〔Day Digest〕HVLP Copper Demand Jumps 260% to 24,000 Tons, Indium Phosphide Gap Tops 70% as Japan Adds Five Export Curbs and EU Slips in Critical Minerals - Two Bottlenecks, One Substitution Trade

0. Weekly Arc

The AI compute build-out is tightening two specialty-metal bottlenecks in parallel rather than one. Soochow Securities pegs 2026 HVLP copper foil demand for AI servers at 24,000 tons, +260% YoY, with 2027 reaching 50,000 tons ^[1]. At the same time, multiple institutions forecast a 2026 global indium phosphide supply gap above 70%, with Chinese producers shipping only about 10% of global supply ^[2]. Yet supply is structurally constrained on both metals: high-end HVLP capacity sits overseas, and the InP chain shows "strong resources, weak manufacturing" ^[1][2]. Japan tightened five more tech items, while the FT reports the EU continues to lose ground in the critical-minerals race ^[3][4]. The trade is long domestic substitution, short import dependence.

1. AI Compute Pulls Specialty Metals

• **[NEW] HVLP copper foil, Soochow Securities:** 2026 global AI-server dedicated HVLP demand at 24,000 tons, +260% YoY; 2027 at 50,000 tons ^[1]. Beyond core AI compute, demand from high-end industrial control, precision instruments, automotive high-end electronics, and high-speed communication "provides sustained support" for growth ^[1].
• **[NEW] Indium phosphide, multiple institutions:** 2026 global InP supply-demand gap expected to exceed 70%; Chinese makers' shipments only ~10% of global supply ^[2]. Driver: AI computing-power buildout from H2 2025, with tension now propagating up and down the chain ^[2].

2. Supply Bottlenecks and Domestic Substitution

• **[NEW] HVLP, structural:** capacity is "highly concentrated in overseas manufacturers"; HVLP high-end process barriers are high, capacity ramps slowly, and ordinary copper-foil lines cannot switch to HVLP ^[1]. Domestic third-generation-and-above HVLP is "highly dependent on imports," leaving "broad domestic substitution space" ^[1]. Top Chinese firms are expanding via line upgrades and new high-end lines ^[1].
• **[NEW] InP, structural:** upstream refined indium is sufficient but high-purity indium capacity is insufficient; midstream substrate and epitaxial wafer are still ramping; high-speed optical chipmakers "generally depend on imported substrates" ^[2]. Localization is accelerating but the chain remains small versus global demand ^[2].
• **Source caveat:** the InP bottleneck detail comes from a single Science and Technology Innovation Board Daily interview round and the headline numbers are forecasts, not reported actuals ^[2].

3. Policy Curbs: Japan and the EU

• **[ESCALATED] Japan, August 16:** five items added to the technology-management system - solder resist, GaN substrates, permanent magnets, perovskite cells, and scintillators - on the stated rationale of diversion risk to military end-uses ^[3]. The mechanism dates to a 2024 "public-private dialogue on strengthening technology management" and is now widening beyond high-end machine tools to "seemingly basic" inputs ^[3].
• **[NEW] EU critical minerals, Financial Times:** Washington and Beijing are beating Brussels in competition to secure components for defense and green technologies, with the EU "slipping further behind" ^[4]. The piece does not quantify the gap, only the direction ^[4].

4. Contradictions and Falsifiable Tests

• Demand and supply do not match on either metal: HVLP demand grows 260% against a supply base that cannot be retooled from ordinary copper-foil lines, and InP demand runs against an industry where domestic share is ~10% and high-purity indium is the binding constraint ^[1][2]. The Japan control list now covers five more inputs while the EU concedes it is losing the procurement race ^[3][4]. The falsifiable test for the substitution thesis is late-2026 domestic third-generation HVLP output and first InP substrate output at scale, neither of which appears in current reporting.
• Source quality: HVLP and InP supply figures originate with Soochow Securities and unnamed "multiple institutions," not with the producers themselves ^[1][2]; the EU-versus-US/China critical-minerals gap is asserted in a single FT headline item without underlying figures ^[4]. Treat the 260% and 70% numbers as forecasts, not prints.