate basis in §2 is therefore not directly comparable to 7/27, as noted in that section. Data sources and basis: single-stock/ETF closing prices, volumes and after-hours prices come from the stockanalysis.com quote API, timestamps uniformly Jul 28, 2026, 4:00 PM EDT (close) and 5:30–6:05 PM EDT (after-hours); indices, VIX, 10Y/2Y Treasuries, the dollar index and Brent come from the CNBC quote API; the S&P 500 and VIX were independently cross-checked with yfinance and matched digit for digit (7,428.78 / 18.21). ⚠️ yfinance rate-limited on a broad scale for a second consecutive day (indices, Treasuries, crude, FX all returned YFRateLimitError); only ^GSPC/^VIX were retrieved before the throttle hit; everything else switched to the dual sources above.

---

0. One-Line Recap

1. Today was not a continuation of the rotation, it was an escalation of it — a momentum unwind. The Dow +1.03% made a new interim high, XLV +2.36% and XLF +1.27% both set all-time highs, while the Nasdaq was −0.22% and SMH −3.45%. Yesterday's "money moving out of semis and into something else" widened today into a collective de-leveraging of the entire high-multiple AI chain, and it spread to places that were considered safe yesterday.

2. The most important conclusion of yesterday's recap was falsified today by its own validation point. The 7/27 recap elevated "not touching wafers means safe" (ARM/CDNS/RMBS rising against the tape) into a theme and wrote "seven to three, zero counterexamples." Today: ARM −8.11%, RMBS −8.99%, SNPS −1.33%, PLTR −6.08% — all four "wafer-free" names were sold off across the board, and ARM's decline (−8.11%) was even more than twice that of SMH (−3.45%). The real dividing line is not "does it touch wafers" but "is it a high-multiple momentum stock inside the AI narrative." See §3.

3. The most completely wrong call was the "one-day wonder" judgement on the oil-price chain. The 7/27 recap gave the oil-price beneficiary chain a "medium-low ⬇" and listed "chasing the oil-price chain" as avoid item #1. Reality: RCL +5.72%, CCL +4.09%, DAL +3.12%, AAL +2.74%, UAL +2.65%, LUV +2.05%, and JETS +2.87%, outperforming SPY by about 2.6pp for a second consecutive day. Most names rose more on day two than on day one.

4. The most completely correct call was semiconductor equipment. The 7/27 recap gave "semi equipment multiple compression" a high continuation ⬆, listed "bottom-fishing semi equipment" as avoid item #2, and named it explicitly: "KLAC's 7/28 after-hours report walks straight into the line of fire." Today KLAC −6.18% → a further −8.01% after hours, AMAT −7.82%, LRCX −7.54%; of 20 semiconductor names sampled, 19 fell, and the only gainer was NVDA (+0.25%).

5. Tone for the next day: FOMC decision day — direction is set at 14:00, not by today's tape. VIX 18.21 (−2.46%) actually fell on a down day for the Nasdaq — confirmation this is rotation, not panic. The 10Y Treasury yield was 4.606% (≈flat) and the dollar index 101.393 (−0.14%). 7/29 brings a triple overlay: FOMC decision 14:00 + Warsh press conference 14:30 + MSFT/META/ARM earnings after the close. The CME FedWatch data missing for two straight days has been filled in today: hold about 63.5–65%, a 25bp hike about 35–36.5%; the September hike probability has risen to about 82%.

---

1. Market Overview

⚠️ Three basis notes, must read:

1. Conflicting prior close on the 10Y: CNBC gives a 7/27 prior close of 4.604%, while the 7/27 recap recorded 4.641% (yfinance ^TNX basis). The 3.7bp gap is a basis difference between data sources at the bond-market close (17:00 ET) versus the equity-market close. Today's yfinance throttling makes a cross-check impossible. Take only the direction: the 10Y was essentially flat today; yesterday's decline neither continued nor reversed.
2. The Brent prior close of $88.36 does not match the $87.68 recorded in the 7/27 recap: CNBC's quote has rolled to the Sep'26 contract, whereas 7/27 used the front-month contract — this is a roll spread, not a data error. This table therefore does not accept the −3.80% absolute decline and switches to ETF cross-verification.
3. The WTI figure is unreliable: CNBC @CL.1 returned a timestamp of 01:00 ET and was marked UNCH, clearly not updated. This figure is for reference only and is not used as a basis for any conclusion.

