CNBC Daily Open: Micron steals Nvidia's margin king crown

Micron reported a record 84.9% gross margin, up from 74.9% last quarter and 39% a year ago, while Qualcomm raised its fiscal 2029 non-handset revenue forecast to $40 billion from $22 billion. SK Hynix filed for a potential $29.4 billion Nasdaq ADR listing, the second-largest U.S. listing on record, and its shares rose as much as 11%. Offsetting the tech strength, oil prices fell below $70 and Europe’s record heatwave caused outages and disruption, while the White House requested $87.6 billion in supplemental funding tied to the Iran conflict.

Analysis

The market is starting to price a second-order AI bottleneck: not compute, but memory and power efficiency. Micron’s margin inflection implies the most asymmetric part of the AI supply chain is shifting from accelerator branding to component scarcity, which should keep pricing power elevated for HBM and adjacent memory suppliers even if broader AI capex decelerates. That also raises the odds that foundry, substrate, and advanced packaging constraints become the next valuation drivers rather than GPU unit growth.

Qualcomm’s guidance reset is more important than the headline number because it extends the AI monetization window beyond smartphones into a multi-year datacenter roadmap. The market is likely underestimating the option value of an alternate CPU architecture pitched on energy efficiency: if hyperscalers genuinely diversify away from x86 in edge inference and agentic workloads, the winners will be the vendors that can offer lower TCO rather than raw FLOPS. The risk is execution timing — any slip from 2028 production pushes this story from near-term rerating into a long-duration call option.

SK Hynix’s U.S. listing introduces an interesting capital-markets wrinkle: it can tighten the valuation gap between Korean memory leaders and U.S.-listed AI infrastructure peers, while also creating a more liquid way for global funds to express the memory upcycle. On the other side, Alibaba’s alleged model extraction dispute reinforces that AI competition is moving from product features to IP enforcement and export-control-adjacent friction; that increases headline risk for Chinese AI beneficiaries and may keep a discount embedded in China tech multiple expansion. The main contrarian takeaway is that the AI trade is not broadening evenly — it is becoming more concentrated in scarce-enough-to-price assets, while everything else remains hostage to scrutiny, regulation, and cycle timing.

Key Decisions for Investors

• Long MU vs short NVDA on a 3-6 month horizon: the market is paying up for GPU growth while underpricing memory margin durability; use a 1:1 dollar-neutral pair, with stop if HBM pricing rolls over or NVIDIA reaccelerates guide cadence.
• Initiate a tactical long QCOM into the next 6-12 months with upside optionality into 2028: the rerating is driven by datacenter credibility, not handset beta; best expressed via call spreads to cap time-decay risk.
• Watch for a dip-buy in SK Hynix via ADR if/when listed: the IPO could create a valuation re-anchor for the memory complex; prefer a basket long HYNIX/KRX-listed peers against weaker analog or networking names to isolate memory scarcity exposure.
• Stay structurally cautious on BABA for 1-3 months: any escalation around AI IP or geopolitical scrutiny can compress multiple expansion; prefer puts or a short against a more domestically insulated China internet name to hedge idiosyncratic policy risk.
• Trim overexposed broad AI baskets on strength and rotate toward power-efficient infrastructure beneficiaries: the next leg of the trade likely rewards suppliers tied to memory, packaging, and low-power compute rather than the highest-beta AI platform names.