This AI Stock's Valuation Makes No Sense to Me. Here's Why That's a Good Thing.

Sandisk has surged about 4,100% since its February 2025 IPO, with the stock up roughly 520% year to date and still trading at 23x forward earnings. Fiscal Q3 revenue jumped 251% year over year to $5.9 billion, net income rose 350% to $3.6 billion, and gross margin expanded to 78.4%, while Q4 revenue guidance of $7.75 billion-$8.25 billion implies another 36% sequential increase at the midpoint. The article argues the rally is supported by rapid earnings growth, AI-driven memory demand, and backlog extending into 2027, leaving the stock looking relatively inexpensive despite the massive run.

Analysis

The key market implication is that SNDK is no longer trading as a “story stock”; it is behaving like a quasi-cycle monopolist with visible supply scarcity and near-term earnings power that is still being revised upward. That matters because the stock can keep rerating even after a massive move if backlog visibility extends into 2027 and pricing remains firm, but the marginal buyer is now paying for peak-ish margins rather than a normal-cycle multiple. The real issue is not whether the business is good — it is whether the market is underestimating how long the shortage can persist before competitors and customers force normalization.

Second-order effects are likely to show up in the rest of the AI supply chain. If hyperscalers are already locking in memory capacity, that usually pulls forward procurement budgets and can crowd out less strategic storage buyers, which should pressure smaller NAND-dependent peers and any OEMs without long-term supply agreements. It also creates a “scarcity premium” across adjacent infrastructure names, but the flip side is that a single large customer reset in capex can unwind pricing faster than investors expect once inventories normalize.

The contrarian miss is that valuation looks cheap only if the current earnings base is durable. A 23x forward multiple on rapidly compounding earnings can still be expensive if the cycle turns down in 6-12 months, especially in a memory business where marginal supply can arrive abruptly. The path dependency is the risk: the stock can keep working for weeks to months on backlog and guidance upgrades, but the downside becomes severe if gross margin peaks before revenue growth does.

The setup favors momentum continuation in the near term, but it is increasingly a late-cycle quality-vs-cyclicality debate disguised as a valuation story. The strongest tell will be whether management keeps extending backlog duration and pricing power into 2027; if not, the market will start discounting the next leg of the memory cycle much earlier than consensus expects.

Key Decisions for Investors

• Maintain a tactical long SNDK only on pullbacks, not strength; use a 4-8 week horizon and size for cycle volatility, because the trade is still supported by backlog momentum but vulnerable to any guide-down in margin trajectory.
• For a relative-value expression, short a basket of memory-adjacent or AI-infrastructure laggards against long SNDK for 1-3 months; the thesis is supply scarcity plus pricing power will separate winners from the rest of the storage chain.
• Buy downside protection on SNDK via 3-6 month puts or put spreads if already long; the risk/reward improves if the market starts pricing peak margins before peak earnings are reached.
• Avoid chasing NVDA/INTC on this read-through; the article’s positive signal is much more specific to storage bottlenecks than to compute, so any benefit to semis is indirect and likely already reflected.
• If looking for a contrarian entry, wait for the first evidence of backlog duration flattening or gross margin inflection; that is the point where the multiple can compress quickly despite still-strong headline growth.