✅ ETF cross-verification of the crude direction (bypassing the contract-roll issue): USO −3.42%, BNO −4.06%, OIH (oil services) −3.72%. None of the three ETFs has a roll-basis problem, and they agree: crude fell for a second consecutive day today, by roughly 3.4–4.1%.

Market Breadth

⚠️ NYSE/Nasdaq advance-decline ratio: no reliable data (second consecutive day). The set of figures retrieved — "NYSE 1,374 advancers / 773 decliners," "S&P 500 advance-decline ratio 1.19:1," "15.8 billion shares traded" — was checked and comes from the Zacks piece "Stock Market News for July 28," whose body text actually describes the 7/27 session (the Nasdaq level it cites, 24,932.08, plus XLK −0.9%, XLE −2.1%, XLP +1.5%, are all 7/27 values). This is the same trap the 7/27 recap fell into — that series datelines its headline one day ahead of its body text. Excluded, not cited.

Breadth is instead derived from our own ETF data:

Sentiment read: risk-on and de-risking happening at the same time — a "momentum unwind." Evidence: ① VIX fell 2.46% on a down day for the Nasdaq (in a systemic flight to safety, VIX could not fall); ② XLV/XLF at all-time highs; ③ 7 sectors up, breadth positive; ④ no safe-haven signature across equities, bonds or FX (dollar down, 10Y flat); ⑤ but the only heavy volume is still in the falling semiconductors. This is not the market cutting exposure, it is the market switching exposure; the money taken out did not leave, it went into healthcare, financials and staples.

---

2. Reconciliation: How the 7/27 Recap's "Next-Day Outlook" Played Out

Basis statement: this section reconciles the forward-looking judgements in §6 of the 7/27 recap, not a pre-market single-stock list. The 77.8% hit rate on 7/27 was on a "pre-market single-stock direction" basis and is not directly comparable to this section. This section counts three categories: theme continuation (5 items) · watch-list validation points (6 sets) · avoid directions (6 items).

2.1 Theme-Continuation Calls (5 items)

Theme subtotal: of 5 items, 2 fully correct, 2 half correct, 1 wrong.

2.2 Validation Points for Watch-List Names (6 sets) — the most valuable part of this reconciliation

The 7/27 recap wrote a falsifiable validation point for each name. Today all 6 sets of validation points got a definite answer, none left hanging. That is a methodological success, and is a separate matter from whether the directional call was right.

Validation-point subtotal: all 6 sets answered (6/6); 5 sets support the 7/27 judgement, and 1 set (ARM) answers with "falsified."

2.3 Avoid Directions (6 items)

Avoid subtotal: of 6 items, 3 correct, 1 half correct, 2 wrong.

2.4 Hit Rate and Post-Mortem

Directional judgements in total: 5 themes + 6 avoids = 11 items → 5 fully correct, 3 half correct, 3 wrong

Strict basis 45.5% · half-credit basis 59.1%

Validation-point design: 6/6 falsifiable and all answered today = 100%

Three post-mortems that must be recorded:

1. [Most serious · theme-level error] The new theme distilled from data yesterday was overturned today by the same body of data. The 7/27 recap wrote: "Semiconductors must be priced by splitting them on 'does it touch wafers'... seven to three, zero counterexamples," and listed it as the No. 2 highest-priority input to the next day's pre-market list. Today's counterexamples are overwhelming: ARM −8.11%, RMBS −8.99%, SNPS −1.33%, PLTR −6.08%, all "wafer-free," all down hard; ARM's and RMBS's declines rank in the top 3 among all 20 sampled semiconductor names. Root cause: a single day's cross-sectional difference with a sample size of 3 (ARM/CDNS/RMBS) was taken for a structural rule, and self-graded as "zero counterexamples." Three samples cannot support a conclusion as strong as "zero counterexamples" — the phrase "zero counterexamples" is itself a signal of overconfidence. The correct move was to tag it a "hypothesis pending verification," not "the most valuable actionable insight."

