Morning Bid: Bond bruise won’t heal

The 10-year U.S. Treasury yield rose above 5.3%, reaching a 24-year high, with its quarterly increase the largest since 1994 despite softer-than-expected August inflation data. Persistent long-end bond pressure pushed stocks mostly lower, while Micron's AI-driven earnings beat—following a near-fourfold share-price gain this year—provided limited support to technology stocks. France's 10-year spread over Germany widened above 120bps for the first time in 14 years, while foreign investors withdrew ¥4.6 trillion ($29.2 billion) from Japanese debt securities in the week through September 26.

Analysis

MU is now a two-factor trade: HBM/DRAM earnings revisions versus duration-multiple compression from rising real yields. The near-term fundamental signal is constructive if management’s implied HBM mix, pricing and gross-margin trajectory exceed already elevated sell-side assumptions; memory’s operating leverage means modest upside to bit shipments or ASPs can create disproportionate EPS revisions. The more non-obvious beneficiary is equipment demand: sustained HBM capacity additions should support deposition/inspection intensity at LRCX, KLAC and AMAT, although MU’s own margin expansion requires disciplined industry supply rather than simply stronger AI end-demand.

For the next 1-3 months, macro can dominate the earnings setup: a further real-yield spike would compress high-multiple AI hardware regardless of improving estimates, particularly after MU’s sharp rerating. Conversely, evidence of labor-market cooling and lower term premium could unlock a catch-up move because MU has more direct exposure to AI-memory scarcity than broad semiconductor indices. The structural 6-18 month risk is that SK Hynix and Samsung close qualification and yield gaps faster than expected, converting a supply-constrained HBM market into a pricing competition before new fabs earn adequate returns.

Consensus appears too focused on headline AI demand and insufficiently on the durability of pricing after the next capacity wave. A positive quarter alone is not enough: the thesis requires management to demonstrate that HBM allocation is displacing lower-return commodity output while conventional DRAM/NAND pricing remains rational. Falsification points are sequential gross-margin guidance below consensus, commentary indicating HBM customer inventory normalization, or a sustained rise in 10-year real yields that prevents estimate upgrades from translating into equity returns.

Key Decisions for Investors

• Initiate a 1-3 month long MU / short SOXX pair on a 1:1 beta-adjusted basis, preferably after any rates-driven weakness rather than chasing an overnight reaction. This isolates MU-specific HBM estimate upside; exit if next-quarter gross-margin guidance fails to exceed consensus or MU underperforms SOXX by 10%.
• Add selective longs in KLAC and LRCX on semiconductor-equipment pullbacks for a 6-12 month HBM capacity-build thesis. Prefer KLAC for process-control exposure; reduce if memory makers signal capex discipline tightening rather than node-transition investment.
• Do not add unhedged AI-semiconductor beta ahead of payrolls and Fed communication. Use a break lower in long-end yields as confirmation; if the 10-year yield continues to reprice materially higher, maintain the MU/SOXX hedge or use SMH puts to protect the long leg.
• Monitor MU’s HBM revenue mix, customer qualification commentary, conventional DRAM/NAND pricing, and capex plan at the next earnings update. Any indication that supply additions are arriving ahead of qualified demand should convert the tactical long into a neutral/short relative view versus diversified semiconductor exposure.