Nvidia stock falls: why can't it benefit from increased AI capex

Nvidia shares fell 2.18% on Thursday despite Alphabet raising 2026 capital expenditure guidance, suggesting investors are rotating from megacap tech into memory/semi peers. Micron rose 2.5% and SK Hynix gained 2.8%, while NVDA snapped a 3-day winning streak amid a relative strength shift toward memory specialists.

Analysis

The market is not rewarding the headline AI spend; it is repricing where the marginal dollar lands in the supply chain. If hyperscaler capex is moving from a pure compute story toward memory-heavy architectures, the economic transfer shifts from the largest platform vendors to the bottleneck suppliers, which is why MU and SK hynix can outperform while NVDA derates on a relative basis. That is a flow and positioning effect first, but it can become a fundamentals story if memory lead times and pricing power begin to extend.

Over the next 1-3 months, the key question is whether this rotation is supported by order-book evidence. If HBM/DRAM allocations tighten, memory names can see estimate revisions before revenue shows up, creating a cleaner catalyst path than the broader AI complex. If not, the move is likely a short-duration factor swap: investors will rotate back to the names with the clearest shipment visibility and strongest balance sheets.

Contrarian view: the consensus may be overreading one capex guide as a durable regime change. In AI buildouts, early spend often looks like a broad beneficiary trade, but returns can reconcentrate in the company controlling system design, software stack, and customer lock-in; that favors NVDA longer term if its platform capture remains intact. The move is therefore more attractive as a relative-value trade than as a thematic all-in bet, with the main falsifier being any evidence that memory pricing or hyperscaler memory attach rates fail to improve in the next quarter.

Key Decisions for Investors

• Initiate a 1-3 month long MU / short NVDA pair trade on any further NVDA strength; target 8-12% relative outperformance if memory pricing tightens, stop if NVDA regains leadership and MU guidance does not inflect.
• Add a smaller long SKHYV position versus SOXX or QQQ as a higher-beta HBM beneficiary; use reduced sizing because ADR liquidity can amplify reversals if the rotation is just a flow trade.
• Use NVDA call spread selling or put spreads for a defined-risk short into strength over the next 4-8 weeks; this expresses the view that multiple compression can persist even without a fundamental miss.
• Treat MU/SKHYV as a watchlist upgrade only if channel checks show HBM/DRAM lead times extending over the next 30-60 days; if pricing stays flat, fade the move rather than chase it.