2. [Time scale · the same error committed in reverse for a second time] "One-day wonder" and "the long side only lasts 0 days" are wrong in the same place. The 7/27 recap had just examined "the default for a long trade should be 0 days, not 1–3 days" (the memory-chain lesson), and on the very same day committed the same class of error in the opposite direction on the oil chain: judging a still-running driver (crude down 3.4–4.1% for a second consecutive day) as "a one-shot release." The shared root cause of both errors is not a mis-set duration parameter but the failure to first ask "is the driver itself still there." The memory chain's driver (a cross-time-zone sentiment gap) was gone at the open, hence 0 days; the oil chain's driver (crude still falling) is still here today, hence there was a second day. What should be judged is the persistence of the driver, not a default number of days assigned to a theme.

3. [Single stock vs sector] The UHS error exposed a missing category: a single-stock negative must first clear the sector gate. UHS cut full-year guidance and fell 4.86% after hours; the fundamental negative is real. But today XLV +2.36% set an all-time high, and sector beta erased the single-stock negative outright and pushed it to +4.34%. The 7/27 scoring model has no "sector direction" input, so single-stock logic was evaluated in isolation. This is really two sides of the same problem as item 1 in §2.3 (the oil chain): both of today's wrong calls were "right at the single-stock/theme level, but the sector-level fund flow was never counted."

---

3. Theme Verification for Today

⚠️ Falsification Box: Where "Not Touching Wafers Means Safe" Went Wrong

Among software companies that equally "do not touch wafers," today produced two completely opposite groups:

• The fallers: ARM (AI architecture licensing), RMBS (AI memory-interface IP), SNPS (AI chip design tools), PLTR (the most expensive AI application name) — what they share: high valuations, and stories entirely hung on AI.
• The risers: IBM, CRM, NOW, CDNS — what they share: installed-base cash-flow businesses, valuations not propped up by an AI narrative.

The rewritten actionable insight (the most important output of this recap, and simultaneously a correction of yesterday's error): What was sold today is not "semiconductors," it is the "AI premium." Semiconductors are merely where the AI premium is most concentrated, which is why they fell hardest; but as long as a stock's valuation rests on an AI narrative, it was being sold today, wafers or no wafers. Counter-check: NVDA is the core asset of the AI narrative, yet it was the only gainer among the 20 sampled semiconductor names (+0.25%). That is the sole counterexample to this insight; the reason is in "Unexpected Theme B." ⚠️ This is a single-day cross-sectional inference over a 20-name sample observed for 1 day — per the lesson in post-mortem 1 of §2.4, it is explicitly tagged here as a "hypothesis pending verification," not a conclusion.

⚠️ Unexpected Themes Not Covered on 7/27

---

4. After-Hours Earnings Moves (7/28 after the close → next-day catalysts)

After-hours prices from stockanalysis.com, timestamps Jul 28, 2026, 5:28–6:05 PM EDT. EPS/guidance figures come from company press releases and financial-media summaries; other than BE, they were not checked line by line against the original SEC 8-K.

The one thing most worth remembering from this after-hours set: TER +13.53% and KLAC −8.01% happened on the same night. Both are semiconductor equipment companies, both fell intraday with SMH (−4.22% / −6.18%), and they went in completely opposite directions after hours. The difference is where they sit: KLAC does front-end inspection/metrology and is tied to fab capex — and "capex peaking" is precisely the core of the current bear narrative; TER does back-end test and is tied to AI chip unit volumes — the more AI chips ship, the more test demand there is, and revenue +104% year over year is the evidence. The "China equipment breakthrough + capex peak" negative hits capex exposure, not every equipment company. This distinction replaces the now-falsified 7/27 "does it touch wafers," and it is backed by earnings numbers, not just a price cross-section — one grade higher on the evidence scale.

