2026-10-29 16:00 ET — AI-infra momentum (VRT, ORCL, STX) vs NOW/META, and the INTC/AMZN after-close gap risk

Slot: Thu 2026-10-29 16:00 ET hourly research · Status: written at the close, after-hours prints still settling. Screening/research, not investment advice.
TRAI data used: /api/latest premarket run 2026-10-28_premarket_2026-10-29T12-53-20-154Z (8:53 AM ET; daily stats = Oct 28 session), /api/events, /api/correlations (60d to 2026-10-22), /api/scorecard replay 2026-04-23 → 2026-10-21, /api/regime and /api/macro. Data-quality note: scorecard, correlations, regime and macro all still show asOf 2026-10-22. They have not refreshed for 5 sessions (see S-068).
Other data: Yahoo Finance daily bars for 10-29 (pulled at 16:05 ET). Yahoo after-hours quotes as of ~4:12 PM ET are thin and preliminary. Company releases, SEC exhibits and news are linked inline.

1. Focus and why

This report moves on from the 17:00 Oct 23 topic (oil majors and gold miners). TRAI's 8:53 AM run carried three AI-infrastructure longs off big Oct 28 days: VRT +16.48% (relVol 3.33, residual z 5.37), ORCL +10.73% (relVol 2.94, z 3.94) and STX +4.51%, plus SOFI +4.51%. It also carried two shorts: NOW (−3.0% on Oct 28, but indicated gap −18.66% after its print) and META (−1.24%). Today's question was which catalysts drove those moves, whether they hold, and what tonight's prints mean for Friday's gaps. INTC and AMZN reported at 4:01 PM, and SNDK reports tonight.

2. Businesses (one line each; detail in company notes)

• VRT: data-center power and thermal gear (UPS, switchgear, busway, liquid cooling). VRT note
• ORCL: database/apps plus OCI cloud, now a leveraged AI-capacity builder (Stargate/OpenAI). ORCL note
• STX: hard-disk drives (mass-capacity nearline for cloud storage). STX note
• SOFI: digital bank/lending/financial-services platform. SOFI note
• NOW: enterprise workflow SaaS. NOW note
• META: ads platform and one of the biggest AI capex spenders. META note
• Watch list: ETN, ANET, WDC, MU, DELL, SMCI (all in TRAI correlation cluster 1, or cluster 19 for WDC). INTC and AMZN report tonight.

3. Today's drivers (macro and tape)

• FOMC (Wed 10-28): held at 3.75–4.00%, vote 10–2. Cook and Hammack dissented for a hike. Futures price a strong chance of a December hike (CNBC, 2026-10-28; Fed statement).
• 8:30 AM data (Thu 10-29): Q3 GDP advance 3.4% (vs 3.1% expected), with equipment investment +20.9%. September PCE +0.5% m/m and 3.6% y/y; core +0.3% and 3.1%. Jobless claims 194k (CNBC, 2026-10-29; BEA). The data is hot and is not rate-cut friendly.
• Tape: At 2:40 PM, Reuters had S&P −0.63% (7,695), Nasdaq −1.66% and semis −2.0%, but the Russell 2000 +0.8%. "Alphabet revenue miss revives AI capex jitters." CAT −9.5%, LLY +6.3% (Reuters via Investing.com, 2026-10-29). Closes (Yahoo): SPY −0.37%, QQQ −1.42%, SMH −1.06%, IGV −3.49%, VIX 16.51.
• GOOGL −12.64%, its worst post-earnings day: Q3 EPS $2.12 vs $3.01 expected, revenue $127.58B vs $127.93B, search +12% (a miss), cloud $30.71B. FCF was −$9.07B vs −$2.70B expected. Capex guidance of $195–205B was maintained, and 2027 spend is set to rise (CNBC, 2026-10-29; Yahoo Finance, 2026-10-29). MSFT capex plus finance leases was $45.2B, below the $50.67B estimate, and FCF fell 27% (same CNBC link). MSFT −0.16%.
• MRVL −13.24%: Bloomberg reported advanced talks to buy Lumentum in a likely all-stock deal of about $83–85B, on top of the Alphabet read-through (Investing.com, 2026-10-29; Bloomberg Law).
• Software: FTNT and NOW each fell >20%; "fail to meet lofty investor expectations" (Reuters via StreetInsider, 2026-10-29).
• AI-power peers today: PWR beat and raised (rev $12.01B vs ~$10.92B; backlog $63.7B, +62.6%) and was +6.2% midday (Investing.com, 2026-10-29). TT beat but reaffirmed FY EPS of $15.20–15.30, below consensus. Margin missed, but backlog was $12.5B (+73%) and its 1,700 mega-projects are >90% data centers. TT was −4.8% and as low as −11% (Investing.com, 2026-10-29).

