SUMCO

The trough is in and the shortage setup is real: major wafer makers cancelled or deferred greenfield 300mm capacity during the 2024-2025 downturn while AI logic and HBM fab buildouts accelerate; wafers are the one semiconductor input you cannot double-order your way out of. SUMCO exits the trough with the industry's newest capacity, TSMC/Samsung/SK hynix/Micron qualifications, mid-20s EBITDA margins even while loss-making, and capex already down from 214.9B yen to 79.9B yen -- a violent FCF and EPS inflection if 300mm tightens in 2027.

• 300mm supply/demand: the top-5 makers control ~90% of supply and deferred expansion industry-wide; management sees AI DRAM wafer consumption roughly tripling to ~1.5M wafers/month over 3-4 years, the fastest-growing wafer end-demand
• Operating leverage: ~30B yen quarterly depreciation on a mostly-new asset base means incremental wafer volume drops through at very high margin; EBITDA margin printed 27.5% in Q4 FY2025 and 23.1% in Q1 FY2026 near the bottom of the cycle
• Capex cliff: 214.9B yen (FY2024) to 79.9B yen (FY2025) to a 9.4B yen Q1 FY2026 run-rate cut the FCF deficit from ~-178B yen to ~-11B yen in one year -- FCF flips positive into the recovery
• LTAs with prepayments de-risk the leading-edge book and give pricing power when spot tightens
• Strategic scarcity: only ~5 companies globally can make leading-edge 300mm epi wafers; Japan's government treats SUMCO as critical infrastructure (economic-security subsidies)
• A 20 yen dividend maintained near trough (only trimmed from 21 yen) despite the net loss signals balance-sheet confidence, backed by a JCR A rating affirmed January 2026

A net-loss-making, indebted pure cyclical that has already almost quadrupled on a shortage that is still a forecast -- one management itself declined to call imminent in February 2026. The stock trades near a $10B market cap (~4x sales, no earnings) while the company guides continued net losses through H1 2026, non-AI demand (the volume majority) stays weak, customers sit on wafer inventory, and Chinese entrants structurally erode the legacy business. If the 2027 shortage slips or AI capex cools, there is no earnings floor under the multiple.

• Fundamentals vs price: +278% stock move in 12 months against FY2025 operating profit of 1.3B yen, an FY2025 net loss of 11.8B yen, and a guided H1 FY2026 net loss of 15.4B yen; the recovery is fully in the price before it appears in the P&L
• Q1 FY2026 missed street EPS consensus and the stock hit daily limit-down on the H1 loss forecast -- evidence the market is priced for perfection
• Management's own February 2026 call said no imminent 300mm shortage, with meaningful wafer-capacity additions not landing until late 2026/early 2027 -- the July 2026 shortage-headline rally runs ahead of the company's own framing
• Non-AI demand (auto, industrial, consumer) is the volume majority and remains in prolonged inventory digestion that management expects to take through 2026 to normalize
• China wafer localization is structural, not cyclical: legacy 200mm/300mm pricing may never recover to prior peaks, and subsidized Chinese capacity keeps coming
• Subsidy cut (75B yen approved -> 19.3B yen) and shelved greenfield fabs mean SUMCO has less low-cost optionality to capture the very shortage the bull case needs -- restarting greenfield takes 2-3 years; gross DER of 0.62x plus heavy D&A leaves thin room for error if the downturn extends into 2027, and yen appreciation is an additional uncontrolled risk