---

5. Flows and Sentiment

Sector ETF Rotation Panorama (7/28)

11 SPDR sectors: 7 up, 4 down.

Key Readings

• VIX 18.21 (−2.46%), falling on a down day for the Nasdaq. This is today's single most important sentiment reading: in a systemic risk release, VIX could not fall. The market is selling AI, not selling stocks. Versus 7/27 (VIX spiked to 19.93 intraday), the level of fear has materially declined.
• XLV and XLF set all-time highs on the same day, while XLK fell 1.84%. This is the hardest evidence that "the money did not leave, it just changed seats."
• XLI −0.39% is a number that is easy to misread: BA +4.76% was a main engine of the Dow, but it was entirely offset by UPS −6.57%. The industrials decline is not cyclical weakness, it is a single-point logistics negative.
• XLU −0.35%, down a second consecutive day, while the 10Y yield was flat — consistent with the 7/27 read: this is the AI power narrative ebbing, not defensive money withdrawing. VRT −6.27% today provides new evidence; but BE +10.88% after hours puts counter-pressure on that conclusion, and the theme needs re-evaluation the next day.
• 10Y 4.606% flat, DXY 101.393 (−0.14%), 2Y 4.281% flat — on the day before an FOMC decision, bonds and FX barely moved, the classic pre-meeting lockdown. ⚠️ The "no data for the dollar index" gap left over from the 7/27 recap has been filled today.
• Volume mix (same pattern for a second consecutive day): INTC 152 million shares, NVDA 133 million shares, AAL 86.33 million shares, MU 58.72 million shares. The heavy volume is still mainly in the falling semiconductors; the rising sectors are still being absorbed on light volume.
• ✅ CME FedWatch (a two-day data gap, filled today): for the 7/29 decision, hold at 3.50–3.75% about 63.5–65%, a 25bp hike to 3.75–4.00% about 35–36.5%; the September hike probability has risen to about 82%. ⚠️ These are ranges, cross-derived from multiple secondhand accounts, not a first-hand snapshot from the CME website.
• The Conference Board consumer confidence index fell back to 90.8, with weakening perceptions of the job market.

Qualitative Conclusion

A "momentum unwind" — risk-on and de-risking happening at the same time. Supporting risk-on: VIX down, 7 sectors up, XLV/XLF at all-time highs, dollar weaker, 10Y flat, Russell 2000 up. Supporting de-risking: SMH −3.45% on heavy volume, high-multiple AI names (ARM/RMBS/PLTR/VRT/MRVL) collectively −6% to −9%, Nasdaq down a second consecutive day.

The two sets of signals do not contradict each other: the market is migrating from one style (AI momentum) to another (value/defensive/cyclical), and during the migration the headline index barely moves (S&P +0.21%) while the internal structure changes drastically. 7/27 called it "rotation"; today's magnitude has outgrown rotation — two consecutive days, spreading into IP/software/AI applications, and accompanied by two sectors making all-time highs. Per Baird, this momentum unwind has been running for six to eight weeks, and only in the last two days has it shown up at the index level.

---

6. Next-Day Outlook (Wednesday, 2026-07-29)

① Theme Continuation

② Tomorrow's Earnings and Macro Calendar (ET)

Macro (weight overrides everything)

• ⭐ FOMC rate decision 14:00; Chair Kevin Warsh press conference 14:30. Current pricing: hold about 63.5–65%, hike about 35–36.5%; September hike probability about 82%.
• What matters this time is not what they do in July, it is how the statement and press conference guide September. If Warsh confirms a September lean, the already multiple-compressed high-valuation AI chain takes a second hit.

After-hours earnings (weight second only to the FOMC)

• ⭐⭐ MSFT (FQ4): consensus EPS about $4.22–4.24, revenue about $87.50–87.67 billion. The keys are Azure growth, AI capacity supply, and data center capex guidance.
• ⭐⭐ META (Q2): EPS expected about $7.18–7.24. The keys are ad revenue growth and the size of the capex raise.
• ⭐ ARM (FQ1): down 8.11% today, with "7/27 was a short squeeze" now confirmed; earnings are its next binary event.

