Skip the Subsidy Story: Six Japanese Chip Stocks That Win Without Tokyo's Help

By Kenichi Ogura - Former Chief Editor of “PRESIDENT” (Japan’s largest business magazine)

Everyone wants to talk about the subsidies. I'd rather talk about what happens after they stop. A subsidy is a line item, not a structural advantage - it survives until a budget committee decides it doesn't. I analyzed six Japanese chip companies that would keep winning even if the checks from Tokyo never arrived, and three that would not. One of those three is worth walking through in detail, because it shows what happens when a subsidy and a cyclical downturn hit the same company in the same year.

World semiconductor demand is expected to grow 26.3% in 2026, to roughly 151 trillion yen (approx. $932.1 billion at 162 yen/$1). Japan's share of the finished-chip market is only about 5%, a number that gets repeated constantly and understood rarely. Japan doesn't need to make the chip. It makes the machine that etches it, the wafer it sits on, and the chemical that develops the pattern on top - and once you control those, the 5% figure stops mattering nearly as much as people assume.

Tokyo Electron Limited<8035>

Tokyo Electron ranks third worldwide in chipmaking equipment, and it's the only supplier making tools for all four core process steps - etching, deposition, coating and developing, and cleaning. Its coater/developer share runs near 89-90%, and for the EUV version specifically, close enough to 100% that rivals barely register. None of that traces back to government money. It comes from the fact that once a foundry qualifies a Tokyo Electron tool, engineers spend the better part of a year tuning the recipe to hit yield, and tearing that out means starting the qualification over. Buying several sequential steps from one vendor makes the whole line easier to run - which is a large part of why TSMC and Samsung keep coming back.

Fiscal 2026 net profit rose 5.6% to a record 574.4 billion yen (approx. $3.5 billion), even as operating profit dipped 10.4% on heavier R&D spending. First-half fiscal 2027 guidance points to revenue up 33.1% as the AI investment cycle turns.

Here's what could go wrong: China made up about 34% of fiscal 2026 revenue, and there was a quarter last year where cooling Chinese demand knocked revenue down 15.7% in three months. Export controls could still reach a China business worth close to 1 trillion yen a year, though management's own guidance assumes China holds roughly flat through 2026 with AI demand elsewhere filling the gap. Still attractive, and the lock-in alone is most of the reason why.

ADVANTEST CORPORATION <6857>

ADVANTEST makes the equipment that tests whether a finished chip works, and it just took real share from the one company that seriously competes with it. Its share of the SoC tester market rose from 56% in 2024 to 66% in 2025 - a ten-point move, in a two-player market, in a single year. Once a customer qualifies the V93000 platform for something like NVIDIA's H100 or B200, switching away takes eighteen months or more and costs in the hundreds of millions of yen. AI accelerator and HBM testing simply runs longer and costs more per chip than legacy work, and Advantest also landed its first production order for silicon-photonics test equipment - a market it opened on its own, not one anybody handed it.

ROE came in at 47.2% for fiscal 2025, which I'd call unusual for an industrial company anywhere, and revenue grew 44.7% while operating profit more than doubled. A fab-light model - most manufacturing outsourced - keeps fixed costs low and lets profit scale fast as volume rises.

Teradyne still holds roughly 34% of the market and won't sit still forever. There's China exposure too, though about 98% of revenue comes from outside Japan, spread across TSMC, SK Hynix, and Samsung sites in different countries, which blunts any single-market shock. I'd happily own this one.

Shin-Etsu Chemical Co., Ltd.<4063>

Shin-Etsu controls an estimated 42.49% of the global silicon wafer market - above 60% combined with SUMCO - and few companies anywhere can mass-produce the 300mm wafers that sub-5nm chips need. That kind of process took twenty-five years to build, and no amount of Chinese state investment shortens the timeline.

The wafer business isn't the whole story, though. Shin-Etsu also owns Shintech, its US subsidiary and the world's largest vinyl chloride producer, and vinyl chloride runs on a cycle that has nothing to do with semiconductors. When one slows, the other tends to hold up - in 2024-2025 it was vinyl that dragged the company down on soft US construction demand, part of why ROE landed at just 10.6%, the lowest of the six names here. I read that less as a weakness than as the diversification doing its job in a direction nobody happened to be watching.

The equity ratio sits at 78.7%, with almost no interest-bearing debt, and that balance sheet is really the whole case: it lets Shin-Etsu keep investing through a downturn while smaller rivals pull back. Watch US tariff policy and an anti-dumping probe into Chinese dichlorosilane exports on the vinyl side. Worth owning, mostly on the strength of that balance sheet rather than the wafer share alone.

DISCO CORPORATION <6146>

DISCO holds 70-80% of the global market in each of three separate steps - dicing, grinding, and polishing - and more than fifty years of accumulated technique gives it sub-micron precision that advanced packaging below 7nm more or less requires.

The interesting part isn't the machines. It's what happens after you buy one: a dicing saw sells for 200-500 million yen, and every saw then needs proprietary blades and consumables that only DISCO makes. That recurring business now runs about 32% of revenue, and it's a big part of why the operating margin came in at 42.2% last year - a number I'd expect from a software company rather than a precision toolmaker. Requalifying a competitor's tool takes eighteen months or more and can cost over 1 billion yen, so almost nobody bothers trying - really the same advantage as the machines themselves, just showing up twice, once at the point of sale and again every time a wafer gets cut.

The equity ratio climbed to 78.9% and net debt is zero. China has historically made up 25-35% of revenue, and Chinese rivals are already undercutting DISCO on mature-node dicing, though not below 7nm. Further out, laser dicing and plasma etching could chip away at the consumables model - a real risk, just not an urgent one.

