Our top 10 things to watch in the stock market Wednesday

U.S. equity futures pointed lower as the 2-year Treasury yield reached 4.794%, its highest since July 2024, while the Nasdaq came off two consecutive record closes. Microsoft was upgraded to buy at Stifel with a $575 target, while Citi lifted Micron's target to $1,300 on stronger expected DRAM pricing; CoreWeave received a UBS buy initiation despite leverage and competition risks. General Mills beat low fiscal Q1 expectations but reported adjusted EPS down 13% year over year and warned input-cost inflation would reach the high end of its 4%-5% range, while McDonald's announced an $8.5B decade-long franchisee investment plan.

Analysis

AI infrastructure remains the cleaner expression of enterprise adoption than application-layer seat counts. Discounting enterprise copilots can accelerate user penetration but delays ARPU realization, making MSFT's next two quarters more dependent on Azure consumption and gross-margin discipline than on reported paid-seat growth; AMZN has comparatively greater upside if incremental AI workloads improve AWS utilization without equivalent pricing concessions. The near-term risk is that investors treat adoption metrics as monetization, creating a valuation air pocket if Q4 guidance does not show accelerating AI revenue per customer.

MU is the highest-beta earnings catalyst: the relevant variable is not spot-memory optimism but whether contracted pricing and supply commitments extend visibility beyond the next pricing cycle. A strong long-term-agreement update would support estimates through the next 6-12 months and tighten supply available to smaller memory buyers; absent that, elevated expectations leave the stock exposed to a guide-versus-whisper miss. CRWV is a less attractive way to express the same AI-capex theme: its equity duration, customer concentration, and refinancing sensitivity can dominate any utilization upside while front-end yields remain elevated.

The financials selloff on AI disintermediation is likely too indiscriminate. Personal agents may pressure low-value brokerage cash balances and order-routing economics, but exchanges and data businesses such as NDAQ monetize market complexity, indexing, and institutional workflow rather than consumer cash sorting. Conversely, GS is more exposed to capital-markets activity and curve-sensitive financing conditions than to retail-agent disruption; a sustained restrictive front end would matter more than the technology narrative over the next 1-3 months. In consumer staples and restaurants, cost pressure is more likely to separate operators with pricing power from those funding value promotions, limiting the read-through from one company's maintained guidance.

Key Decisions for Investors

• Initiate a 3-6 month long AMZN / short CRWV pair, sized beta-neutral. This isolates hyperscaler AI-margin leverage from neocloud financing and concentration risk; target 15-20% relative return. Exit if AWS growth decelerates or CRWV demonstrates contracted backlog growth sufficient to cover its next major funding needs.
• Maintain MSFT but do not add solely on reported Copilot seats; add only if the next results show Azure growth acceleration or disclosed Copilot ARPU/attach-rate improvement. Hedge event risk with 2-3 month downside puts if the position is above benchmark weight; discounting without revenue conversion is the thesis break.
• Use MU earnings as a conditional catalyst rather than pre-positioning aggressively: buy only on confirmation of multi-quarter contract-price visibility and disciplined capex, or own a limited-risk call spread into results. A revenue guide driven only by spot-price assumptions, or weak long-term supply commitments, would be a signal to fade the post-earnings move.
• Rotate part of any broad financial underweight into long NDAQ versus short GS over 3-6 months. NDAQ's recurring data/index revenues offer a better hedge against consumer-brokerage disruption and a flatter curve; close the spread if capital-markets activity reaccelerates materially or NDAQ's net-new recurring revenue slows.
• Avoid treating META's developer event as a durable earnings catalyst until management quantifies distribution costs, inference expense, and a credible monetization path. Event-driven upside is possible, but a 1-2 month post-event fade is likely if product announcements lack advertising conversion or paid-agent economics.