Note: MSFT's and META's capex guidance cuts both ways for the market — capex above expectations → positive for semiconductors, negative for software margins; capex below expectations → confirms "capex peaking," a third red candle for semiconductors. Both of today's main storylines are decided by this one number.

③ Watch List (Ticker + Validation Point)

④ Avoid Directions

1. Sizing up in any one direction before 14:00, with the FOMC decision and MSFT/META earnings on the same day. Today is the eve of a triple event overlay; the avoid item raised on 7/27 was proven correct today, and tomorrow's event density is higher.
2. Bottom-fishing KLAC. Still −8% after hours, and the real reason for the guidance disappointment remains unestablished (§4). Buying without knowing why the market is selling is gambling, not trading — exactly the criterion 7/27 applied to the FOMC.
3. Chasing TER's after-hours gain. After hours +13.53% implies an open about 13% higher. The most serious post-mortem in the 7/27 recap was that "names that delivered generally opened at their high" — TER is this week's most typical candidate. Watch first, do not chase.
4. Using "does not touch wafers" as a stock-selection basis. That conclusion was thoroughly falsified today by ARM/RMBS, and §3 formally retires it. Any list entry still using that framework must be deleted.
5. CAPR — 7/29 FDA advisory committee, down another 6.08% today. A binary event, unsuitable for either side (retained for a second consecutive day).
6. Trading BE as a positive before it reclaims $188.18. The +10.88% after hours only recovered part of the day's decline and is still below the 7/27 close.
7. CSGP / NXPI — after hours −13.19% / −4.72%, likely to catch down the next day. ⚠️ This item follows the same pattern as the 7/27 call on UHS, and that one was wrong (UHS +4.34% today). Per post-mortem 3 in §2.4, check the sector first: if real estate / auto semis are broadly strong tomorrow, this judgement can fail the same way.

⑤ Input Hints for the Next Pre-Market List (retraining feedback)

⚠️ First item (a process problem, higher priority than any research conclusion): the 7/28 US pre-market list was not generated, which cost this recap its intended reconciliation target. Please first investigate why the scheduled bin/generate.sh us premarket job did not produce reports/us/2026-07-28.md — with a link of the loop broken, two consecutive days of the retraining chain get distorted.

The following 7 items should feed directly into the 7/29 pre-market scoring corrections:

1. [Highest priority · revoke yesterday's conclusion] Delete the "does it touch wafers" scoring dimension. The 7/27 recap listed it as "the most valuable actionable insight" and self-graded it "zero counterexamples"; today it was thoroughly falsified by ARM −8.11% and RMBS −8.99%. Replace it with §4's "capex exposure vs unit-volume exposure" — the latter is backed by the earnings numbers of TER (revenue +104%) and KLAC (guidance disappointment), one grade higher on the evidence scale.

2. [New mandatory category · the shared root cause of today's two wrong calls] Single-stock scoring must first clear the sector-direction gate. Today's two wrong calls (UHS up against the tape +4.34%, the oil chain continuing to rise) share a root cause: the single-stock/theme logic was right, but the sector fund flow was never counted. Recommendation: before any single-stock conclusion, tag the direction of its sector ETF for the day and the past two days; when a single-stock negative conflicts with sector strength, downgrade rather than hold.

3. [Time scale · a correction to 7/27's correction] Do not assign themes a default number of days; judge instead whether the driver is still there. On 7/27 the "default for a long trade" had just been changed from 1–3 days to 0 days, and on the same day the opposite version of the same error was committed on the oil chain. The memory chain's driver (an overnight sentiment gap) vanished at the open → 0 days; the oil chain's driver (crude still down 3.4–4.1%) is still here today → there was a second day. The criterion is the persistence of the driver, not a duration parameter.

4. [Coverage gap · same class as 7/27] Pre-market must scan the Asian session tape, not just US headlines. The direct transmission path for today's US memory-chain crash was SK Hynix −14%, Samsung −13%. The 7/27 recap already missed an intraday catalyst by "only scanning overnight and pre-market news," and today is the second occurrence of the same gap. Recommend codifying it: the pre-market list must include the Asian/European session performance of SK Hynix, Samsung, TSMC and ASML.

