Intel Sold Its NAND Memory Business for About $9 Billion. Micron Is Now Worth More Than Twice What Intel Is.

Intel’s $9 billion 2020 sale of its NAND memory and storage business to SK Hynix closed in March 2025, with Intel receiving about $1.9 billion net of adjustments at final transfer. Since AI-driven demand lifted memory pricing, SK Hynix reported Q2 revenue of 79.3 trillion won (+257% YoY) and a record 76% operating margin, while Intel’s Q2 segment revenue rose 31% YoY to $5.8B but still posted a $2.1B operating loss. The article argues Intel’s exit looks costly versus Micron’s ~ $1.05T market value (~2.3x Intel’s ~$464B), concluding the foundry has not yet earned real profits despite faster revenue growth.

Analysis

The investable takeaway is not that a legacy asset was sold too early; it's that the memory complex has become the highest-beta claim on AI capex while Intel is still spending like a turnaround story without yet earning like one. That makes SKHY and MU the cleaner beneficiaries of pricing power, but the more important nuance is that memory margin expansion can be self-limiting: once gross margins get this extreme, the industry typically answers with capacity additions, and the earnings step-up becomes a 2-3 quarter window rather than a new regime.

For Intel, the exit matters less as a missed commodity windfall than as a signal of persistent capital misallocation. The market will increasingly value INTC on whether foundry can stop being a cash sink; until there is evidence of operating leverage, every incremental dollar of capex risks compressing the sum-of-parts multiple. That also creates second-order pressure on AI infrastructure names: if memory remains tight, server BOM inflation can slow deployment cadence, which is a subtle headwind to NVDA unit growth even as it supports memory vendor ASPs.

The contrarian point is that consensus may be too eager to declare a permanent strategic blunder from Intel's sale. In a cyclically inflated tape, the better question is whether MU/SKHY are now pricing peak scarcity, while INTC has the option to disappoint less if foundry losses narrow. The falsifier for the bullish memory thesis is a flattening of contract pricing or commentary from hyperscalers on capex moderation over the next 1-2 quarters; the falsifier for the bearish Intel thesis is a clear foundry margin inflection and evidence that losses are peaking by mid-2026.

Key Decisions for Investors

• Prefer long MU or SKHY versus short INTC as a 1-3 month pair trade: the spread benefits if AI memory pricing stays firm while Intel's foundry losses remain unrewarded by the market. Risk/reward is attractive because the short leg has a visible catalyst path (another quarter of cash burn), while the long leg is supported by contract pricing momentum.
• If entering a new long in memory, stagger in on any 5-8% pullback rather than chasing strength; this is a cyclical pricing trade, not a secular annuity. Use MU as the more liquid expression and SKHY only if ADR liquidity and borrow are acceptable.
• Avoid adding to INTC on the basis of the NAND exit story alone; wait for a foundry operating-loss inflection or a material guide-up in gross margin before considering a long. Until then, INTC remains a candidate for relative underperformance versus semis with actual pricing power.
• Watch NVDA and AI infrastructure names for a second-order read-through: if memory shortages begin to constrain server availability, that can become a short-term headwind to deployment velocity. Use any evidence of slower hyperscale capex as a signal to trim memory longs rather than as a reason to short the whole AI chain.