Nvidia Stock Rallied 39% in 2025. This Year, It Could Go Much Higher

Nvidia recovered from a 37% plunge in 2025 to finish the year up 38.9%, reporting fiscal 2026 Q3 revenue of $57.0 billion (+62% YoY) and EPS of $1.30 (+67%); management guided Q4 revenue of $65.0 billion (implying +84% YoY). CEO Jensen Huang said next‑generation Vera Rubin chips are in full production six months early and will cut AI inference costs ~90% versus Blackwell, CFO says the prior $500 billion revenue estimate for 2026 is too conservative, and limited Chinese approvals for H200 chips could meaningfully expand addressable demand—supporting a bullish case at under 25x next‑year earnings.

Analysis

Market structure: Nvidia is the primary beneficiary — the Q3 $57B revenue run-rate and management pushing Q4 to $65B (84% YoY growth) imply widening TAM and pricing power for high-end datacenter GPUs. Suppliers of HBM memory, advanced packaging, and TSMC capacity will see demand shocks; GPU competitors and low-margin inference vendors face margin pressure if Rubin drives a 90% inference-cost drop. Cross-asset: strong NVDA-led risk-on could steepen yields (higher growth expectations), compress IG spreads, lift semiconductor capex stocks and bid HBM/metal substrate commodity prices (copper/aluminum up 3–8% in prior ramps), while options markets should see elevated NVDA IV around catalyst windows.

Risk assessment: Tail risks include a renewed US/China export blockade, unexpected Rubin production issues, or a global data-center capex freeze — any could wipe 20–40% off forward revenue expectations in 3–12 months. Immediate (days) risk is event volatility around CES/earnings; short-term (weeks/months) hinge on China H200 approvals and supply ramp; long-term (years) rests on Rubin adoption and cloud providers re-pricing services. Hidden dependencies: China revenue potential (~$50B/yr cited) is highly conditional on selective approvals and customer credit/FX, and lower inference costs may force cloud pricing competition that compresses ASPs for OEMs.

Trade implications: Direct: establish a controlled NVDA exposure to capture Rubin-driven upside but hedge execution risk. Use relative-value: long NVDA vs short AMD to express platform share consolidation. Options: favor debit call spreads to limit capital at risk into known catalyst windows (earnings, China approvals). Sector: overweight semicap, HBM suppliers and cloud infra; underweight legacy GPU OEMs and small AI chip entrants.

Contrarian angles: Consensus underestimates the risk that a 90% cost cut commoditizes inference hardware leading to ASP deflation and margin erosion across the stack over 12–24 months; NVDA’s revenue beat could be front-loaded. The market may be underpricing geopolitical tail risk — if China approvals stall, a 10–20% revenue haircut is plausible. Historical parallels: prior rapid GPU ramps (2016–2018) produced both vendor concentration and subsequent cyclical overcapacity; watch capex orders and wafer starts for signs of overheating.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (size relative to portfolio) within 1–4 weeks to capture Rubin and Q4 guidance, add a second tranche on any pullback >7% from current levels, target a 6–12 month horizon and consider trimming at +40% realized gain or if forward P/E exceeds 30x.
• Buy a limited-risk call spread on NVDA expiring in Mar 2026 (debit) sized at 0.5–1.0% portfolio: buy ATM call, sell 25% OTM call to express bullishness into earnings/China-approval catalysts while capping premium paid; close 3–5 trading days before earnings if IV spikes >30% above median.
• Construct a pair trade: long NVDA 1.0 / short AMD 0.5 notional to express share gain bias for 3–6 months; rebalance if NVDA outperforms by >15% or if AMD reports a surprise win in datacenter design wins.
• Overweight semiconductor-capex and memory suppliers (target 3–6% overweight to base) for 6–12 months (vendors with HBM, advanced packaging, TSMC capacity exposure). Increase exposure by 50% if China H200 approvals are formalized within 30–60 days; reduce if wafer-start indicators fall two consecutive months.