Substrates/CCL +20-30%, 300mm wafers expanding, Marumae net profit +21% — pricing power is quietly flowing back upstream behind the 'Japan loses' headline
Two Headlines, One Truth
On October 8, 2026, Nikkei Business ran two nearly identical pieces: "Japan falters as AI boom drives DRAM prices to historic highs" and "Japan loses chip race as AI boom triples DRAM prices" (日経ビジネス). The same day, Samsung Electronics reported Q3 operating profit surging roughly 8.8x year-over-year to a record (NHK, 読売新聞). The narrative was clean: AI memory cycle dollars go to Korea. Japan watches its consumer electronics get expensive.
But directly beneath that narrative, inside the same 72-hour window, three quiet prints pointed in exactly the opposite direction. Maeil Business Japan reported that substrate and copper-clad laminate (CCL) prices had jumped 20-30%, expanding the market influence of Japanese and Taiwanese materials and equipment suppliers (매일경제). Nikkei reported that semiconductor equipment components supplier Marumae's FY2027 net profit rose 21% on strong sales to chip-equipment makers (日本経済新聞). And Dream News reported that Japan's 300mm silicon wafer market is expanding on AI demand (ドリームニュース).
Three prints. One pattern. Wherever the memory dollars flow, they have to pass through someone's substrate, someone's photoresist, someone's wafer. Right now that someone is still largely Japan.
Anatomy of the Repricing
Substrates/CCL +20-30% is not a cycle-peak spike. It is the structural result of empty inventories, new capacity additions pushed out to H2 2027, and the fact that every CoWoS and FO-PLP line requires ABF substrate or high-layer-count CCL. The October 8 news that Nanya New Materials broke ground on an AI-server-focused CCL factory in Jiangxi (note) confirms the pressure is pulling new investment even into China. But new capacity takes 18-24 months to ramp, and in the meantime, the pricing flows to the Japanese materials chain: Mitsubishi Gas Chemical, Ajinomoto Fine Techno (ABF film), Shin-Etsu Chemical (silicon wafers), Tokyo Ohka Kogyo (photoresist).
Marumae +21% is a different axis. The company supplies precision cleaning and polishing parts for semiconductor manufacturing equipment. Which means every time TEL, Lasertec, SCREEN sells a tool, Marumae's revenue follows — it is a second derivative of the equipment cycle. FY2027 +21% signals that the equipment cycle is not past peak and sinking; it is still accelerating. The October 9 IMAPS 2026 conference focusing on advanced packaging technology for AI semiconductors (JETRO) confirms where that demand terminates.
And the 300mm wafer expansion is the slowest but thickest layer. Shin-Etsu Chemical (4063) and SUMCO together hold roughly half of global 300mm supply. AI-driven 300mm demand is now fully offsetting the 2024-25 legacy digital softness in 2026-27 — and this is not a capex cycle, it is an opex (consumable) cycle. It repeats every quarter.
Positioning: The Floor Nikkei Isn't Looking At
Rohm's (6963) October 8 upward FY2027 guidance revision (Reuters) is another confirmation point from the materials and power-semi side that demand remains strong. Advantest (6857), by contrast, fell 1.9% on US semiconductor weakness (日本経済新聞) — showing how volatile the test-equipment names remain. Nomura strategists warned on October 9 that the AI/semiconductor rally may consolidate as valuations compress (野村證券).
That warning is correct at the index level. But below the sector, in the materials layer, the direction is opposite. What compresses is the multiple, not the pricing power. CCL +30% is ASP, not multiple. 300mm wafer expansion is volume, not beta. Marumae +21% is P&L, not narrative.
With DDR5 16Gb spot sitting at $58.9 on October 11, memory peak-cycle calls are getting louder. But whether those calls are right or wrong, the materials layer runs on a different clock. Substrates are capacity-short through H2 2027. 300mm wafers do not face ASP compression until the AI mix shift completes. And every time TEL, Lasertec, SCREEN ship a tool, Marumae and its peers' second derivative stays positive.
The Trade
Shin-Etsu Chemical (4063) is the center position. 300mm wafer demand, photoresist demand, and a diversified non-semi book (PVC, silicones) that provides ballast — but the core of the 2026-27 repricing cycle is in semi materials. Tokyo Ohka Kogyo (4186) offers more concentrated exposure to EUV photoresist. Rohm (6963) is confirmed by its guide-up in power semis. DISCO (6146) and SCREEN (7735) are the first-derivative equipment names of which Marumae is a second derivative.
What to avoid: short positions based on the index-level "Japan loses" narrative. That narrative is right about where the DRAM dollars go, but it misses the materials layer those dollars have to pass through. The moment Nikkei declares defeat, the prices of the substrates, wafers and photoresists that enable the defeat quietly go up 20-30%. That is the real print of Japan's semiconductor chain in the second week of October 2026.
Key Sources: - Substrate, copper-clad laminate prices surge; Japanese, Taiwanese suppliers gain (매일경제, 2026-10-09) - Marumae Net Profit Rises 21% as Semiconductor Equipment Demand Surges (日本経済新聞, 2026-10-09) - Japanese silicon wafer market expands on AI demand, 300mm growth (ドリームニュース, 2026-10-09) - Rohm Raises FY2027 Earnings Forecast Amid Strong Semiconductor Demand (Reuters, 2026-10-08) - Japan Loses Chip Race as AI Boom Triples DRAM Prices (日経ビジネス, 2026-10-08) - plus 26 more
If this analysis was helpful · ☕ Support Us · ✈️ Telegram