Better Semiconductor Stock During the Crash: Nvidia or AMD?

Nvidia reported Q4 FY2026 revenue of $68.1B (up 70% YoY) with 75% gross margins; AMD reported Q4 FY2025 revenue of $10.3B (up 34% YoY) with 54% gross margins. AMD's CPU market share rose to 29.2% in Q4 2025, it announced partnerships with OpenAI and Meta, and it has a trailing-12-month PEG under 0.5; AMD stock is up ~88% over the last year (as of Mar 20) versus Nvidia's ~46% gain. The author recommends Nvidia based on superior revenue growth and GPU dominance but notes a case for owning both to capture each firm's strengths.

Analysis

The real contest is not GPU vs. CPU in isolation but who captures the higher-margin layers above silicon: software stacks, optimizer toolchains, and long-term hyperscaler validation. That creates asymmetric optionality — a firm that locks-in compiler/runtime or model-optimization flows converts one-time silicon wins into multi-year annuity-like revenue and pricing power, and that is the axis investors should track over the next 6-24 months.

Second-order supply effects are underappreciated. Foundry and advanced packaging capacity will become the choke points that determine who can scale throughput to hyperscaler demand; wins in procurement cadence (multi-quarter commitments from a few cloud buyers) will amplify gross-margin differences faster than product-perf benchmarks. Watch booking patterns at TSMC/ASML and OSAT lead times as early warning signals.

Downside drivers are classic but timing-sensitive: a hyperscaler inventory digestion cycle, an AI model architecture pivot away from current accelerators, or regulatory pressure on software bundling could compress multiples quickly. Conversely, incremental software/SDK monetization or an enterprise contract that forces standardized deployment can re-rate share prices abruptly — catalysts that typically materialize inside 3–12 months.

Key Decisions for Investors

• Pair trade (6–12 months): Long NVDA / Short INTC (equal dollar exposure). Rationale: capture asymmetric upside from continued accelerator-led data center spend while hedging broader x86/server cyclicity. Position size: keep net delta <0.5 of portfolio to limit systemic equity risk. Risk/reward: expect 25–50% upside on the long leg vs 0–15% downside on the short if AI demand executes; primary risk is AI cycle disappointment.
• Defined-risk options (9–15 months): Buy NVDA 1-year ATM call and sell a 25–35% OTM call to fund ~50–70% of premium (vertical call spread). Rationale: levered exposure to multiple expansion and sustained data-center demand with capped max loss (premium net). Target payoff: asymmetry where >40% equity move produces >2x option ROI; loses limited to paid premium if thesis fails.
• Income/convexity on AMD (3–9 months): Buy-and-write AMD (buy shares, sell 3-month OTM calls) sized as a tactical overweight. Rationale: capture upside optionality from CPU foothold while harvesting premium to reduce basis. Risk/reward: reduces downside by premium received; cap upside to strike — roll calls if upside materializes.
• Portfolio hedge (3–6 months): Buy NVDA 3–6 month put spread sized to existing long NVDA exposure (or long equity basket skewed to AI names). Rationale: cheap tail protection against rapid multiple compression or hyperscaler inventory drawdown. Cost: limited to premium; benefit is non-linear protection during headline risk events.