Exclusive-China’s No. 2 chipmaker readies 7 nm production as Beijing ramps up self-suffiency drive

Hua Hong Group (via Huali Microelectronics) has developed a 7-nanometre chipmaking process and is targeting initial production capacity of a few thousand wafers per month by year-end, a material step toward China-built AI chips. Hua Hong Semiconductor plans to acquire a controlling stake in Huali and raise 7.56 billion yuan (~$1.10bn) to fund upgrades; Huawei and domestic supplier SiCarrier are reported collaborators. While this reduces China’s reliance on foreign AI chip supplies amid eased U.S. export controls, production yields and equipment sources remain unverified, leaving execution risk.

Analysis

The incremental closing of the gap on advanced nodes inside China should be treated as a change in competitive geometry, not an immediate revenue cliff for Western equipment and chip leaders. Domestic process advances lower the marginal value of a single tile of restrictive policy (i.e., one sanction or one shipment delay) because Chinese customers can substitute locally at the margin — that flattens near-term upside for suppliers to Chinese fabs while lengthening the runway for a broader industry bifurcation over 2–5 years.

A key second-order effect is on system-level demand: lower-yield, domestic nodes push Chinese cloud and AI developers to optimize models for different TCO curves (more chips at lower perf per die) which favors cheaper GPUs/accelerators and local software/hardware co‑design. That will increase TAM for mid-tier accelerators but preserve high-end demand where power/efficiency matters — a bifurcated market that sustains premium real estate for best-in-class architectures and puts pressure on the “full-stack” outsourcing model.

Geopolitics remains the dominant asymmetric risk. Policy actions that target tooling or IP transfers can compress scale economics overnight, while sustained investments in domestic equipment suppliers can gradually erode export dependence over several years. The most probable path over 12–24 months is a mix: continued Western sales into China for the highest‑end nodes, alongside faster incremental capacity from local suppliers at trailing-edge advanced nodes — this favors semiconductor designers with global fabs and diversified end markets.

For portfolio construction, treat exposure to AI demand (NVDA) as convex but event-sensitive and exposure to capital equipment (ASML) as monocline with binary policy tail risk. Position sizing should reflect a two‑year view: capture upside from secular AI re‑acceleration while actively hedging political/legal outcomes that would truncate cross‑border flows.

Key Decisions for Investors

• Long NVDA directional with convexity: buy Jan-2027 call spread (size 2–3% NAV). Rationale: captures multi-year datacenter AI spend while limiting premium decay; target 2–4x payoff if cloud AI cycles continue. Hedge: sell shorter-dated calls into rallies and reduce notional by 25% if US export tightening triggers negative headlines (monitor 30‑day realized volatility).
• ASML tactical dip buy with policy hedge: accumulate shares on >8% pullback and buy 6–9 month protective puts (size 1.5–2% NAV). Rationale: secular lithography scarcity supports 20–40% upside in 12–24 months; downside tail (policy embargo) could be -30% to -50%, hence the put. Exit/trim on confirmation of a new export ban or on >40% downside to preserve capital.
• TSMC selective long as defensive foundry exposure: buy Jan-2027 calls or modest stock position (size 1.5–2% NAV) paired with a short China-revenue sensitivity overlay (e.g., out-of-the-money puts on ETFs with high China fab exposure). Rationale: global demand and advanced-node leadership should outperform if yields at domestic competitors remain inferior; risk is secular market-share erosion if domestic yields materially improve within 18 months.