Maruwa raised operating profit guidance to ¥38.3 billion on AI optics; Tokuyama's powder line lands in 2028
The Tokyo ceramics maker beat its own half-year forecast on both profit lines and raised full-year guidance, naming AI communications demand first. The aluminum nitride shortage has moved from supply-chain talk to audited numbers, and the new capacity on the books tells you when it ends.
Richard Tang · 3 min read
Maruwa−8.50% — Maruwa, down 8.50 percent today, a Tokyo-listed electronic-ceramics maker worth about ¥655.7 billion, printed record half-year results on October 27 12. Net sales came in 11% above the company's own forecast and operating profit 19% above it, and every reported profit line set a record 12.
Management named next-generation high-speed communications demand first among the reasons, lifted the interim dividend to ¥55 per share from ¥51, and raised full-year guidance to ¥101.5 billion in sales with ¥38.3 billion in operating profit 34. The aluminum nitride bottleneck in AI optics has moved from supply-chain talk to an audited profit and loss statement.
Maruwa's first half beat its own forecast on both lines
- Forecast
- Actual
Data
| Forecast | Actual | |
|---|---|---|
| Net sales | 42,200 | 46,921 |
| Operating income | 14,800 | 17,605 |
The part everyone skipped is now the choke point
An 800G or 1.6T optical transceiver packs more lasers, driver chips and signal processors into less space, and 1.6T-class modules run at 20 to 30W 5. Heat becomes the binding constraint, because a laser drifts and degrades as it warms.
So the laser sits on an aluminum nitride substrate: a ceramic that conducts heat at 170 to 230 W/(m·K), against roughly 24 for ordinary alumina, while also insulating, carrying the fine metal wiring, and holding its shape as the chip heats and cools 5. As data-center modules switched to cheaper non-hermetic packaging, this one ceramic became circuit board, insulator and heat sink at once.
Three independent confirmations in one quarter
The shortage is not Maruwa's claim alone. Zhongci Electronics, a Chinese maker of optical-module ceramics, told institutions on September 17 that aluminum nitride thin-film substrates are supply-short 5. Tokuyama, the Japanese leader in high-purity aluminum nitride powder, is building a second powder base to lift capacity about 30%, with commercial operation set for April 2028 5.
TrendForce puts the AI optical transceiver market near $26 billion this year, and LightCounting expects 800G module shipments to more than double in 2026, with 1.6T rising to the order of tens of millions of ports 5.
MaxLinear−7.05% — MaxLinear, down 7.05 percent today, whose PAM4 signal processors sit in those same modules, reported infrastructure revenue up 210% year on year on October 22 and guided the December quarter up as much as 19% again, with connectivity sales more than tripled and order visibility into 2028 67. We covered MaxLinear's print three weeks ago 8. What is new this week is the bottom of the chain confirming it.
The premium has an expiry date
The skeptics are right that this cannot last as built. Maruwa's new communications building at its Seto plant starts operating in the second half, and Tokuyama's powder line lands in 2028 35. The profit reflects process capability that is genuinely scarce today, but both companies are building the capacity that erases it, so the scarcity premium carries a defined end date.
What is scarce is the process, not the powder
What is genuinely scarce is not nitrogen or aluminum but process capability. Thermal conductivity is highly sensitive to oxygen impurities in the powder lattice, sintering needs temperatures above 1700°C with tight atmosphere control, and the finished substrate then has to pass grinding, polishing and thin-film metallization that hold flatness and surface roughness where the customer certifies them 5. Each step gates yield, which is why capacity moves in years rather than quarters 5.
The H2 test is concrete
The bill for that scarcity lands on optical-module makers competing for the same ceramic. There is a catch on quality too: Simply Wall St flags a high share of non-cash items in Maruwa's reported profit, and three months of choppy trading going into the print 4.
Turning scarce powder into qualified, reliably shipping substrate is exactly the work the reward should follow. The H2 test is concrete: whether Seto's new communications building converts the raised guidance into cash while optics orders hold into 2028. If it does, Maruwa keeps what it earned until Tokuyama's powder line lands. If orders slip first, the skeptics' case strengthens.
Get Chips by email, free
How AI demand is reshaping chipmakers, memory and the fabs behind them. In our daily email, every weekday.
