The Most Secretive AI Winner

null

Author | Freddie

Data Support | Gougu Big Data

This year, in the global AI chain sector, not only have the core big players absolutely dominated, but second- and third-tier supporting small manufacturers have also feasted, and even some companies whose businesses seem completely unrelated have been hyped to the moon.

In the Japanese and South Korean stock markets, a group of traditional manufacturing companies have seen their gains leave many AI concept stocks in the dust.

TOTO, known for its sanitary ware, saw its stock price hit a five-year high, rising 145% over the past year.

Driving this revaluation is not its century-old toilet main business.

But rather its semiconductor precision ceramics, which it has been deeply involved in for nearly four decades.

01 Positioning in the AI Track

TOTO's ceramics business began in 1984. That year, the company established a new materials development department, attempting to transfer the high-temperature sintering process accumulated over decades of toilet manufacturing toward industrial precision ceramics. In 1990, it began co-developing etching chamber components with U.S. semiconductor equipment giant Lam Research, stepping into the semiconductor supply chain.

But for the next thirty years, this business segment had an extremely weak presence. High process difficulty, low yield, and low capacity utilization—semiconductor ceramics had been dragging down the group's profits for years. Until five years ago, the profit margin was still only 9%.

The real turning point came in 2020. A new factory in Oita Prefecture introduced fully automated production lines and AI quality inspection systems, significantly improving yield. Then, in late 2022, AI demand began to explode, NAND manufacturers ramped up production wildly, and orders for electrostatic chucks flooded in.

The combination of these two variables completely transformed the ceramics business.

In fiscal 2025, semiconductor ceramics revenue reached 67.4 billion yen, up 34% year-on-year. Operating profit was 28.9 billion yen, up 42%. The profit margin was 43%. The nearly century-old sanitary ware main business had a profit margin of only 5%. The ceramics business accounted for only 9% of total revenue but contributed 54% of operating profit.

TOTO's identity in the capital market had been stable—a building materials stock, a sanitary ware stock. Its P/E ratio had long hovered between 18 and 20 times. During the most狂热 period of the semiconductor cycle in 2021, it briefly touched 39.5 times, but by the end of 2024, it had fallen back to 18.8 times.

The market wasn't ready to price a toilet company as a semiconductor equipment component manufacturer. But four catalysts in 2026 broke through this barrier:

January 22: Goldman Sachs upgraded TOTO from "neutral" to "buy," raising the target price from 4,800 yen to 6,100 yen. The stock rose 11% that day.

February 17: Activist investor Palliser Capital issued an open letter, calling TOTO "the most undervalued AI beneficiary in the market," estimating its intrinsic value at over 8,800 yen.

April 30: Annual report disclosed—EPS of 71.16 yen, 79% higher than market expectations. The stock rose 18% that day, its largest single-day gain in five years.

June 3: Management announced plans to invest 80 billion yen over the next five years to expand semiconductor ceramics capacity, with the semiconductor share of capital expenditure jumping from 11% to over half. The stock rose another 11%.

After these four catalysts, the stock price soared. But by this point, there was a significant divergence in market perception.

Is TOTO a "sanitary ware company with a semiconductor business" or a "semiconductor equipment component company with a sanitary ware business"? These two correspond to completely different valuation multiples.

This judgment is difficult because the position TOTO occupies in the semiconductor chain is truly unique.

The more advanced the chip, the harsher the manufacturing environment. EUV lithography must be done in a vacuum, and temperature fluctuations in each process step cannot exceed micron levels. Traditional mechanical clamps simply can't handle it—only ceramic electrostatic chucks can meet four conditions simultaneously: withstand thousands of degrees of heat, resist strong corrosive plasma, provide ultra-high insulation, and not release gas in a vacuum environment.

3D NAND is stacking from 200 layers to 500 layers. Each additional layer requires an additional low-temperature etch, and each etch requires an electrostatic chuck. As chips go from a single large chip to a multi-chip package, thermal density surges, and ceramics become the only solution.

This logic leads to a counterintuitive conclusion: the more the chip industry pursues "advanced" technology, the deeper its reliance on traditional material processes.

So the question is: what gives TOTO the ability to capture this demand?

While competitors can manufacture aluminum oxide ceramic parts, maintaining high purity, uniform grain size, and precise dimensions during large-scale sintering—a complete set of know-how that only TOTO possesses. From 1995 to 2026, it was the company with the most patent applications for electrostatic chucks globally. Since co-developing chamber components with Lam Research in 1990, the two have been bound for over 35 years, with Lam awarding it the Supplier Excellence Award for two consecutive years.

In terms of capacity, TOTO's Kyushu factory is already at full capacity, while a new firing workshop in Fukuoka is expected to begin production in 2027. The 80 billion yen investment plan announced in June this year exceeded market expectations significantly.

