Sumitomo Bakelite raises 2026 profit forecast on AI packaging boom

Sumitomo Bakelite's 57.9% operating profit surge is not a cyclical blip—it is the financial fingerprint of advanced packaging shifting from a process discipline to a materials bottleneck. CoWoS and HBM stacking push material layers per die from 3-4 to 10+, causing EMC, underfill, and TIM consumption to compound exponentially. The company monetizes layers, not volume. Upstream, this compresses margins for legacy 2D packaging material suppliers. Namics retains a first-mover edge in low-CTE epoxy for HBM, but Sumitomo Bakelite's positioning in high-thermal-conductivity TIM and thick-film underfill for AI servers is more surgically targeted. The critical signal: materials-side profit growth dwarfs the OSAT segment (ASE, JCET at roughly 15-20%), confirming an irreversible value migration from assembly to formulation. Risk vector: OSAT houses in Taiwan, China and mainland China are accelerating domestic EMC adoption through 2025-2026. A price war in mid-tier packaging resins will surface within 12 months. However, HBM and CoWoS-grade materials remain locked by three players—Sumitomo Bakelite, Namics, and Entegris—with no credible near-term substitute. 12-24 month outlook: AI inference chips will demand larger interposer footprints than training chips. Low-warpage EMC consumption per unit runs 4-6x traditional BGA. The 35 billion yen full-year target is likely conservative. The real wildcard is the HBM4 volume ramp in H2 2026, where material ASP could climb another 15-20%.