Market Brief(X) — Oct 26–Oct 29, 2026

2026-10-29 • Twitter

Executive Summary

Anthropic’s S-1 restarted the AI trade on Monday, and the cohort spent the rest of the window testing whether the hyperscalers could pay for it. The filing showed Q2’26 revenue of $11.6B, a first GAAP operating profit and about $132B of cash, which killed the “labs run out of money” story (@FundaAI). Wednesday then split the field: Azure accelerated to 48% with positive free cash flow, while Alphabet fell about 9% after hours on an ads miss and a negative $9.1B FCF despite 103% cloud growth (@qinbafrank). Warsh’s hawkish hold let the long end do the tightening, with the 10Y closing near 5.05% (@laochenusa). The tension to hold is that AI demand looks stronger than ever, yet the market is now paying only for companies whose cash returns keep pace with their spending, and Oracle just put the first “capex peak” date on the calendar.

Key Themes & Trends

Anthropic’s S-1 Moves the AI Debate From Demand to Supply

The filing was the window’s dominant event, with 9 distinct voices across all four profile types, including 3 VIPs. The key disclosures were Q3’26 revenue backed out at about $16.9B, H1’26 implied gross margin of 85%, NRR near 900%, and 5–7% of revenue coming from customers who violated usage terms through distillation (@FundaAI). TJ_Research (VIP) highlighted the speed of the path to GAAP profit (@TJ_Research). NullableX (VIP) read the bounce as proof that the market is trading “earnings expectations, earnings expectations, earnings expectations” (@NullableX). ShanghaoJin argued that slower ARR growth reflects a shortage of compute, not of demand, and that profitable labs holding about $200B after the IPO will not cut back on spending (@ShanghaoJin). The bearish counterpoint is about liquidity. With the roadshow possible from Nov 9, labubu_trader and qinbafrank both expect long-only funds to sell Mag7 shares to free up cash for IPO subscriptions (@labubu_trader, @qinbafrank).

  • High-signal tickers / exposures: ANT (IPO), AMD, CRWV, Mag7 (as a source of IPO funding)

Cloud Up, Ads Down: Alphabet Splits the Cohort

GCP revenue was $30.7B, up 102.6%, at a 36% operating margin. But Search came in at $63.4B against a $65B estimate, and EPS was hit by roughly $8.7B of unrealized SpaceX losses (@qinbafrank). NullableX (VIP) said Meta’s slowing ad impressions had already “smelled” like weakness for Google (@NullableX). FundaAI went further, arguing that Meta is taking North American ad budget in travel, autos and financial services, a “gray rhino” the market ignored (@FundaAI). The investors disagreed with each other. ShanghaoJin sold a decade-long GOOG position because “Search… will never be the way it was” (@ShanghaoJin). ivanalog_com called the selloff “absolutely wild” given GCP’s lead (@ivanalog_com). Microsoft was the clean winner, with Azure guided to 49–50% constant-currency growth, $18.6B of FCF and a $677B backlog (@qinbafrank).

  • High-signal tickers / exposures: MSFT (long), GOOGL (contested), META (ad share gains)

“Peak Capex” Gets a Date, Then Gets Postponed

At its investor day Oracle guided capex to peak in FY27–28, FCF to turn positive in H2 FY28, and at least $30B of FCF in FY29. The stock rose about 11% (@laochenusa). Franktradinglog called this the first time a cloud provider had talked about a capex peak and a “huge negative” for SOXX (@Franktradinglog). labubu_trader agreed it was bullish for clouds and bearish for semis (@labubu_trader). Microsoft’s apparent capex miss ($39.4B against a $45.9B estimate) looked like a “triple kill” until FundaAI pointed out that Microsoft had moved some finance leases, which count toward capex, into operating leases (@FundaAI). labubu then pushed the scare to “next quarter” but stayed pessimistic about 2028 because of funding and power limits (@labubu_trader). TJ_Research (VIP) noted that Google guided 2027 capex “significantly higher” (@TJ_Research). laochenusa (Medium) put the bear case plainly: if buyers spend less, who buys the chips (@laochenusa)? Support here is broad but split: traders lean bearish on semis, while analysts and investors see a postponement.

