Better AI Buy: Nvidia vs Micron

Nvidia reported fiscal-year revenue up 65% to $215 billion driven by AI GPUs and platform expansion, with ongoing product cadence (Blackwell, Blackwell Ultra, Rubin). Micron delivered record revenue, gross margin, EPS and free cash flow and expects to double sequential FCF next quarter, but says it can only meet ~50–66% of AI-driven demand due to supply constraints. Micron shares are up ~50% YTD while Nvidia is down ~5% YTD; the author prefers Nvidia as the better long-term AI buy based on valuation and roadmap despite Micron's stronger near-term performance.

Analysis

AI-driven demand is now bifurcating the semiconductor value chain: compute (NVDA) is becoming a platform business where software, interconnects, and recurrent upgrade cycles create high-margin annuity-like flows, while memory (MU) behaves as an industrial good with much shorter inventory cycles and acute capacity elasticity. That dynamic creates asymmetric timing opportunities — compute benefits from multi-year platform lock‑in that compounds gross margins, whereas memory’s upside is concentrated in 3–9 month windows when capacity lags demand and spot ASPs spike.

Micron’s current inability to meet demand is a two‑edged sword: it supports near-term pricing power and FCF outperformance but invites accelerated capex and customer diversification by hyperscalers. Expect competitors and captive hyperscaler fabs to accelerate HBM/DRAM supply expansion within 9–18 months, which can mechanically reverse MU’s margin tailwind if Micron doesn’t convert excess demand into contractual higher-ASP commitments.

For Nvidia, the second‑order risk is platform complacency: annual cadence buys time, but large cloud providers are already deploying custom inference silicon and software stacks that can undercut GPU utilization economics for certain workloads within 12–24 months. Geopolitical export controls and a softer enterprise AI refresh cycle are the primary catalysts that could reset multiples quickly, despite a durable structural TAM.

Net-net, tradeable edges are timing and structure — capture Micron’s near-term constrained pricing with time‑limited, delta‑positive exposure, and express conviction in Nvidia’s multi-year platform by buying long‑dated, mildly financed optionality while hedging a 12–18 month risk of competitive or policy shocks.

Key Decisions for Investors

• Long NVDA via a 9–15 month call spread to buy convexity while capping cost: buy NVDA Jan-2027 LEAP call / sell NVDA Jan-2028 higher strike call sized to risk 0.75–1.5% NAV. Target +40–80% on spread if platform re-acceleration; stop-loss: worsen to -40% of premium if cloud capex guidance misses by >10% over a quarter.
• Tactical long MU for the next 3–6 months to capture supply-driven ASP upside: buy MU outright or 3‑month call spread sized 0.5–1% NAV. Risk/reward: target +30–50% vs max downside -25% if spot memory prices roll over; trim into signs of announced 12–18 month capacity additions from SK Hynix/Samsung.
• Pair idea (risk‑managed): long MU 3–6 month calls funded by short NVDA 1–3 month calls to monetize elevated NVDA short-term vol while maintaining directional long NVDA via LEAPs. Expected outcome: collect premium if NVDA consolidates near-term, while retaining upside to NVDA’s multi-year story; monitor for earnings-driven vol spikes and adjust 48–72 hours ahead of prints.
• Hedge/insurance: buy small INTC or broad Nasdaq put protection (0.25–0.5% NAV) with 3–6 month tenor to protect against a policy/export-control driven drawdown that would hit both NVDA and MU heavily within days to weeks.