3 Stocks That Crushed Nvidia With 300% or More Gains Over the Past Year

Memory and AI chip beneficiaries are highlighted: Sandisk is up ~3,990% over the past year (despite a 19% pullback) after revenue surged 251% and gross margin jumped from 22.5% to 78.4% on a NAND shortage, with Wedbush raising its fiscal Q4 outlook. Micron is up ~721% as DRAM/DRAM-HBM demand accelerates, expanding gross margin and enabling long-term contracts covering 40% of revenue; the stock trades at a forward P/E under 6.5x. AMD is up ~302% over the year, supported by inference/agentic AI momentum, $100B GPU partnerships (OpenAI, Meta), and a stated $120B agentic AI CPU TAM outlook.

Analysis

The cleanest read-through is that AI spend is shifting from pure compute scarcity to memory/content scarcity. That is structurally better for MU and SNDK than for GPU-only exposure because every incremental inference dollar requires more bytes moved, stored, and served; if that mix persists, memory can keep pricing above marginal-cost returns for multiple quarters, while GPU multiples become more sensitive to any capex normalization.

AMD is the second-order beneficiary if inference adoption broadens: buyers optimizing for throughput-per-dollar, not training prestige, will keep testing alternatives to NVDA. The risk is that NVDA’s software stack and packaging roadmap can still recapture share if it counters with bundled systems and memory-heavy configs; in that case AMD outperformance becomes a tactical trade, not a structural one.

The contrarian miss is that the market may be underestimating how cyclical "supply shortage" stories are once suppliers lock in contracts. Those contracts improve visibility, but they also mark the point where future upside gets pulled forward; if spot pricing stops rising for 4-8 weeks or contract renewals come in weaker than expected, these names can de-rate fast despite still-strong reported growth. For the broader semis basket, the key is whether AI capex expands or merely re-allocates from GPUs into memory and storage.

Key Decisions for Investors

• Long MU vs short NVDA, 3-6 month relative-value trade: express the view that AI dollar intensity is shifting toward memory content faster than toward incremental GPU share. Falsify if NVDA reaccelerates on enterprise/Blackwell demand or MU guides down on DRAM pricing.
• Buy SNDK on post-rally pullbacks rather than chasing strength; use a 1-2 quarter horizon and treat it as a cyclical momentum trade, not a hold-to-perpetuity compounder. Exit if NAND spot prices flatten or next gross margin print fails to expand.
• Initiate a small AMD/NVDA pair on a 6-month horizon, targeting inference and agentic-AI share gains. Risk/reward is attractive only if AMD execution stays on schedule; cover if AMD commentary slips while NVDA maintains >80% incremental AI capex share.
• If you want lower beta expression, add MU through SOXX/SMH instead of single-name size. The edge here is broader memory exposure with less idiosyncratic risk; thesis breaks if memory ASPs roll over for two consecutive months.
• Watch META capex commentary as a leading indicator for inference spend mix. If the company shifts more budget toward memory-rich inference deployments, it is a confirming signal for MU/AMD and a mild negative for compute-only leaders.