Micron vs. SanDisk: One Is Building Capacity. The Other Is Buying It.

The Market Has Misread the Structure of This Comparison

Both Micron (MU) and SanDisk (SNDK) are framed as AI memory stocks. Both have delivered extraordinary returns this year — Micron is up 207% year-to-date and SanDisk is up 411%. The implication is that they are competing for the same investment dollar. That framing misses the actual structural question: one of these companies is building its own fabrication capacity while the other is purchasing wafers from someone else's fab.

The distinction between a fab-based manufacturer and an asset-light assembler determines who controls supply, who bears cycle risk, and which business model survives when the upcycle ends.

Capex: The Smoking Gun

Micron spent $25.26 billion on capital expenditures over the trailing twelve months. SanDisk spent $177 million. That is a 142-to-1 gap.

Capex is the load-bearing metric for any semiconductor company because it tells you what the company actually owns and what risks it carries. Micron is constructing fabrication facilities in the U.S. and Taiwan to produce advanced DRAM and high-bandwidth memory (HBM), the stacked memory technology that sits directly on GPU modules and feeds AI workloads. The $25 billion buy builds a moat: HBM production requires multi-billion-dollar fabs, years of qualification cycles, and three global suppliers — Samsung, SK Hynix, and Micron.

SanDisk's $177 million in capex tells the opposite story. SanDisk does not own NAND fabrication lines at a comparable scale. It purchases NAND wafers from external foundries — primarily through its longstanding alliance with Kioxia, its former Japanese joint-venture partner — and assembles them into solid-state drives, memory cards, and enterprise storage modules. It is closer to a design-and-packaging house than a chipmaker.

This is not a value judgment. It is a structural classification. It means SanDisk's entire profit trajectory depends on external wafer supply and external pricing. When NAND is tight, SanDisk earns the spread. When NAND is loose, that spread collapses. SanDisk has no capacity of its own to constrain.

Margin Convergence, Different Risk Profiles

The numbers are strikingly similar right now:

• Micron: 72.57% gross margin, 65.63% operating margin, 58.63% ROIC
• SanDisk: 71.47% gross margin, 61.19% operating margin, 80.63% ROIC

Both companies are earning elite margins. But the risk profiles are inverted. Micron's margins are supported by fixed-price contracts covering its entire 2026 HBM supply, giving it multi-year revenue visibility. SanDisk's margins are a function of current NAND tightness — a condition it did not create and cannot control. SanDisk has a higher ROIC precisely because it owns fewer assets.

The free cash flow contrast sharpens the picture. Micron generated $26.17 billion in free cash flow TTM (trailing twelve months) while spending $25.26 billion on capex — meaning operating cash flow was $51.43 billion. SanDisk generated $11.49 billion in free cash flow TTM while spending $177 million on capex — essentially all operating cash flow, $11.67 billion, flowed through. SanDisk's cash conversion is extraordinary. Micron's is enormous but capital-intensive.

Revenue Trajectory

Micron's Q2 FY2026 revenue was $23.86 billion, up 167% year-over-year. Q3 guidance points to $33.5 billion with approximately 81% gross margins. EPS was $12.20 in Q2 and is expected to reach $19.15 in Q3.

SanDisk's Q2 FY2026 revenue was $3.025 billion, up 175% year-over-year. EPS was $6.20 in Q2. The consensus estimate for Q3 is $0.88.

Micron is the larger company by an order of magnitude. SanDisk's growth rate is similar, but the base is much smaller and the trajectory is driven by one commodity cycle: NAND pricing. Micron's revenue is diversified across DRAM, HBM, and NAND — and the HBM segment provides the structural advantage that justifies the capex burden.

The Two-Market Split

The memory market has bifurcated into two distinct sub-markets:

1. HBM/DRAM for AI training and inference compute — a three-supplier oligopoly with high barriers to entry, multi-year qualification cycles, and fixed-price contracts. Micron participates here. The constraint is fabrication capacity, and the supplier controls both technology and volume.

2. NAND flash for AI storage — a broader competitive set with Samsung, SK Hynix, Kioxia, and Micron as wafer producers, and SanDisk, Samsung, and others as SSD assemblers. The constraint is tighter but more fragile. Samsung or SK Hynix can expand NAND wafer output faster than any company can build new HBM capacity. SanDisk has a collaboration with SK Hynix on High Bandwidth Flash — a new technology targeting AI inference workloads — but this partnership makes SanDisk dependent on SK Hynix's supply priorities, not independent of them.

In the first market, the supplier sets the terms. In the second market, the wafer producer does. SanDisk is a customer in the NAND supply chain.

The Valuation Question

Micron trades at a market cap near $1.2 trillion, up from roughly $160 billion a year ago. The stock trades at approximately 44x trailing earnings but below 9x forward earnings — a gap that reflects expected earnings acceleration from HBM volume ramps and margin expansion toward that 81% gross margin target.

SanDisk trades at approximately $169 billion in market capitalization, up from roughly $38 per share at its February 2025 spinoff. The stock has risen 6,500% from spinoff lows. At current levels, SanDisk trades at roughly 196x its most recent reported EPS of $6.20.

Both stocks have been bid into positions where any stumble carries disproportionate consequence. The market has priced both companies to deliver flawless execution through the remainder of the AI infrastructure buildout.

What Breaks Each Thesis

Micron: If hyperscaler capex decelerates — if the projected $700+ billion in AI data center spending moderates — HBM demand softens and Micron's $25 billion annual capex becomes overcapacity risk. The buildout takes 18–24 months to come online; if demand turns before capacity arrives, the cycle inverts. Micron also faces competitive pressure if Samsung aggressively expands HBM4 capacity.

SanDisk: If NAND supply discipline breaks and Samsung or SK Hynix expand wafer output, pricing pressure arrives quickly. SanDisk has no capacity of its own to constrain. Its 80.63% ROIC was earned on $177 million in capex — it is a margin play, not a capacity play. The spinoff created a pure NAND entity riding a pricing upcycle. When the cycle turns, the asset-light model becomes asset-vulnerable.

Investor Takeaway

The key issue is not which company has the better AI exposure. Both do. The more important question is which business model survives when the current supply constraint loosens — and when that loosening happens.

Micron owns the bottleneck. Its $25 billion capex builds a structural position in HBM that cannot be replicated quickly by competitors. If hyperscaler spending remains on its current trajectory, Micron's capacity becomes increasingly scarce relative to demand.

SanDisk rents the bottleneck. Its extraordinary margins and 80.63% ROIC are a function of NAND tightness that it did not create. The company is well-managed and well-positioned within the current upcycle, but its entire earnings profile is contingent on external wafer producers maintaining supply discipline.

The investor's choice is not between two AI memory stocks. It is between owning the supply constraint (Micron) and riding it (SanDisk). The latter earns the higher current returns. The former earns the higher durability. The question is which matters more as the cycle ages.

The forward condition to watch: NAND contract pricing in the next two quarters. If Samsung or SK Hynix signal aggressive capacity expansion — or if hyperscaler storage spending decelerates — SanDisk's thesis narrows rapidly. If NAND tightness persists, the current spread in favor of SanDisk's asset-light model holds. But the constraint is always temporary in a market where wafer producers set the volume.