These 3 Stocks Have Been the Hottest Buys on the Nasdaq-100 This Year, and Here's Why They Could Still Go Higher

Sandisk, Micron Technology, and Marvell Technology are all up more than 200% in 2026, driven by surging AI-related demand for memory and custom chips. Sandisk posted revenue of just under $6 billion, up 251% year over year; Micron reported $23.9 billion in revenue, up 196%, with net income up 771%; and Marvell’s revenue rose 28% to $2.4 billion, with management guiding to 35% growth on exceptional AI bookings. The article argues valuations remain reasonable relative to growth, and Jensen Huang’s endorsement adds to the bullish long-term case.

Analysis

The equity market is signaling that AI capex is still in the acceleration phase, but the second-order winner here is the supply side of the memory stack: pricing power is shifting from hyperscalers to component vendors with constrained capacity and long qualification cycles. That typically extends the cycle longer than consensus expects, because buyers cannot easily switch memory architectures once designs are locked, which supports both near-term margin expansion and delayed mean reversion.

The main asymmetry is that Sandisk and Micron are benefiting from the same scarcity dynamic, but they are not equally positioned if supply normalizes. Micron has the cleaner operating leverage to HBM and AI infrastructure demand, yet it is also the most exposed to an eventual memory downcycle; Sandisk’s valuation can stay elevated as long as retail flows and narrative momentum persist, but it is more vulnerable to multiple compression once growth decelerates. Marvell is the better “pick-and-shovel” exposure because custom silicon and connectivity spend can keep compounding even if memory pricing cools, though its current valuation leaves less room for disappointment.

The consensus miss is that the easy money may already have been made in the obvious AI beneficiaries, while the next trade is in the supporting ecosystem and in hedging the eventual inventory unwind. If hyperscaler capex growth slows over the next 2-3 quarters, memory names will likely re-rate first, while names tied to networking, interconnect, and custom ASIC design should hold up better. Jensen Huang’s endorsement of Marvell matters less as a fundamental signal than as a demand-validation event that can sustain multiple expansion for a few more months.

Near term, the key catalyst is continued earnings guide-up from AI bookings; the key risk is any evidence of channel inventory build or a pause in datacenter capex, which would hit these names hard within one reporting cycle. On a 6-12 month horizon, this is still a favorable tape for the group, but the risk/reward is now more selective and favors relative value over outright beta.

Key Decisions for Investors

• Long MRVL / short SNDK basket for 3-6 months: express a view that custom silicon and networking can sustain multiple support better than pure memory, with lower downside if AI spend remains healthy.
• Take profits on a portion of MU longs into strength over the next 2-4 weeks; keep a core position, but trim because the stock is most exposed to a future memory-price inflection.
• Buy MRVL Jan-2027 calls or call spreads on any post-earnings pullback: upside remains tied to AI capex duration, but valuation leaves less margin for error, so structure matters.
• Avoid initiating fresh long SNDK at elevated levels unless there is a 10-15% correction; momentum is strong, but this is the most vulnerable name to multiple compression if retail flows slow.
• Monitor hyperscaler capex commentary this earnings season; if guidance decelerates, rotate from memory names into higher-quality AI infrastructure names within 1-2 trading sessions.