What it will take for us to buy the dip in DuPont. Plus, profit-taking hits cyber stocks

Stocks were lower as Treasury selling pushed the 10-year yield back to the psychologically important 5% level, while AI-related industrial and semiconductor names generally outperformed. DuPont has fallen about 13% since Aug. 5 amid pressure from higher oil prices and rates, though Wells Fargo maintained an overweight rating and $166 target, viewing Middle East exposure—about 4% of sales—as manageable. Costco, down more than 18% from its May 19 record high, reports next Thursday; key upcoming macro releases include PMIs, jobless claims, new-home sales, durable-goods orders and consumer inflation expectations.

Analysis

A 5% 10-year yield is a more consequential discriminator within AI than a directional signal for the group. ETN and GEV retain relatively near-term backlog conversion and grid-capex scarcity value, while long-duration semiconductor multiples—especially NVDA and AVGO—become vulnerable if real yields continue higher without a matching upward revision to hyperscaler capex. The second-order pressure is on rate-sensitive end markets: KBH faces both mortgage-payment affordability and a higher hurdle rate for land inventory, while industrial valuation compression can persist even if underlying orders remain intact.

The semiconductor strength should not be read as a clean all-clear. MU and AVGO have identifiable earnings-power upside if memory pricing and custom silicon demand remain firm, but INTC's recent momentum now requires concrete evidence of execution rather than incremental narrative; a reversal in peer-led risk appetite could be sharper there. Cybersecurity profit-taking is potentially more interesting than it appears: PANW and CRWD have comparatively recurring revenue and AI-security demand, but their premium multiples still need billings/remaining-performance-obligation acceleration to offset the rate headwind over the next 1-3 months.

For consumer names, COST's pre-earnings setup is primarily a valuation and expectations problem, not a question of operational quality. A routine membership-led beat without an acceleration in traffic, renewal economics, or margin may fail to rerate the shares; conversely, a post-print gap down with stable comparable-sales and membership metrics would create a better entry than positioning ahead of the release. The key near-term falsifier for the broader risk thesis is a sustained 10-year yield above 5% following PMI, labor, and inflation-expectations data; a retreat below that threshold on softer growth data would favor re-expansion in quality growth and cybersecurity.

Key Decisions for Investors

• Initiate a 1-3 month pair: long GEV / short a beta-adjusted basket of KBH and XHB. Grid demand is less mortgage-rate elastic than residential construction; exit if the 10-year yield falls sustainably below 4.75% or GEV order/backlog commentary weakens.
• Maintain ETN exposure but do not add broadly to AI semiconductors while the 10-year yield is above 5%. Prefer MU over INTC for incremental semiconductor risk; reassess after the next memory-pricing and gross-margin update, with a stop if industry pricing rolls over.
• Use further 5-8% weakness in PANW or CRWD to build a staged 6-12 month long only if billings/RPO guidance remains intact. Avoid chasing a one-day rebound; the thesis is invalidated by material forward-growth deceleration, not by sector volatility alone.
• Do not position long COST into earnings absent evidence of revised consensus estimates or supportive options skew. Place an alert for a post-results decline of 8%+ with unchanged membership and comparable-sales trends; that would offer a more favorable risk/reward entry for a 6-12 month holding.
• Hedge the growth sleeve with short QQQ or reduced NVDA/AVGO gross exposure if the 10-year closes above 5% for multiple sessions after macro data. Cover the hedge on a yield reversal below 4.75% or clear evidence that rising yields reflect stronger nominal growth rather than inflation risk.