CMB International: Optimistic about the rapid increase in AI PC penetration rates during 2026-27; companies such as Foxlink Precision (06088) are expected to benefit.

CMBI's research note from CES 2026 highlights that next-generation AI server and AI PC initiatives are progressing well: Nvidia's Vera Rubin platform is ahead of schedule with full-scale production underway and capacity set to expand in H2 2026, while AMD's MI400/Helios upgrades are proceeding smoothly. Qualcomm, Intel and AMD launched new AI PC processors capable of supporting large models, prompting CMBI to forecast rapid AI PC penetration gains in 2026–2027 and identifying supply-chain winners including Honfun Precision and Luxshare (interconnects), BYD Electronics (liquid cooling), Foxconn Industrial Internet (AI server ODM) and Lenovo (AI PCs).

Analysis

Market structure: NVDA's Vera Rubin early production and H2 2026 capacity expansion crystallize a two-tier market — hyperscaler GPUs (NVDA) and diversified accelerators (AMD, QCOM, INTC for PCs). Expect NVDA to retain pricing power in servers through 2026 (20–30% gross margin premium vs peers) while AMD and Intel/Qualcomm fight for edge/PC share, boosting upstream suppliers (interconnects, liquid cooling, ODMs) where pricing and lead times will tighten.

Risk assessment: Tail risks include expanded US/China export controls, wafer/yield setbacks in new nodes, or enterprise AI budget reallocation; each could shave 20–40% off consensus revenues for affected suppliers in 12 months. Immediate watch items (days–weeks): CES order announcements and commentary; short-term (3–9 months): shipment cadence and HBM/interconnect lead times; long-term (12–24 months): actual AI PC penetration (forecast 15–25% by end-2027) and NVDA margin sustainability.

Trade implications: Favor concentrated long exposure to NVDA around the H2 2026 capacity ramp (6–12 month horizon), tactical longs in AMD for MI400 tailwinds, and selective long positions in Luxshare/Honfun/BYD Electronics for interconnect/liquid-cooling beta. Use pair trades (long NVDA, short INTC) to express server vs legacy CPU divergence; implement 6–12 month call spreads to limit premium spend and sell covered calls on mature OEMs to harvest elevated IV.

Contrarian angles: Consensus underprices supply-chain bottlenecks (HBM, liquid cooling) that could compress OEM margins — a risk to Lenovo/ODM profitability despite unit growth. Conversely, AI PC adoption may be overhyped near-term; a 10–20% pullback in AI PC demand would disproportionately hurt lower-margin suppliers, creating short opportunities and selective supplier buys at discounted multiples.

Key Decisions for Investors

• Establish a 2–3% long position in NVDA (NVDA) via a staggered buy over 5–10% pullbacks, add a 6–12 month bull-call spread (debit) to capture H2 2026 volume ramp while capping premium risk; take profits if NVDA outperforms by +30% or margins fall >200bp sequentially.
• Initiate a 1.5–2% long in AMD (AMD) to play MI400/Helios server upgrades; complement with 3–6 month out-of-the-money calls (20–25% delta) before quarterly shipment updates, trim on signs of HBM shortages or missed OEM wins.
• Put on a pair trade: long NVDA (1.5%) vs short INTC (1%) to express GPU/server share gains over legacy CPUs; use equal notional sizing and stop-loss if NVDA/INTC spread narrows <10% on a trailing 30-day basis.
• Allocate small tactical stakes (0.5–1% each) to supplier plays: long Luxshare (interconnect exposure), Honfun (interconnect), BYD Electronics (liquid cooling) — accumulate on >15% pullbacks and exit on order-book updates showing <6 months of forward visibility.
• Monitor within 30–60 days: US export-control announcements, HBM lead-time data, and NVDA/AMD server order disclosures; if export controls expand or HBM lead times lengthen >20% vs current levels, reduce cyclicals/ODM exposure by 50% and rotate into software/service providers benefiting from on-premise to cloud AI shifts.