  • High-signal tickers / exposures: ORCL, SOXX, VRT, NVDA

Memory’s Supercycle Gets Contract Backing

SK hynix beat consensus by 7% on revenue (KRW 104.6T) and 9% on operating profit (KRW 83.3T, an 80% margin). Long-term supply agreements (LTAs) with 14 customers cover more than 50% of 2027–28 capacity; it says DRAM and HBM are “virtually sold out” through 2029, and plans about KRW 100T of 2026 shareholder returns (@jukan05, @ArtofSpecuycky). Samsung followed by saying the shortage will “worsen in 2027” and that 60–70% of its capacity could be under LTAs through 2030 (@jukan05). qinbafrank concluded that treating 2027 as the profit peak is “too mechanical” and that the next trough should be higher (@qinbafrank). The bearish thread is China. CXMT is starting HBM3E production at 20% yield, but zephyr_z9 estimates its gross margin is already 60–71% and expects yields of 50–70% by Q2 (@zephyr_z9). ShanghaoJin says the state-backed DRAM expansion is so large he “dare not say” how big (@ShanghaoJin). Support spans 2 VIPs, industry analysts and macro voices (jukan05 and zephyr_z9 count as one). The KOSPI jumped 6.67% on Hynix, then Samsung fell 6% intraday on its own record print (@laochenusa).

  • High-signal tickers / exposures: SK hynix, Samsung, MU, KLAC, SNDK

Warsh’s Pause: Bonds Do the Tightening

The Fed held at 3.75–4.00% on a 10–2 vote, with Cook and Hammack voting for a hike. The statement upgraded growth to “strong” and dropped its pledge of a “timelier” return to 2% (@qinbafrank). The curve steepened, with 2s30s wider by 11.2bp (@Corsica267). Franktradinglog drew out a new reaction function: if credit tightens, hikes become less likely, so the Fed is “constrained by financial stability” while inflation is still above target. His conclusion was that this means a weaker dollar and stronger gold and crypto (@Franktradinglog). labubu_trader called it a “conditional Warsh put” (@labubu_trader). Corsica267 was blunter: “why don’t you fxxking hike?” (@Corsica267). RichTerry123 (VIP) warned that the rally may be a “dead-cat bounce” and that the real test comes in December (@RichTerry123). The S&P forward P/E has already compressed to about 18.5x from 23x (@laochenusa).

  • High-signal tickers / exposures: TLT/long duration (short), DXY, gold, 2s30s steepener

AI Swaps Payroll for Compute

Microsoft offering buyouts to 60,000 staff (about 27% of its workforce) became a cross-profile thesis. TJ_Research (VIP) framed it as companies paying for AI with operating expenses and expects the labor-market impact in Apr–Jun 2027 (@TJ_Research). ShanghaoJin said, “you won’t see token demand fall… you’ll see layoffs,” and named bank back offices as next (@ShanghaoJin). ivanalog_com says mass layoffs are the next profit lever for SaaS and one reason he owns software (@ivanalog_com). Software still sold off: NOW fell 18% and FTNT 25% after hours (@laochenusa).

  • High-signal tickers / exposures: MSFT, PLTR, NOW, FTNT, CRWD

Market Sentiment

The cohort went from relieved to conflicted. Monday’s S-1 bounce lifted SOXX almost 4%, but on volume only near its 20-day average, (@laochenusa). The split by profile is sharp. Traders de-risked: Franktradinglog cut most of his book, labubu_trader is still waiting, and LinQingV said not to “click buy before late November” (@LinQingV). Fundamental voices stayed constructive on AI demand, and LeoYuen13 still targets SPX 8,000 after Thanksgiving (@LeoYuen13). By the close of the window, ruth_capital (VIP) described the dispersion as a “slaughterhouse of individual names” (@ruth_capital). Conviction is high on demand and low on the price anyone should pay for it.