5. [Confidence calibration · methodology] Prohibit drawing "zero counterexamples"-grade conclusions from a single-day cross-section with a single-digit sample. On 7/27, 3 samples (ARM/CDNS/RMBS) and 1 day of data produced "seven to three, zero counterexamples," overturned the next day. Recommendation: tag every newly appearing theme as a "hypothesis pending verification" on first appearance; it may only be promoted to a scoring dimension with at least 2 trading days or earnings numbers behind it. (§3's "what is being sold is the AI premium" in this report has already self-tagged as a hypothesis under this rule.)

6. [Data gaps · 2 of the two-day backlog resolved] ✅ Dollar index obtained (101.393, −0.14%); ✅ CME FedWatch obtained (hold 63.5–65% / hike 35–36.5%, September 82%). Still missing: NVDA 5-year CDS spread (second consecutive day), KLAC China revenue mix (second consecutive day), NYSE advance-decline counts (second consecutive day).

7. [Source hygiene · the same trap for a second time] The Zacks "Stock Market News for July X" series datelines its body text one day behind its headline. The 7/27 recap already had to re-date data because of this, and today's search hit the same trap again (the piece is headlined "7/28" but its body is actually 7/27: Nasdaq 24,932.08, XLK −0.9%, XLE −2.1%). Recommend codifying a blacklist rule in the pre-market/recap workflow: figures from that series are always to be treated as headline date minus one day, or simply not used.

---

Appendix: Data Gaps and Pending Items Actively Flagged by This Recap

Following the no-fabrication principle, the following are known but unconfirmed:

1. reports/us/2026-07-28.md pre-market list does not exist — this recap's reconciliation baseline was replaced with the next-day outlook of the 7/27 recap, and the hit-rate basis is not comparable to 7/27.
2. The real reason KLAC's guidance was "disappointing" was not established — the guidance midpoints are numerically above consensus ($1.16 vs $1.13, $4 billion vs $3.91 billion), yet the stock fell 8.01% after hours. Multiple sources merely say "disappointing" without explaining the mechanism. This report does not speculate.
3. KLAC China revenue mix — the core validation point set on 7/27, not obtained for two consecutive days.
4. NVDA 5-year CDS spread — the validation point set on 7/27; primary credit-market data not obtained for two consecutive days.
5. NYSE / Nasdaq advance-decline ratio — no reliable source for a third consecutive day. The retrieved data was checked and belongs to 7/27 (see §1) and has been excluded; the breadth read is instead derived from the 11 sector ETFs and the relative strength of DIA/SPY/IWM/QQQ.
6. Conflicting EPS basis for Visa — adjusted EPS $3.32 beat the $3.23 estimate, while another source reports GAAP EPS $2.97 below the $3.19 estimate. The two can coexist, but the 8-K was not checked, so both are presented side by side without adjudication.
7. The specific WTI figure — CNBC @CL.1 is stuck at a 01:00 ET timestamp and marked UNCH, and yfinance was rate-limited. The crude direction is cross-confirmed by three ETFs, USO −3.42% / BNO −4.06% / OIH −3.72%; the absolute figure is not accepted.
8. Two-source difference on the 10Y prior close — CNBC 4.604% vs the 7/27 recap's 4.641% (yfinance basis), a 3.7bp gap that could not be cross-checked today. Direction only.
9. CME FedWatch figures are ranges — cross-derived from multiple media accounts (63.5–65% / 35–36.5%), not a first-hand snapshot from the CME website.
10. After-hours EPS / guidance figures — other than BE (checked against the company press release), all come from financial-media summaries and were not checked line by line against company IR originals or SEC 8-Ks.
11. The causal link between Nvidia's $250 billion backstop and NVDA rising against the tape — an inference, not confirmed.
12. XLV / XLF "all-time highs" — relayed from media statements; this report did not independently review their historical price series to verify.
13. KO's "best single day since 2009" — relayed from media, not independently verified; the +5.00% move for the day is confirmed by the quote API.
14. SK Hynix −14% / Samsung −13% — relayed from CNBC, primary KRX data not obtained.
15. yfinance rate-limited on a broad scale for a second consecutive day — indices, Treasuries, crude and FX switched to the CNBC quote API, single stocks/ETFs to stockanalysis.com. The S&P 500 and VIX were cross-checked digit for digit between the two sources; the remaining items were not checked item by item for consistency.