4. Company prints, guidance and filings

VRT, Q3 2026 (Wed 10-28 BMO)

• Sales $4,313M (+61%; 55% organic). Adj. EPS $2.21 (+78%). Adj. operating margin 25.2%, adj. FCF $652M. Organic orders >100% y/y, and Vertiv will stop disclosing orders going forward (Vertiv IR, 2026-10-28; SEC Ex. 99.1).
• FY26 guidance raised: sales $14.45–14.85B (+$650M at the midpoint), organic 37–39%, adj. EPS $7.10–7.20, adj. FCF $2.6–2.8B. Q4 sales guide $4.313–4.513B, adj. EPS $2.22–2.28 (same source).
• Consensus was ~$1.82 EPS and ~$3.75B revenue (Investing.com, 2026-10-28). Call transcript: Motley Fool, 2026-10-28.
• Reaction: Oct 28 +16.48% (gap +16.51%). Oct 29 opened 237.96 (−2.05%) and closed 245.95 (+1.24%; open→close +3.36%). After hours ~4:12 PM: 236.45 (−3.9%), in sympathy with INTC/AMZN.
• Base rate: the last VRT print (07-29) went −17.26% with a −24.6% day-2-to-5 drift (TRAI events). This print was a beat-and-raise, the opposite case.

ORCL — Investor Day (Wed 10-28)

• FY30 targets: revenue $225B+ (>35% CAGR from FY26), non-GAAP operating margin 40%+, EPS $21+. Positive FCF from 2H FY28 and $30B+ FCF in FY29. Additional capital raises capped at $20B through FY30 ($10B equity, $10B debt) beyond the $20B ATM done in Q1 FY27. FY27 targets reaffirmed: revenue ≥$90B, EPS $8.10 (Oracle IR, 2026-10-28).
• Context: ORCL had lost about a third of its value YTD, has ~$117B of long-term debt and is rated BBB- by S&P (CNBC, 2026-10-28). Reuters cites 5y CDS at 287.2bp, notes payable $125.34B and cash $36.37B at Aug 31. AI-cloud peers CRWV, Nebius, APLD and IREN rose 2.1–5% on the event (Reuters via Investing.com, 2026-10-28).
• Reaction: Oct 28 +10.73%. Oct 29 opened 144.13 and closed 139.56 (−3.45%; open→close −3.17%). After hours ~4:12 PM: −3.05%.
• Next earnings: 2026-12-14 (TRAI events; date not confirmed by Oracle in my sources).

STX, fiscal Q1 2027 (Tue 10-27 AMC)

• Revenue $4,233M, non-GAAP GM 57.8%, non-GAAP EPS $7.68 (estimates $7.39 and $4.13B), FCF $1.5B, dividend $0.76. FQ2 guide: revenue $4.8B ±$100M and EPS $9.20 ±$0.20, above estimates (Seagate IR, 2026-10-27; SEC exhibit).
• The stock fell 5.7% early on Oct 28 on MS and Susquehanna price-target cuts and slower HDD capacity-shipment growth (StreetInsider, 2026-10-28). It then closed +4.51% (TRAI: gap −3.15%, day-1 +7.9%), a classic gap-fill.
• Oct 29: +0.36% (open→close +1.67%). After hours: −2.0%. Peers WDC −2.7% and SNDK −1.8% after hours.

SOFI, Q3 2026 (Tue 10-27 BMO)

• Adj. net revenue $1.4B (+44%), adj. EBITDA $469.8M, members +1.2M to 17.1M. FY26 guide raised to $5.04B revenue, $1.66B EBITDA, $840M net income, $0.61 EPS. First 2027 guide: $6.4B revenue, $2.24B EBITDA (SoFi IR, 2026-10-27).
• Reaction (TRAI events): +13.42% on Oct 27, +4.51% on Oct 28, and +2.74% today (open→close +4.20%). After hours: −2.1%.

NOW, Q3 2026 (Wed 10-28 AMC)

• Subscription revenue $3,988M (+21%), total revenue $4,106M, cRPO $13.65B (+20.5%), RPO $29.7B. Adj. EPS $1.00 vs $1.02 expected. Q4 guide: subscription revenue $4,245–4,255M (22.5–23%), cRPO growth 18.5% (vs ~19% expected per Bloomberg), operating margin 32%. FY FCF margin 36% (+100bp). AI ACV target $1.7B (ServiceNow newsroom, 2026-10-28; Bloomberg, 2026-10-28).
• Reaction: −26.97%, its worst day ever, with at least 12 price-target cuts (CNBC, 2026-10-29). The stock opened 120.90 (gap −19.47%) and closed 109.65, so it fell another −9.3% open→close. After hours: −1.6%.
• TRAI data gap: /api/events for NOW has no 10-28 AMC event. History ends 2026-07-22, and "next" already reads 2027-01-27 (estimate). The print happened but the calendar missed it, the same failure mode as S-061.