Key Figures & Assets

Trading Activity & Holdings (optional)

  • jdhasoptions (VIP): Covered a CRWV short after its weekly chart broke the lower Bollinger band, citing a poor risk/reward for staying short (@jdhasoptions). Went long PLTR for a breakout, having bought on Friday (@jdhasoptions). His long-term account holds only GOOG shares (@jdhasoptions).
  • TJ_Research (VIP): Keeping MSFT; may sell AMZN and GOOG after Amazon reports (@TJ_Research). Still holds more than 10% in AMZN (@TJ_Research) and is waiting for a chance to add TSLA (@TJ_Research).
  • RichTerry123 (VIP): Holds MRVL through the Lumentum headlines (@RichTerry123).
  • Franktradinglog (High): “Nuked” most of his book after FOMC and Oracle. He kept puts on INTU, XLP, BKNG and MS, a long INDA position and small QQQ puts (@Franktradinglog).
  • ShanghaoJin (High): Sold GOOG and has left most of the proceeds uninvested (@ShanghaoJin).
  • labubu_trader (High): Plans to buy semis LEAP calls (INTC, BE, LITE, AMD) on a Thursday dip if NQ and SOXX hold key levels, and targets SOXX at 590–600 by year-end (@labubu_trader).

Off-Theme Highlights (optional)

  • MRVL / LITE: Reports that Marvell is near a deal for Lumentum sent MRVL down about 10% and LITE up about 10% (@ArtofSpecuycky). FundaAI expects heavy dilution, antitrust questions and a lift for COHR (@FundaAI).

Notable Perspectives & Insights

  • ShanghaoJin’s compute futures curve: “Not only is spot compute short, the forward curve is in deep contango,” while equity investors price the opposite. “Buying compute vs. imagining compute, one of them will fall.” In his framework, semis are “oil producers” and neoclouds earn a carry trade (@ShanghaoJin).
  • qinbafrank on retiring ARR: He argues that $50–60B of real annual revenue is “far higher quality” than $100B of annualized ARR, and that once Anthropic and OpenAI list, the market should judge them on reported quarterly revenue (@qinbafrank).
  • TJ_Research’s 2027 playbook (VIP): Cash-cow software; healthcare as the source of alpha; semis, mainly INTC, as the source of beta; and the agent platforms (META, GOOG, AAPL, QCOM), buying “whoever is weakest.” He would add bonds once big-tech yields pass 7% (@TJ_Research).

What to Watch

Follow-up on last brief’s list

  • Israel election: Still pending as a risk. ShanghaoJin was surprised by Netanyahu’s strong vote (@ShanghaoJin), and NullableX noted Trump’s talk of striking Iran after the midterms barely moved the market (@NullableX).
  • FOMC plus MSFT and GOOGL: Mixed. The Fed held with a hawkish tone, MSFT was bullish and GOOGL bearish (@laochenusa).
  • Neocloud credit tape: Improved. Meta’s choice of equity over debt cut CRWV’s CDS by about 110bp (@FundaAI), though ORCL’s CDS is still near 300bp (@laochenusa).
  • Month-end liquidity: Pending. ShanghaoJin says the bleeding has stopped at the source but liquidity hasn’t recovered (@ShanghaoJin).

Next one to two weeks

  • AMZN and INTC earnings, Sep PCE and Q3 GDP (Thu Oct 29): The AWS bar is about 40%. For INTC, Q3 revenue above $16.8B and a Q4 guide midpoint of at least $17.6B would justify upgrades. Misses would reopen the capex-return debate (@qinbafrank).
  • Term premium and month-end rebalancing (Oct 30 to November): This is Corsica267’s key concept. The long end is pricing a funding squeeze, and he expects the Treasury to drain liquidity intermittently in November and sees “something brewing overseas.” A 30Y rising toward 5.5% would be bearish for long-duration equities. A long end that stabilizes after the rebalance would support risk assets (@Corsica267).
  • US midterms (Tue Nov 3): The “golden 8 days” window (Oct 27–Nov 6) has been positive in 18 of 19 cycles since 1950. A break would be a bearish signal (@qinbafrank).
  • Anthropic roadshow (from Nov 9 at the earliest): Watch for Mag7 selling to fund subscriptions. A strong oversubscription could “pump up the animal spirits” (@labubu_trader).
  • Marvell–Lumentum announcement (possibly next week): A confirmed all-stock deal is bearish for MRVL near term and bullish for COHR and the other laser suppliers (@RichTerry123).