2 Trillion-Dollar AI Infrastructure Stocks With Up to 136% Upside, According to Select Wall Street Analysts

Raymond James analyst Simon Leopold set a $515 Nvidia price target, implying 136% upside from its Sept. 11 close, citing persistent AI data-center hardware shortages, potential manufacturing-capacity expansion, and GPU pricing power. Melius Research's Ben Reitzes sees Micron reaching $2,200 per share, or 126% upside, after the company secured at least $100 billion of memory strategic customer agreements through 2030. The bullish case centers on sustained demand for Nvidia GPUs and Micron HBM, DRAM, and NAND, with long-term supply contracts potentially supporting Micron's elevated margins despite memory's historical cyclicality.

Analysis

The key investable distinction is not headline AI demand but where incremental capacity converts into durable economics. NVDA remains exposed to rack-level bottlenecks—advanced packaging, HBM availability, networking and power delivery—so loosening any one constraint can accelerate shipments but may simultaneously shift bargaining power toward hyperscalers and compress system-level pricing. The better second-order beneficiaries are capacity enablers such as TSM, AVGO and ANET; each captures AI spend with less dependence on a single accelerator architecture.

MU’s valuation case hinges on whether contracted HBM volume represents enforceable take-or-pay economics with price-reset protection, rather than nonbinding supply understandings. Memory pricing remains structurally vulnerable to Samsung and SK Hynix adding qualified HBM output; a faster qualification cycle would cap margins well before end-demand weakens. Over the next 1-3 months, earnings commentary on HBM mix, yields, contract duration and customer prepayments matters more than aspirational long-dated revenue figures; over 6-18 months, the risk is a conventional memory downcycle disguised by temporarily scarce AI-grade product.

Consensus is likely underweighting the financing constraint on AI infrastructure. If hyperscaler capex remains elevated but inference monetization disappoints, buyers can reduce accelerator orders without abandoning AI roadmaps, creating a sharp inventory correction through the supply chain. NVDA is more vulnerable to a multiple reset because expectations embed sustained scarcity; MU is more vulnerable to gross-margin disappointment because its earnings are more cyclical. The cited extreme target prices should not be treated as price-discovery signals absent independently verified assumptions for unit volumes, ASPs, and normalized margins.

Key Decisions for Investors

• Maintain a relative-value long MU / short NVDA basket for a 3-6 month horizon only if MU trades at a meaningful discount on forward EV/EBITDA and HBM gross-margin guidance is raised; target 15-20% relative outperformance. Exit if MU guides HBM bit growth below plan or NVDA data-center backlog conversion accelerates without incremental discounting.
• Prefer long AVGO and ANET as diversified AI infrastructure exposure over adding NVDA beta ahead of results; use 5-8% downside stops or hedge with SMH puts. Their upside catalyst is continued scale-out networking and custom silicon demand even if GPU share dynamics normalize.
• Do not underwrite a standalone MU long solely on long-term customer-agreement claims. Upgrade to a directional long only after confirmation of customer prepayments, take-or-pay terms, and HBM supply extending beyond the next fiscal year; otherwise treat MU as a cyclical trading position around earnings.
• Buy 3-6 month SMH put spreads only if hyperscaler capex guidance begins to decelerate sequentially or NVDA lead times normalize materially; this is the clean hedge for a broad AI hardware inventory reset, with risk limited to premium.