Jensen Huang Believes that NVIDIA is TSMC's Premiere Customer

NVIDIA CEO Jensen Huang said NVIDIA is now TSMC's largest customer, and industry chatter claims NVIDIA accounts for nearly 13% of TSMC revenue as foundry capacity is reallocated toward AI workloads. Separate reports allege TSMC has pressured Apple for sizeable price increases and may have deprioritized Apple production even as Apple purportedly holds ~50% of 2 nm capacity for upcoming A20/A20 Pro chips; Apple is also reportedly developing an inference-only 'Baltra' GPU targeted at TSMC 3 nm (N3E/N3P). The developments, driven by demand for AI datacenter silicon, are rumor-driven but could influence competitive positioning and wafer-allocation dynamics across NVIDIA, Apple and TSMC.

Analysis

Market structure: If true, NVIDIA (NVDA) gains incremental pricing power and throughput for AI datacenter ASICs at TSMC (TSM) where NVDA now reportedly ~13% of revenue; expect NVDA to capture a larger share of early 2nm wafers for datacenter GPUs and accelerators over the next 12–24 months, squeezing rivals’ access. Apple (AAPL) being deprioritized on bleeding-edge capacity would pressure its A-series/AI roadmap cadence and could delay A20 launches by 1–2 quarters or force migration to N3E/N3P, increasing unit costs by an estimated 3–8% on advanced nodes. Semiconductor equipment suppliers and OSATs (AMAT, KLAC, ASX/ASE) are secondary beneficiaries from higher TSMC mix toward high-margin 2nm products.

Risk assessment: Tail risks include a TSMC operational outage (fire/contamination) that would spike GPU spot prices and NVDA inventory shortfalls within days, or regulatory intervention (US/ROC export controls or antitrust) within 3–12 months that could curtail wafer allocations. Hidden dependencies: NVDA still needs advanced packaging (TSMC + ASE) and HBM supply — HBM bottlenecks could cap revenue even with wafer priority. Catalysts to watch in 30–90 days: TSMC Q1 guide, NVDA datacenter order flow commentary, and Apple WWDC / A20 leaks.

Trade implications: Tactical: establish a 2–3% long NVDA equity position and stagger buys over 2–6 weeks; complement with 3–6 month NVDA call spreads (buy 1–3 month ATM call, sell 3–6 month higher strike) if NVDA implied vol >50% to cap cost. Pair trade: long NVDA, short AAPL (0.8:1 notional) sized to risk — target 20% relative outperformance in 6–12 months, stop if NVDA underperforms AAPL by 10% in 30 days. Consider a 1–2% long TSM position for secular foundry upside but hedge with out-of-the-money put protection if TSMC capex guide disappoints.

Contrarian angles: Consensus may underplay Apple’s bargaining power — AAPL could absorb higher wafer prices or shift some designs to N3E without meaningful consumer impact, making an outright short AAPL risky. NVDA concentration at TSMC raises counterparty risk: if TSMC pushes prices up 5–10% across the board, NVDA gross margins could be compressed if end customers resist price passthrough. Historical parallel: prior node squeezes (7nm) produced intense spot premiums for GPU SKUs but also accelerated competitor node investments; expect competitor capex announcements within 9–18 months that could normalize pricing.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (ticker NVDA) over 2–6 weeks, target 20–40% upside in 6–12 months; hedge half the notional with 3–6 month call spreads (buy ATM, sell +15–25% strike) if IV >50% to limit premium.
• Initiate a tactical pair trade: long NVDA (size X), short AAPL (size 0.8X) to express foundry priority shift; set relative profit target 20% and hard stop if NVDA underperforms AAPL by 10% within 30 trading days.
• Buy a small, hedged TSM position (1–2% of portfolio) to capture secular foundry upside; protect with 6–12 month OTM puts (5–7% delta) sized to limit downside to ~3% portfolio impact if TSMC guidance weakens at next quarterly report.
• If seeking asymmetric downside exposure to AAPL, purchase a 6–9 month put spread (buy 1 ITM/ATM put, sell 1 lower strike put) sizing to 1–2% portfolio — enter if market implies AAPL downside >10% or if supply-news confirms wafer deprioritization within 60 days.