yifeng.ai
← All days
Market · No. 10

Mon, Oct 12, 2026

Daily summary

10/12 (Columbus Day, bond market closed): Equities were mixed amid rotation, deleveraging across AI/semiconductors, and resilient large caps.

Semiconductor index roughly -2.7% (pre-earnings profit-taking, with the memory supply chain weighed down by Samsung's results and CXMT competition); META gained over 3% (Muse Wonder launch); NVDA fell for a 5th straight day.

This week's catalysts: Wednesday CPI + ASML earnings + Anthropic investor day; Thursday TSMC earnings + PPI + retail sales; Friday monthly options expiration.

Key levels: SPX >7820 targets 7880-7900; <7780 risks a retest of 7735-7700; extreme downside at 7640-7650.

Sources broadly agree on a near-term wait-and-see stance ahead of data and earnings; research notes place greater emphasis on AI valuations' sensitivity to how private labs report ARR.

Highlights
Mixed

AI cools as capital rotates within equities; CPI and earnings ahead

  • The three major indexes closed mixed on October 12, broadly steady despite pronounced internal rotation and widespread weakness across AI-related sectors
  • Macro: The Cleveland Fed president struck a hawkish tone, citing broadening inflation pressures and insufficient policy restraint, favoring small, frequent rate hikes; Wednesday's CPI is key, with an upside core CPI surprise likely to pressure growth stocks and shift rate expectations higher; Middle East tensions worsened, Russia-Ukraine ceasefire reports remained unconfirmed, and oil prices were volatile
  • Sectors: The Philadelphia Semiconductor Index fell about -2.7% on pre-earnings profit-taking; memory chips weakened on Samsung's earnings miss and concerns about CXMT competition; optical communications stocks fell broadly; META rose over 3% as institutions sought safety; NVDA fell for 5 consecutive days, with support at the 50-day moving average in focus
  • Earnings: Major banks report first this week, but the main event is the concentration of Magnificent Seven results in the week of October 26; S&P 500 third-quarter earnings growth is expected to reach over 30% year over year
  • View: Stay on the sidelines near term ahead of data and earnings; capital is concentrating in large-cap tech with strong cash flow
NVDAMETA
Mixed

Monday morning: Semiconductors lead declines at -1%, software strengthens; focus on Anthropic investor day

  • QQQ -25bps, SPY +2bps; Treasury market closed; oil +1% following a Houthi attack on Riyadh airport; last weekend's rotation continued, with semiconductors -1% and software +50bps
  • META +1.3% led gains; SPCX +1% (Musk said Starship's 15th flight is coming in the next few months); CRM +4% (BofA upgraded to Buy/$285 from UW/$160); NFLX +1%
  • Asia: KOSPI -3.65%, Samsung/SK Hynix -4.5%, memory/HDD names followed with declines of 2-3%; Largan hit limit-up on positive CPO project commentary
  • This week's calendar: ASML and TSM earnings, OCP and Semicon conferences; Anthropic investor day Wednesday, with market expectations for end-September ARR around $78-80B
  • Ratings: Wells downgraded ADBE to UW/$225; Barclays downgraded HUBS and MNDY to Neutral; Wells upgraded TYL to OW/$400; Evercore added WDC and LITE to TAP; coverage initiated on WULF (TD Cowen Buy/$21) and HOOD (Evercore OP/$130)
METASPCXCRMASMLTSMADBEWDCLITE
Mixed

Meta launches Muse Wonder; semiconductors lead declines at 2.7%; OpenAI reporting change fuels AI concerns

  • QQQ -70bp in light holiday trading; megacaps held up (META +3.4%, AAPL +1.3%, MSFT +1%), while deleveraging continued in AI capex/financing-linked names; semiconductors -2.7%
  • META launched Muse Wonder: second only to Claude Opus 5.5 on the AA Intelligence Index; SWE-Marathon 57.5% vs. Opus 5.5's 46.3%; pricing of $4/$20 per million tokens; the strongest version will be sold directly to enterprises and remain closed-source
  • Zuckerberg said work on the next iteration is already underway and spending will not slow; Prometheus cluster in Ohio >1GW, Hyperion in Louisiana to reach 5GW; the author estimates costs at approximately $15 billion/GW
  • Thursday's change in OpenAI's annualized revenue reporting from 70 billion to 50 billion (gross to net) did not alter the broader trend, but highlighted the AI trade's heavy dependence on ARR from two opaque private labs; investors fear an Anthropic ARR miss could trigger a selloff
  • Bearish interpretation: frontier models are shifting from a duopoly to an oligopoly, weakening frontier labs' fundraising case; Muse Wonder delivers a double blow to Anthropic through stronger competition and lower spending by a major customer
METAMSFTAAPLNVDAAMDSOXXQQQ
Neutral

The Information: Lamanna turns to Anthropic models to fix Copilot; enterprise users exceed 30 million

  • After customer complaints about Copilot in summer 2025, Charles Lamanna proposed paying for Anthropic models to Nadella; commercial chief Althoff questioned the cost, as OpenAI models were nearly free for Microsoft, but demonstrations showing Anthropic's stronger Excel math and PowerPoint performance won support
  • Copilot enterprise users have more than doubled since the start of the year to over 30 million; customers including AT&T report an improved experience
  • Lamanna was promoted to EVP overseeing Copilot engineering in March and takes charge of Office this month (Rajesh Jha retires in July); he also oversees Teams, leading an organization of roughly 25,000 people, and is viewed as a potential successor to Nadella
  • The current priority is enabling Copilot to select the right mix of models for different tasks to reduce costs; Lamanna says Microsoft should become the "Costco of AI"
  • Anthropic and OpenAI are both key Microsoft partners and fierce competitors; enterprise customers are scrutinizing AI budgets
MSFT
Neutral

Morgan Stanley CIO survey: Around 45% of companies use open-weight models, rising to 90% in 3 years

  • Morgan Stanley's 3Q26 CIO survey: Around 45% of respondents are deploying open-weight models, expected to reach around 90% in 3 years; these models currently account for 8% of AI workloads, rising to 19% in 3 years
  • Adoption is consistent across company sizes at around 40-50%; large companies with revenue above $15 billion are expected to have the highest adoption in 3 years at 95%
  • Adoption is higher in healthcare, manufacturing, finance, and energy, driven by regulation, data sovereignty, and low-latency edge requirements; retail and technology are expected to see the largest increases
  • The bank believes U.S. companies are willing to deploy Chinese models on U.S.-based infrastructure, though geopolitical uncertainty persists; security, data sovereignty, and lower token costs are driving on-premises deployment
  • The bank proposes a stock-selection framework spanning 'three worlds'—closed-source wins, hybrid, and open-source wins—covering GOOGL, AMZN, MSFT, NVDA, PANW, DELL, and others
GOOGLAMZNMSFTNVDAPANWDELLHPEMETA
← Sun, Oct 11, 2026Tue, Oct 13, 2026 →