---

Data Sources

• TheStreet — Stock Market Today (July 28, 2026): Dow rises amid earnings reports, chip sell-off¹
• Yahoo Finance — Stock market today: Nasdaq bounces back, S&P 500 and Dow rise as earnings, falling oil prices offset chip weakness (July 28)²
• CNBC — Micron, Nvidia fall after SK Hynix plunges nearly 15% as chip sell-off deepens³
• CNBC — Nvidia and OpenAI in talks for up to $250 billion backstop to fund AI infrastructure plans⁴
• CNBC — Nvidia's potential $250B backstop for OpenAI is another strike against the AI trade⁵
• CNBC — Stocks making the biggest moves after hours: Ford, Teradyne, Visa, KLA Corp and more⁶
• Benzinga — Boeing Flashes Recovery Signs: Cash Flow Turns Positive, Backlog Hits All-Time High⁷
• Yahoo Finance — Boeing Q2 revenue tops estimates as jet deliveries climb, cash turns positive⁸
• Businesswire — Bloom Energy Reports Record Second Quarter 2026 Financial Results and Raises Full Year 2026 Guidance⁹
• Benzinga — Bloom Energy Blows Past Q2 Estimates, Shares Jump¹⁰
• Investing.com — KLA Corp posts Q4 beat, shares tumble amid underwhelming guidance¹¹
• Benzinga — KLA Stock Dives Further After Q4 Report — Here's Why¹²
• Yahoo Finance — KLA CORPORATION REPORTS FISCAL 2026 FOURTH QUARTER AND FULL YEAR RESULTS¹³
• StreetInsider — Teradyne (TER) Tops Q2 EPS by 41c; offers guidance¹⁴
• FinancialContent / StockStory — Teradyne (NASDAQ:TER) Reports Upbeat Q2, Stock Jumps 12.8%¹⁵
• MarketScreener — Earnings Flash (V) Visa Posts Fiscal Q3 Adjusted EPS $3.32, vs. FactSet Est of $3.23¹⁶
• Benzinga — Visa Reports Strong Q3 Results: CEO Says 'Spending Remains Resilient'¹⁷
• Investrade — Mid-Morning Look: July 28, 2026 (sector rotation and the XLV/XLF new highs)¹⁸
• TradingKey — Fed July FOMC Meeting Preview: Rate Hike or No Change? (FedWatch probabilities)¹⁹
• PrimeXBT — Microsoft, Meta, Apple and Amazon report earnings within 48 hours on July 29 and 30²⁰
• ⚠️ Yahoo Finance / Zacks — Stock Market News for July 28, 2026 (checked: the body text is actually the 7/27 session; this report has excluded its figures)²¹
• Quote data: stockanalysis.com quote API (single-stock/ETF closes, volumes and after-hours, timestamps Jul 28 2026 4:00PM / 5:28–6:05PM EDT) · CNBC quote API (Dow/S&P/Nasdaq/Russell 2000, VIX, 10Y & 2Y Treasuries, ICE dollar index, Brent) · yfinance (^GSPC 7,428.78 and ^VIX 18.21, used for independent cross-check; all other tickers rate-limited)
• Reconciliation baseline: reports/us/2026-07-27-recap.md §6 "Next-Day Outlook" (substituted because that day's pre-market list is missing)

---

⚠️ Risk disclaimer: this recap is only a post-close summary of information and observations and does not constitute investment advice. Data may differ in timeliness or basis; please defer to company disclosures / SEC filings. It must not be used directly as a basis for trading.