Nvidia CEO promises to attend Computex Taipei in June

Nvidia CEO Jensen Huang confirmed he will return to Taiwan in June to deliver a keynote at Computex (June 2–5) and said the company will make multiple announcements, while praising Taiwan suppliers after a record year for 2025 (citing Foxconn, TSMC and Wistron). Macronix said it will spend NT$22 billion (US$699.1 million) on capacity expansion—about 11x last year’s NT$1.8 billion—focused on MLC NAND for eMMC, and TrendForce forecasts global server market growth of 12.8% with AI servers accounting for 16.5% and shipments up ~28% to >2.7 million units. Taiwan’s composite economic indicator flashed its first “red” light in a year (index of economic indicators at 38), while separately gold surged near US$5,600/oz, forcing Italian jewellers to reduce gold content amid price pressure.

Analysis

Market structure: Nvidia (NVDA) and Taiwan Semiconductor (TSM) are primary beneficiaries — NVDA as the GPU/AI stack consolidator and TSM as the critical foundry bottleneck — with AI server shipments forecast +28% to ~2.7M this year implying GPU demand could rise 20–40% y/y. Small chip startups, legacy CPU vendors and low-margin jewelers (gold exposure) are losers as pricing power shifts to hyperscaler/accelerator providers. Cross-asset: stronger capex and export-led growth in Taiwan supports TWD and raises front-end rates risk; gold and safe-haven flows remain a tail-hedge if geopolitics worsen.

Risk assessment: Tail risks include tightened US export controls or China market access cuts (low-probability, high-impact), a TSM capacity shortfall delaying NVDA supply ramps, and a hyperscaler demand pause creating inventory build (3–6 month shock). Immediate catalysts are Computex (June 2–5) and NVDA’s keynote (days–weeks); medium-term (3–12 months) risks center on TSM fab build timelines and Macronix capex translation into supply. Hidden dependency: NVDA’s moat is operationally coupled to TSM/Foxconn/Wistron — supply-chain outages amplify revenue risk.

Trade implications: Bias long NVDA into Computex but use option structures to limit drawdown: staggered 2–3% position via 3-month call spreads and a small protective long-dated put (6–9 months) to guard against policy shocks. Add 1–2% exposure to TSM (direct or 12–18 month call LEAP) to capture fab utilization upside. Underweight pure-play legacy CPU and discretionary luxury retail names sensitive to commodity inflation.

Contrarian angles: The market may under-appreciate capacity constraints at TSM — a tighter supply could keep NVDA pricing power intact longer, so TSM leverage may be underpriced. Conversely, consensus underestimates regulatory tail risk; buy asymmetric protection (cheap long-dated OTM puts) rather than naked long size. Historical parallel: 2017 GPU boom turned volatile when capacity caught up; expect similar oscillations and trade around discrete catalysts.

Key Decisions for Investors

• Establish a 2–3% notional long position in NVDA over the next 2–6 weeks via staggered 3-month call spreads (buy calls and sell higher strikes) to capture Computex upside while limiting capital at risk; simultaneously purchase a single 6–9 month protective put (0.25% notional) to hedge regulatory/announcement tail risk.
• Add 1–2% exposure to TSM via shares or a 12–18 month call LEAP (delta ~0.6) to play higher fab utilization and structural AI demand; trim if TSM issues guidance indicating >6 month delay in capacity ramps or if foundry utilization falls below 90%.
• Implement a pair trade for relative value: long NVDA (1.5–2%) and short a smaller AI-chip competitor or lagging CPU vendor (0.75–1%) to capture share consolidation; choose the short where FY+1 revenue growth consensus >20% but gross margin <30%.
• Deploy asymmetric protection: buy 9–12 month OTM puts on NVDA sized 0.5% notional as tail insurance and reduce tech beta by 100–150 bps in duration-sensitive fixed income allocations if Taiwan export leading indicators slip for two consecutive months.