But what truly leaves competitors in the dust is not capacity, but time. Qualifying a new supplier for electrostatic chucks takes at least five years. Even if competitors invest heavily in building factories now, from the start of certification to qualified shipment, it will take at least five years.

The market's hype of TOTO is still ongoing, and the path of shifting its valuation anchor from building materials to semiconductor equipment components is not yet complete.

02 Not Just TOTO

TOTO is not an isolated case. The same logic is playing out in different industries.

Nittobo, a Japanese textile company that has been making glass fiber for 128 years. Its stock price rose 325% last year.

Driving this surge is a product called T-glass, a low-thermal-expansion glass fiber cloth. As AI chip packaging substrates grow larger and stack more layers, the requirements for the coefficient of thermal expansion (CTE) of substrate materials have become extremely tight. Ordinary electronic fabrics can no longer meet the needs of advanced packaging, and T-glass has become the only choice.

Approximately 90% of the global supply of T-glass is concentrated in Nittobo's hands, with capacity already booked through 2027. The supply gap for high-end products exceeds 40%, directly triggering two rounds of price increases—a 20% increase across the board in August 2025, and another 20% to 30% increase in April 2026. The price pressure traveled down the supply chain, with Apple bypassing multiple tiers of channels to directly lock in capacity with Nittobo.

The same identity mismatch occurs with another, more famous Japanese company.

Ajinomoto, the world's largest monosodium glutamate (MSG) producer, leveraged its amino acid chemical technology to develop an insulating film called ABF in the late 1990s, used for interlayer insulation in chip packaging substrates.

Source: Ajinomoto official website

For over two decades, ABF has been the industry default standard, with a global market share of about 80% to 95%. The advanced packaging substrates for AI chips have gone from 8 layers to 16 layers, with each additional layer requiring an additional layer of ABF film. This business segment accounts for only 6% of Ajinomoto's group revenue but contributes 30% of profits, with a profit margin of over 50%.

The surge cases of Nittobo and Ajinomoto both point to the same conclusion: the high-profit-concentration positions in the AI supply chain are not necessarily at the very forefront of technology. They can also be found in seemingly inconspicuous but critical path positions where capacity cannot be quickly expanded.

The same logic is playing out in the A-share market, but the narrative is somewhat different. In A-shares, the story is about the time window opened by the combination of domestic substitution and supply-demand gaps.

Precision Ceramics Direction

The domestic substitution rate for high-end electrostatic chucks in China is less than 1%, with 12-inch products almost entirely dependent on imports. CETC Advanced Materials (Zhongci Electronics) is currently the fastest-progressing domestic company—its electrostatic chucks have passed on-machine verification by leading domestic equipment manufacturers and are entering mass supply; its aluminum nitride thin-film substrates have also begun delivery to customers.

In the first quarter of 2026, the company's revenue grew 79% year-on-year, and net profit attributable to shareholders grew 57%. Its 52-week stock price rose from 45 yuan to 176 yuan, a gain of nearly three times. Following closely are Kema Technology and Xianfeng Precision, but it will still take time for them to achieve scale shipments.

Electronic Fabric Direction

The price of high-end electronic fabrics has cumulatively increased by 250% to 300% since the beginning of 2024, with some extreme models seeing even higher increases. Honghe Technology (Taiwan) is the leader in ultra-thin fabrics (16 microns and below), with a market share of about 26%, and has passed certification from Nvidia and TSMC. In the first quarter of 2026, its single-quarter net profit reached 140 million yuan, a 354% increase year-on-year.

Feilihua (Philippine Hua) is the only domestic company capable of mass-producing quartz fabric, and it has also received Nvidia certification—quartz fabric costs 200 to 400 yuan per meter, with a gross margin exceeding 60%. According to Huatai Securities, the market size for special low-dielectric electronic fabrics (Low-Dk and quartz fabric) is expected to surge from 3.9 billion yuan in 2025 to 29.2 billion yuan in 2027, a compound annual growth rate of 173.3%. This material has become one of the fastest-growing sub-segments in the AI hardware space.

The core tension in the A-share mapping is: the supply-demand gap provides a time window, and the speed of substitution determines elasticity. The real test lies in whether capacity can be released on schedule and whether yields can reach levels comparable to Japanese competitors.

03 Epilogue

The inertia of industry classification is extremely strong. A company that has been making toilets for over a century will not automatically be classified as a tech stock just because its semiconductor business contributes more than half of its profits. The same applies to textile factories, MSG factories, and daily chemical companies—their traditional labels won't automatically fall off.

But changes in profit structure will not wait for market perception to catch up. The difference is only whether the market adjusts gradually in hesitation or jumps all at once when the logic is clear enough.

The structural trend of cross-industry migration will not reverse. AI's demands on chip precision will only increase, and its reliance on traditional material processes will only deepen. But the pace must be清醒—logic takes time to materialize, and stock prices often run ahead of logic. (End of full text)