3 AI Stocks Up 500% or More in the Past Year That Could Have More Room to Run

AI-driven demand for HBM and optical networking is powering outsized gains: Sandisk is up ~2,900% YoY (as NAND supply/demand tightness supports higher pricing and ~80% adjusted gross margin), Micron is up nearly ~700% YoY on constrained DRAM/HBM supply and low valuations (forward P/E ~6), and Lumentum is up ~550% YoY benefiting from data centers shifting from copper to optical networks. Sandisk also disclosed long-term agreements worth at least $42B (covering one-third of expected fiscal 2027 capacity) with revenue growth projected in the mid-to-high teens (FY2028–2030) while keeping gross margin around 80%. The piece frames these stocks as “cheap” on forward P/E despite strong momentum, implying continued upside if the AI memory/optics supercycle sustains.

Analysis

This is less a generic AI-beta story than a pricing-power story in constrained manufacturing steps. The market is rewarding companies that sit on scarce capacity with long qualification cycles, which means the real beneficiaries are the bottlenecks upstream of AI deployment rather than the obvious compute names. ASML is the stealth winner here: if HBM continues to consume EUV capacity, the capex funnel stays tight and memory pricing can remain elevated longer than consensus expects.

The near-term risk is not demand evaporation; it is supply response and sentiment mean reversion after a parabolic move. Over the next 1-3 quarters, watch for lead-time compression, customer inventory normalization, and any language about incremental wafer starts or capex re-acceleration from Samsung/SK hynix/WDC—those are the first signs the margin tailwind is peaking before the P&L shows it. LITE’s setup is strong but lumpy: certification stickiness creates durability, yet adoption of optical interconnects can disappoint if hyperscaler rollouts slip even one platform cycle.

The contrarian view is that these names may still be under-owned on a 12-24 month basis because investors are anchoring to old memory-cycle behavior. If the industry is moving from spot-market cyclicality to multi-year contracted supply, the correct multiple is not a classic trough P/E. The thesis breaks if gross margins stop expanding while backlog/contract visibility fails to translate into shipments; that would tell us the market has overpaid for duration.

Key Decisions for Investors

• Long MU vs short WDC for the next 1-2 earnings cycles: express memory scarcity with better mix and HBM leverage on the long leg; cover the short if NAND pricing reaccelerates or WDC guides above expectations.
• Buy SNDK on a 5-10% pullback rather than chasing strength; use a tight stop if NAND ASP commentary or contract backlog disappoints, since the stock is already discounting an extended supercycle.
• Prefer LITE via a 9-12 month bull call spread instead of outright stock due to parabolic momentum; the payoff is best if optical adoption continues through the next platform rollout window, and wrong if hyperscaler certification slows.
• Initiate a smaller tactical long ASML as a stealth beneficiary of HBM bottlenecks; this is the lower-beta way to own the EUV scarcity thesis, with the thesis falsified by any sign of EUV capacity loosening or order deferrals.