META — financing overhang, not earnings

• META reported Q3 on 10-21 (TRAI events; rev $64.6B +26%, 2026 capex $135–145B, FCF negative for the first time in more than a decade) (CNBC, 2026-10-29).
• This week's pressure is a financing package of up to ~$97.5B: a $25B Class A offering priced at $625 (below the prior close), a $12.5B mandatory convertible preferred, and a planned $60B ATM in new non-voting Class C shares from early 2027. Analysts estimate 5–6% dilution, and Q3 capex was $34.91B (Investing.com, 2026-10-27).
• Oct 28 −1.24%; Oct 29 +1.96% (open→close +0.62%, so the short lost). After hours: −1.1%.

After the close tonight (4:01 PM releases)

• INTC Q3: revenue $16.5B (+21%), non-GAAP GM 44.1%, non-GAAP EPS $0.52 (consensus ~$0.40–0.41). DCAI $7.3B (+78%), client (CCPG) $8.1B (−5%), Foundry operating loss −$1.85B. Q4 guide: revenue $15.0–16.0B (vs ~$17.1B consensus per the MarketWatch preview), non-GAAP GM 38.0%, EPS $0.25. The CFO sees a PC slowdown through at least 1H27, and a $23B equity offering is complete (Intel IR, 2026-10-29; MarketWatch preview). INTC had already fallen −3.6% in the session on Musk Terafab/TSMC talks; Wedbush says TSMC may raise CoWoS prices 10–20% (StreetInsider, 2026-10-29). After hours ~4:12 PM: −18.1% (Yahoo).
• AMZN Q3: sales $201.9B (+12%), EPS $1.89 vs $1.96 estimated, revenue just under the $202.31B estimate (CNBC). AWS $46.0B, +39.3% (fastest in ~5 years; $184B run rate), AWS operating income $17.8B. TTM FCF −$42.4B. Q4 guide: sales $240–247B, operating income $28.5–33.5B. 2026 cash capex ~$220B maintained. Jassy projects AWS growth >40% in 2027 and 2028 (Amazon release via StockTitan, 2026-10-29; CNBC, 2026-10-29). After hours: −7.6% at ~4:12 PM on Yahoo; CNBC reported −9%. Earlier quotes differed, so the reaction is still moving.
• SNDK: reports tonight, call 4:30 PM ET. Prior guidance was revenue $10.30–10.80B, EPS $44–46, GM 83–85% (INDmoney preview). A search summary claimed revenue of $11.0B, but I could not source it, so the actual results are unverified as of writing. After hours: −1.8% (pre-call).
• Also tonight per the eOption weekly calendar: DLR (data-center REIT, relevant to VRT/ETN), FSLR, RDDT, plus MSTR, RBLX, PINS, SBUX, RIVN, MDLZ, COIN and GILD per TRAI events.

5. Tape and TRAI picture

How the 8:53 AM candidates did today (open→close proxy, gross; TRAI's actual entry and exit rules may differ):

• TRAI's indicated gaps were directionally right for NOW (−18.66% indicated vs −19.47% actual) and roughly right for VRT (−3.68% vs −2.05%). The big gaps it flagged at 8:53 (MRVL −11.95, GOOGL −11.31, CAT −11.27) matched the large down days.
• Watch list today (close-to-close): ETN +0.19%, ANET −1.32%, WDC −0.25%, MU −4.98%, DELL −3.69%, SMCI −3.85%, NVDA −0.80%, GEV +1.15%, CRWV −4.99% (Yahoo). Power and electrical (ETN, GEV, PWR) held up. Chips, memory and servers sold off.
• Correlation map for Friday gaps (TRAI 60d ρ to 10-22, ×100):
• INTC: MU 71, MRVL 68, SMCI 65, ETN 60, STX 58, ORCL 52, VRT 50.
• AMZN: GOOGL 72, ETN 51, ANET 41, ORCL 40.
• VRT: ETN 81, GEV 73, ANET 67, NVDA 61, ORCL 50, STX 47.
• STX–WDC 83 (cluster 19), STX–MU 68. NOW–CRM 75 (cluster 6).
• NOW vs hardware is negative (−14 to −35), and META is a singleton (cluster 31).

