Citi Just Slapped a Massive $2,500 Price Target on SanDisk. Here's Why They're So Bullish

Citi raised its SanDisk target to $2,500 from $2,025, citing tight NAND supply, hyperscaler-driven datacenter demand, and new multi-year customer agreements that provide unusual earnings visibility. SanDisk reported $5.95 billion in revenue, with datacenter revenue up 645% YoY to $1.47 billion, gross margin expanding to 78.4%, and $2.99 billion of free cash flow alongside zero long-term debt and a fresh buyback. Management guided Q4 revenue to $7.75 billion-$8.25 billion and Non-GAAP EPS of $30-$33, but the stock still faces valuation and execution risk at roughly 70x trailing earnings.

Analysis

The market is still underpricing the shift from cyclical NAND exposure to a contracted supply business with more embedded recurring economics. That matters because once a memory vendor has multi-year volume commitments plus capital discipline across the oligopoly, the earnings profile starts to resemble a capacity-constrained toll road: upside comes less from unit growth than from pricing persistence and mix migration. In that regime, the next leg of multiple expansion is usually driven by evidence that the “good quarter” is actually the new run-rate, not by another beat.

The second-order beneficiary is the broader AI infrastructure stack, especially SSD-adjacent vendors and datacenter buildout beneficiaries that need persistent high-bandwidth storage. If NAND stays tight into 2027, hyperscalers will be forced to prioritize cache and storage architecture earlier in the procurement cycle, which can pull demand forward for controllers, packaging, and memory subsystem suppliers. The losers are customers with weak bargaining power and more spot-linked procurement; their margin pressure will show up with a lag, not immediately.

The key risk is that the setup is now self-reinforcing on the way up and vulnerable to air-pocket downside if one variable breaks: price discipline. A modest easing in capex, a digestion quarter at hyperscalers, or any sign that committed supply is being backfilled into weaker channels could compress the multiple quickly because the stock is already discounting a long-duration shortage. The move is more durable over months than days, but the near-term trade is crowded and susceptible to factor-driven de-risking in high-beta tech.

Consensus still seems to be treating this as a strong cyclical, but the underappreciated point is that the company is trying to monetize scarcity through contracting, buybacks, and debt-free balance sheet optionality simultaneously. That combination can justify a premium, but only if the market believes the contracts are enforceable through a downcycle. If not, the premium will peak well before fundamentals do.

Key Decisions for Investors

• Long SNDK on pullbacks over the next 2-6 weeks; use a staggered entry rather than chasing strength. Upside is a continued multiple re-rate if Q4 lands at the high end, but stop-loss discipline is essential because crowded positioning could unwind 10-15% on any macro tech selloff.
• Buy SNDK Jan-2027 call spreads instead of outright calls. This captures the 12-18 month thesis that NAND tightness and contract visibility persist, while limiting damage if the market decides the peak narrative has been fully priced.
• Pair long SNDK / short WDC for 3-6 months. The trade isolates the contract-driven, mix-improving business model versus the more legacy-exposed peer; risk is broad sector de-rating, so size modestly and cover if WDC starts validating similar margin durability.
• For event-driven accounts, sell downside puts after volatility spikes around earnings or guidance updates. The balance sheet and buyback reduce bankruptcy-style tail risk, so implied vol may overstate downside; this is best if you are willing to own the stock on a 10-15% dip.
• Watch hyperscaler capex commentary and NAND spot pricing weekly. If either rolls over for two consecutive prints, reduce exposure quickly because the market will likely de-rate SNDK before fundamentals visibly deteriorate.