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

Sandisk has gained ~2,900% over the past year (as of end-August) as NAND supply shortages pushed NAND prices higher, with long-term capacity agreements totaling at least $42B and guidance implying ~80% adjusted gross margin through 2030. Micron is up nearly 700% over the same period on constrained HBM-driven DRAM demand and supply limits, trading at a forward P/E of ~6. Lumentum is up ~550% as AI data centers shift from copper to optical networks, leveraging its InP laser leadership (up to ~60% market share) and optical interconnect exposure.

Analysis

The cleanest exposure here is not the commodity memory names themselves but the structural bottlenecks around them. If AI buildouts continue, ASML and LITE should retain pricing power longer than MU/SNDK because they sit closer to constrained manufacturing steps and certification-led switching costs; that usually supports higher-quality earnings and a less fragile multiple. By contrast, memory producers can look optically cheap right before supply discipline breaks, which is why the market often overpays for the early innings of a supercycle.

Second-order, the real margin pressure lands on the AI supply chain: hyperscalers, server OEMs, and networking integrators will keep absorbing more memory and optical content per rack. That raises capex intensity and can compress near-term returns on deployed capital if utilization does not ramp fast enough; the first tell will be any pause in cloud capex commentary or order digestion at the GPU/server layer. In the next 1-3 months, momentum can persist, but the stocks most vulnerable to a disappointment are the ones already being priced as if scarcity lasts multiple years.

Contrarian view: the consensus is treating scarcity as a structural state rather than a cyclical phase. The falsifier is not a single earnings beat; it is a flattening in contract pricing plus signs that capacity additions and yield improvements are catching up faster than expected. If that happens, the multiple compression will hit MU/SNDK first, while ASML/LITE should hold up better because their revenue is driven by tooling and design-in friction rather than spot pricing.

Key Decisions for Investors

• Prefer ASML and LITE over MU/SNDK on a 6-12 month basis: long ASML / short MU as a relative-value pair, targeting resilience in bottleneck equipment versus cyclical memory ASP risk; stop if DRAM contract prices keep rising into the next two quarters.
• Initiate or add to LITE only on pullbacks after the recent momentum surge; use a 3-6 month call spread rather than outright chasing stock if implied volatility remains elevated, since the thesis is adoption durability rather than immediate multiple expansion.
• Do not add fresh exposure to MU/SNDK after strength unless you have independent channel checks showing contract pricing still inflecting up; the risk/reward turns poor if management guidance implies capex normalization or customer inventory digestion.
• Set a watch item on hyperscaler capex commentary and server OEM margins over the next 1-3 months; any slowdown there is a tell that memory and optical demand could be pulled forward rather than sustained.
• If the group keeps rallying without fresh price data, fade the most cyclical names via a small basket short in MU/SNDK against a long in ASML as a 6-18 month hedge against mean reversion in the memory cycle.