Here's the One Storage Stock I'd Buy With $2,000

AI data-center storage demand has driven multiyear capacity commitments for Western Digital, Seagate and Sandisk, with Seagate pricing contractually set for all of calendar 2027 and most nearline capacity allocated into 2028. Seagate reported fiscal 2026 revenue of $12.2B, up 34%, with Q4 revenue up 48%, adjusted gross margin expanding to 52.7% from 37.9%, and Q1 revenue guidance of about $4.1B implying 56% year-over-year growth. Western Digital posted $3.75B in fiscal Q4 revenue, up 44%, and guided to roughly $4.1B for the next quarter; Sandisk's fiscal 2026 revenue rose 175% to $20.25B, supported by memory pricing and data-center demand. The article favors Seagate at roughly 15x expected fiscal 2028 earnings because its volumes and prices are more fully contracted, while warning that a slowdown in data-center investment could still pressure storage equities.

Analysis

The key investable distinction is not nominal backlog but enforceability: fixed-price, build-to-order commitments reduce STX's 2027 earnings volatility, while WDC retains more residual spot-price exposure. That makes STX the cleaner near-term earnings compounder if hyperscaler capex remains intact, but also caps upside from further HDD price inflation; WDC has greater operating leverage if tight supply persists into 2027. Verify cancellation provisions, minimum-volume commitments, and customer concentration in filings before assigning contracted-volume multiples to either name.

The second-order beneficiary is the HDD supply chain, particularly component and media suppliers, but the more important competitive issue is technology execution. STX's HAMR qualification/ramp could create an enduring capacity-per-platter and cost advantage in high-capacity nearline drives; a delayed ramp would instead force expensive legacy capacity additions and reopen share opportunity for WDC. Over 6-18 months, hyperscaler storage architecture is the structural risk: improved data reduction, greater use of QLC NAND, or a capex digestion phase can weaken exabyte demand even if headline AI spending remains strong.

SNDK's price floors are economically valuable only to the extent counterparties remain creditworthy and the contracts specify meaningful volume remedies. NAND remains structurally more cyclical than nearline HDDs because industry supply can respond faster; floors may protect downside but can also prompt customers to shift incremental workloads toward HDD or alternative flash suppliers. Consensus appears to be treating all three as equivalent AI-storage exposures, underpricing the dispersion between STX's contracted earnings visibility, WDC's upside beta, and SNDK's NAND-cycle risk.

Key Decisions for Investors

• Initiate a 3-6 month long STX / short WDC pair, sized beta-neutral, only after confirming STX's fixed-price commitments and WDC's uncontracted 2027 mix in next filings. Target 10-15% relative upside from a valuation/earnings-visibility premium; exit if STX guides below contracted shipment expectations or WDC demonstrates comparable multi-year price protection.
• For directional exposure, accumulate STX on post-earnings volatility rather than chase pre-results; use a 9-12 month horizon through 2027 booking conversion. Risk is a hyperscaler capex pause or HAMR execution miss, which should be treated as thesis-breaking if nearline unit guidance is cut materially despite claimed allocation.
• Maintain WDC as a tactical upside watch rather than a core long: buy only if confirmed pricing or capacity commitments extend meaningfully beyond the currently disclosed horizon. Its higher residual spot exposure offers better upside in a renewed HDD shortage, but that same exposure creates faster multiple compression if pricing normalizes.
• Avoid adding SNDK solely on stated price floors until management discloses floor economics, contracted volumes, and remedies for customer nonperformance. A better trigger is evidence that realized NAND pricing remains above floors while gross margin holds through a broader memory-price correction; otherwise use SMH/SOXX rather than single-name NAND risk.