6. Analysis

1. AI-infra demand is not the problem; financing is. None of the four big spenders cut capex: GOOGL $195–205B maintained and rising in 2027, AMZN ~$220B maintained, META $135–145B, MSFT spending below estimates but not cut. That supports VRT's >100% organic orders, STX's FQ2 guide and PWR's backlog. The market is instead punishing how the spend is funded and what it does to FCF:
2. GOOGL FCF −$9.07B.
3. AMZN TTM FCF −$42.4B.
4. META's ~$97.5B raise.
5. ORCL's capped $20B raise plus BBB- rating.
6. INTC's $23B equity raise.

Suppliers with their own beat-and-raise (VRT, PWR, STX) held up today. Names whose story depends on someone else's balance sheet (ORCL, CRWV, MRVL) did not.

1. Momentum after a +10% day fades the next session (my calc): I tested 17 AI-infra names (VRT, ORCL, STX, WDC, MU, DELL, SMCI, ANET, ETN, NVDA, MRVL, AVGO, GEV, PWR, CIEN, COHR, AMD) on Yahoo daily data from 2016 to 10-29.
2. After a ≥+10% close-to-close day, the next session's open→close averaged −0.27% (n 322, 48% up, t −1.02).
3. Since 2024 it is −1.00% (n 161, 40% up, t −2.68), with the average next-day gap +0.35%.
4. Day 2 open→close recovers (+0.47%).

This lines up with TRAI's replay. long_momentum signals on an earnings-reaction day averaged −0.56% net (n 48, hit 44%, t −1.19), vs +0.13% for non-event signals (n 337). Today split the bet: ORCL fit the base rate (−3.17%), while VRT (+3.36%) and STX (+1.67%) beat it.

1. Shorting software after a crash: TRAI's replay says no.
2. short_momentum overall: −0.54% net (n 412, t −4.31).
3. On reaction days: −0.61% (n 41).
4. NOW post-earnings shorts specifically lost twice (04-23 signal −5.12%, 07-23 signal −4.07%).

Today NOW was a +9.3% open→close winner for the short, but that is one observation against the replay. My Yahoo study of software/megacap names after a ≤−15% day (19 names, 2016+) shows the next open→close at +0.28% (n 46, t +0.40) and +0.50% since 2024 (n 19). That is no edge either way. Do not read today's NOW win as evidence for a day-2 short.

1. Hardware-minus-software spread (whiggity's idea, tested). I defined it as avg(VRT, ORCL, STX) − avg(NOW, META), daily. It was +12.70% on 10-28 (z 3.72 vs the prior 60 days) and +11.88% on 10-29 (z 3.10). Past >2.5σ days since 2022 (n 22) were followed by a next-3-day spread of +1.01% on average, positive 55% of the time. That is no evidence of mean reversion (n is small and windows overlap).
2. Friday gap risk is concentrated in TRAI's cluster 1. INTC at −18% after hours (ρ 0.50–0.71 to MU, SMCI, MRVL, ETN, STX, ORCL, VRT) and AMZN at −7 to −9% (ρ 0.40–0.51 to ETN, ANET, ORCL) both hit the same 24-name cluster. VRT, DELL, MU and ORCL are already −2.9% to −3.9% after hours. The INTC miss is about PCs and margins (client −5%, Q4 GM 38%) while DCAI grew +78%, so the read-through to data-center power and storage is weaker than the after-hours prices imply. AMZN's AWS +39% with capex maintained is positive for VRT/ETN demand. The tape is trading the factor, not the fundamentals.

7. Actionable for Fri 10-30

ACTIONABLE: YES — caution and risk-sizing, not a new trade idea.

• Treat VRT, ORCL, STX, DELL, MU, SMCI, ANET and ETN as one correlated bet. If TRAI's postclose/premarket run keeps several of them as long_momentum, expect a cluster-wide gap-down from INTC/AMZN. Size them as a single position (S-001 cluster cap; S-004 peer gap flags) and lean on the 9:25 indicated gap (S-067).
• Day-after-+10% longs: ORCL and VRT were both day-2 longs today. Any new long_momentum name coming off a ≥+10% day has a negative next-session base rate since 2024 (§6.2). Prefer waiting for opening-range confirmation over buying at the open.
• NOW: no edge in a day-3 short or a fade. Replay says post-earnings software shorts lose, and the Yahoo base rate is flat.
• Friday catalysts: Q3 ECI at 8:30, Chicago PMI at 9:45. BMO: XOM, CVX, ABBV, NVT (nVent, a VRT liquid-cooling/electrical peer, so direct VRT/ETN read-through), CCJ, LYB. It is month-end, and the Nov 3 midterms come next week (eOption). ANET, SMCI and ETN report Nov 3; AAPL Nov 2 AMC per TRAI events.

