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

Memory and AI infrastructure names are leading: Sandisk shares are up ~3,990% (still down ~19% after a pullback) as NAND shortages pushed revenue up 251% and gross margin from 22.5% to 78.4%. Micron is up ~721% on the DRAM supercycle, with HBM-driven inference demand expanding gross margins and long-term contracts now covering ~40% of revenue. AMD is up ~302% as the market shifts toward inference and agentic AI, supported by GPU partnerships totaling $100B with OpenAI and Meta and an estimated $120B agentic-AI CPU addressable market.

Analysis

The clearest winner is the memory complex, but the cleaner risk-adjusted expression is MU, not SNDK. In a shortage regime, pricing power migrates upstream to the component with the tightest substitution path; that usually means DRAM/HBM first, NAND second, and commoditized storage assemblers last. That favors MU and also lifts equipment names indirectly, while pressuring AI compute names on cost of goods sold if memory stays scarce enough to compress accelerator gross margins or delay deployments.

The second-order dynamic is that inference-heavy AI is not just “more demand for chips,” it is a reallocation of spend toward memory bandwidth and capacity. That helps AMD relative to NVDA at the margin because a larger share of AI dollars shifts from pure training compute to systems optimized for memory density and CPU orchestration; however, the market may be underpricing how much of that advantage is already in the stock after a big rerating. META and other hyperscalers are the hidden losers if memory inflation forces them to choose between model rollout speed and free-cash-flow discipline, which can show up first in capex scrutiny rather than headline earnings misses.

The main risk is that this is a cyclical pricing story masquerading as a structural one. For SNDK, the immediate catalyst can last weeks to a few quarters, but NAND supply response and customer inventory resets can reverse the move fast if spot pricing stalls; for MU, long-term contracts and HBM demand make the cycle more durable, but the stock is still vulnerable to any sign of capacity additions or a downshift in hyperscaler capex growth over the next 1-2 quarters. The contrarian view is that consensus may be overconfident in the duration of NAND tightness and underestimating how quickly additional wafer starts from Samsung/SK Hynix/others can cap the upside once margins normalize.

Key Decisions for Investors

• Long MU on 1-3 month horizon; use any post-rally pullback or earnings-related weakness to build. Risk/reward is best because long-duration contracts reduce cycle risk while DRAM/HBM remains the tighter part of the stack. Falsifier: DRAM pricing roll-over or a guide-down tied to inventory normalization.
• Fade SNDK strength via a tactical short or call spread overwrite into further spikes; 2-8 week trade. The stock is most exposed to mean reversion if NAND spot prices pause or if management commentary confirms the cycle is being extrapolated too far. Cover if NAND pricing continues accelerating for another quarter.
• Relative long AMD / short NVDA only as a valuation-sensitive pair over 1-3 months, not a fundamental conviction short NVDA. This works if the market rotates toward inference and memory-rich architectures, but it fails if training capex re-accelerates or NVDA extends its software moat. Keep size modest.
• Watch META and other hyperscaler capex proxies for knock-on margin pressure; if memory inflation starts to dent free-cash-flow guidance, short-term downside can move from component names to infrastructure spenders. No action until the next capex reset or commentary from earnings.
• If you want a cleaner macro expression, long SMH/MU versus short a basket of compute-only beneficiaries is the better structure than naked single-name shorts. The thesis is that memory scarcity is a tax on the whole AI stack, but the suppliers with pricing power will capture it first.