SCREEN Holdings Co., Ltd. <7735>

SCREEN is the one name here that's gaining rather than just defending. Its share of single-wafer cleaning equipment went from 34% in 2023 to 42% in 2024, at a moment when cleaning steps are multiplying anyway as HBM and 3D packaging both demand more of them. Growing faster than an already-fast market is harder than it sounds.

Fiscal 2026 revenue beat the company's own plan, and the fourth-quarter semiconductor segment posted a 30% operating margin. December 2025's launch of the DW-3100 - a direct-write system for advanced packaging - plus ongoing work with IBM and imec, suggests SCREEN isn't leaning on any one customer or program to stay relevant into the next process generation. TSMC alone accounts for roughly 14% of revenue.

If there's one thing that could derail this story, it's concentration risk stacked on top of itself: China made up 39% of revenue as of the fiscal 2026 third quarter, the highest of any name here, and the display equipment business, roughly a fifth of revenue, runs on OLED capital cycles that add volatility the core business wouldn't otherwise carry. Still a buy, but size the position knowing it carries the widest risk band on this list.

TOKYO OHKA KOGYO Co., Ltd.<4186>

TOKYO OHKA KOGYO made Japan's first domestically produced photoresist in 1968, and it still holds the largest overall photoresist share in the world, at 24.7%. In EUV resist specifically, its share is about 28%, putting it alongside Shin-Etsu and the now-nationalized JSR as one of three firms that between them control most of the world's supply.

It isn't a patent wall that keeps competitors out so much as a qualification wall. TOK formulates custom photoresist chemistry tuned to each customer's production line, and switching means requalifying that line - six to eighteen months of yield risk most fabs won't accept over a chemical input. In 2023 the company sold its entire equipment business to put everything into photoresist alone, which struck me as an unusually confident bet at the time, and the results since have mostly backed it up: EUV resist revenue alone grew 50% last year, and the dividend rose for an eighth consecutive year.

The equity ratio sits at 71.1%. The weak spot is regulatory rather than competitive - tightening PFAS rules in Europe and the US could raise development costs, since some EUV resist formulations rely on fluorinated polymers. China exposure barely matters here, since EUV resist mostly serves leading-edge fabs outside current export-control reach. Worth owning.

What to avoid

SUMCO CORPORATION <3436> is the clearest lesson in what a subsidy doesn't fix. METI is funding part of its new Saga plant with up to 75.0 billion yen (approx. $463 million), and that funding did nothing to stop the numbers underneath it from falling apart. The January-March 2026 quarter posted an ordinary loss of about 8.0 billion yen (approx. $49.4 million), and the full-year outlook now points to an operating loss.

Two things are happening at once, and they're making each other worse. The wafer market has split - 300mm demand tied to AI is strong, while smaller-diameter wafers for cars and industrial equipment sit in a slump with no clear end. Chinese producers, some of them state-backed, are undercutting SUMCO on price in that exact segment, and operating margin has gone from 24.9% in 2022 to an outright loss. The dividend has been cut three years running, from 81 yen to 20 yen.

SUMCO is closing small-diameter production at its Miyazaki plant by the end of 2026, which is the right call, but unlike Shin-Etsu it has no second business to lean on when silicon turns down. It's a pure wafer play riding the cycle with nothing underneath it, and the irony is hard to miss: the government paid for a factory, and waiting on the other side of it is a Chinese competitor whose own subsidy did the same job more cheaply. Skip this one.

Although they are unlisted, Rapidus and JSR don't make the list either, for a simpler reason. Rapidus had received over 2,354.0 billion yen (approx. $14.5 billion) in cumulative state support as of April 2026, against only 167.6 billion yen (approx. $1.0 billion) in private capital, with roughly 5 trillion yen (approx. $30.9 billion) still needed to hit its 2027 target. JSR was taken private via a roughly 1 trillion yen (approx. $6.2 billion) tender offer by a government-backed fund - nationalization, in effect. Neither company's future is being decided by customers anymore, which puts them outside anything I'd call a merit-based thesis.

China exposure runs through the winners too - 34% for Tokyo Electron, 39% for SCREEN, 25-35% for Disco. In June 2026, China's Ministry of Commerce added twenty Japanese companies to its dual-use export control list, reaching well past chip equipment into broader industrial trade. I'd still argue the same list works in these companies' favor more than against them: you tend to land on an export control list because you make something nobody else can, and if Chinese demand does shrink, AI-driven demand from TSMC, SK Hynix, Samsung, and Intel is already there to take its place.

Add it up and the 151 trillion yen semiconductor market won't lift every Japanese chip stock the same amount this year. It lifts the ones that built something a subsidy never created and an export ban can't easily undo - decades of process knowledge, switching costs running into the hundreds of millions of yen, and balance sheets strong enough to keep spending when everyone else pulls back. Tokyo Electron, ADVANTEST, and DISCO sit on the equipment side, catching the AI cycle directly. Shin-Etsu and TOKYO OHKA KOGYO sit upstream in materials, where a strong balance sheet widens the gap during every downturn. SCREEN is the outlier, taking share it didn't already have.

I'd rather hold six companies that made it through Japan's past semiconductor winters without help than one that needed the help to survive at all. Fiscal discipline in Tokyo won't hold forever on its own, the Bank of Japan's independence remains an open question, and industrial policy that hand-picks winners tends to drift toward the kind of favoritism markets are supposed to prevent. None of that touches the six names above - but it's already touching SUMCO, and it explains most of what's gone wrong at Rapidus and JSR.

*The content of this article is based on the author's personal views and analysis, and is not intended as a recommendation or solicitation to buy or sell any specific stocks. Please make any final investment decisions at your own discretion and risk.