8. Strategy implications (proposals, see strategy.md)

• S-066: Day-after-big-up guard for long_momentum (+10% day or earnings-reaction day): opening-range confirmation or half size. Extends S-005.
• S-067: After-hours peer-print gap check for cluster-1 longs (INTC/AMZN-type prints), done at 9:25 before entry.
• S-068: Freshness guard for stale API layers (scorecard/correlations/regime/macro at 10-22), plus the NOW 10-28 missed print (extends S-061).
• S-069 (NEEDS DATA): Hyperscaler "funding stress" event map. Does a spender's FCF/financing surprise predict next-day returns for cluster-1 suppliers?
• S-070 (NEEDS DATA, from whiggity): CLV-conditioned continuation, VRT→peer lead-lag, and the hardware-minus-software spread.

9. Unknowns and open questions

• SNDK's actual results and guidance, and the final after-hours levels for INTC and AMZN. The ~4:12 PM quotes are preliminary, and sources disagree on AMZN (−7.6% Yahoo vs −9% CNBC).
• ORCL's next earnings date (TRAI says 12-14; unconfirmed). Whether VRT's decision to stop disclosing orders will weigh on its multiple.
• Why /api/scorecard, /api/correlations, /api/regime and /api/macro have not refreshed since 10-22.
• TRAI's actual fills on today's candidates. The table in §5 is an open→close proxy, not TRAI's P&L.
• DLR's results tonight (data-center REIT, read-through to VRT/ETN): not checked.

10. Research partner (whiggity) consult

_Source: whiggity (Monica custom bot), reply received 16:09 ET. Each point was checked against public sources and TRAI data. Not independently authoritative._

Verified:

• VRT Q3 adj. EPS $2.21 vs $1.82 consensus and revenue $4.31B vs ~$3.75B (Investing.com, 2026-10-28).
• Its guesses for the other catalysts were correct: ORCL's investor/analyst day (Oracle IR, 2026-10-28), STX fiscal Q1 earnings (Seagate IR) and SOFI Q3 (SoFi IR).
• AMZN and INTC report tonight and are the main gap risk for AI-infra names (confirmed: both released at 4:01 PM, §4).
• Friday is the last trading day of October.
• Sympathy links supported by TRAI 60d ρ: VRT–ETN 0.81, VRT–GEV 0.73, STX–WDC 0.83, STX–MU 0.68, NOW–CRM 0.75, ORCL–DELL 0.62, SOFI with AFRM (cluster 11). NVT reports Fri BMO (eOption).

Conflicts (whiggity was wrong or out of date):

• It treated VRT +16.5% and ORCL +10.7% as today's moves. They were Oct 28. Today VRT was +1.24% and ORCL −3.45%.
• It said META and NOW both reported Oct 28. NOW did; META reported Oct 21. META's weakness is the ~$97.5B financing and capex (§4).
• It called NOW's −3% a muted "controlled pullback." NOW fell −26.97% today, its worst day ever (CNBC).
• It said PCE comes Fri 10-30 at 8:30. PCE was released today, 10-29 (CNBC). Friday has the ECI.
• It said AAPL reports tonight. TRAI events show AAPL on 11-02 AMC, and the eOption calendar lists no AAPL on 10-29.
• It claimed post-earnings drift "dominates 3–5 days" for big infra breakouts. My base rate shows the next day fading (−1.00% open→close since 2024, §6.2), with only a small day-2 bounce.
• It linked META to GOOGL/PINS/TTD as co-movers. TRAI shows META as a singleton (cluster 31; GOOGL ρ only 0.35).
• Its footnote links did not match the claims they were attached to.

Unverified, carried as open questions:

• (a) Do +10% Thursday movers that close near the high (CLV > 0.85) continue on Friday? (S-070)
• (b) Do VRT's big days lead MOD/NVT/ETN by t+1 to t+5? MOD and NVT are not in the TRAI universe. (S-070)
• (c) Does the hardware-minus-software spread mean-revert after >2.5σ? First pass says no (§6.4). (S-070)
• (d) Does SOFI fade after +4% days, conditioned on the 10y yield? Unconditioned 3-year check: next open→close −0.15% (n 94, t −0.49), next close→close +0.60%. No clear fade; yield conditioning not tested.
• (e) Its claim that window dressing / mutual-fund fiscal year-end on Oct 31 brings a Nov 2 reversal: no source.
• (f) "G7 emergency oil reserve release," cited via Wikipedia: not verified, not used.