Nvidia smashes expectations again, but investors shrug

Nvidia reported blowout Q4 results—revenue $68.1bn vs. $66.2bn expected and adjusted EPS $1.62 vs. $1.53—driven by data‑centre sales of $62.3bn (up ~75% YoY) which now comprise over 91% of sales; net income nearly doubled to $43bn and cash rose to $62.6bn. Management gave striking Q1 guidance with a $78bn midpoint (~77% implied growth) that excludes China, highlighted a 263% YoY jump in networking revenue to $10.98bn, first Vera Rubin samples shipped, and supply‑chain diversification (TSMC Arizona, Foxconn Mexico). The print confirms accelerating AI demand but the market’s muted reaction reflects already elevated expectations and valuation/duration questions about how long the training‑led spending wave will persist and how Nvidia will fare as AI shifts toward inference.

Analysis

Market structure: Nvidia is the clear winner — NVDA captures accelerating training demand (data‑centre revenue +75% y/y; Q1 guide implying ~77% growth) and now benefits from a second‑order boom in high‑speed networking (network revenue +263% y/y). Hyperscalers (GOOGL, AMZN, MSFT, META) are short‑term beneficiaries as customers but face margin and cash‑flow strain from near‑term capex (~$600–700bn combined industry spend). China exclusion shifts incremental demand to Western hyperscalers and increases pricing power for NVDA and its networking partners while depressing Chinese OEM suppliers.

Risk assessment: Key tail risks — renewed China market access reversal (either reopening or harsher export controls), rapid emergence of efficient inference accelerators, or supply disruptions at TSMC/Arizona — could swing revenues ±20–40% over 12–24 months. Near‑term (days/weeks) volatility should be driven by IV and guidance cadence; medium (3–12 months) by Vera Rubin production ramp (H2 2026) and hyperscaler capex pacing; long term (2+ years) by competitive entrants and inference commoditisation.

Trade implications: Supply remains tighter than consensus: prioritize asymmetric long exposure to NVDA (calls/call‑spreads) and industrial suppliers of packaging/fabs (TSM) while hedging via small short exposure to hyperscalers whose margins could compress. Cross‑asset: stronger capex implies higher industrial commodities (copper, power) and modest upward pressure on longer‑dated corporate spreads for hyperscalers; expect USD strength on safe‑haven and tech flows.

Contrarian view: The market underweights the networking/IP advantage (NVLink, Spectrum‑X) and supply‑chain de‑risking (US/Mexico fabs) which extend pricing power beyond pure GPU moats. Conversely, consensus may be overconfident on duration — if inference commoditises within 24–36 months NVDA’s multiple could reprice materially. The best asymmetry is time‑boxed option exposure into Vera Rubin (H2 2026) and hyperscaler Q2–Q3 capex cadence rather than an open‑ended long at current multiples.

Key Decisions for Investors

• Establish a 1.5% portfolio position in NVDA using a 6‑month call spread: buy 6‑month 10% OTM calls and sell 6‑month 30% OTM calls. Rationale: asymmetric upside to capture continued training demand while capping premium; target 40–60% ROI if NVDA > +25% in 6 months. Exit if NVDA falls >15% from entry.
• Add a 2.0% portfolio long in TSM (ticker TSM) via shares or 9–12 month 20% OTM calls to play fabs ramp (Arizona capacity). Target +25% in 12 months on continued fab demand; scale in on any >10% pullback; hard stop at −15%.
• Establish a 0.75% portfolio bearish hedge against hyperscaler margin risk: buy a 6‑month put spread on AMZN (buy 10% OTM put, sell 20% OTM put) sized to 0.75% notional. Rationale: protects against capex‑driven margin shock at customers over 3–6 months.
• Implement a short‑volatility income tactic if NVDA IV >40% and price stalls: sell 1‑month iron condors (sell 5%/25% OTM call/put spreads) sized to 0.5% portfolio to collect premium; remove if IV compresses >20% or if price moves >10% against strikes.