ROHM Co., Ltd.: The Kyoto Chipmaker's All-In Bet on Silicon Carbide

I. Introduction & Episode Roadmap

On May 13, 2026, in a conference room in Kyoto, ROHM's president Katsumi Azuma presented a set of numbers that told two completely different stories at the same time.

The first story was recovery. Revenue for the fiscal year just ended had climbed 7.3% to ¥481.1 billion. Operating profit, which had been a ¥40.1 billion loss the year before, had swung positive to ¥10.9 billion. Inventories were coming down. Automotive demand had stabilized. Orders for silicon carbide traction inverters — the product the entire company had been rebuilt around — were growing.

The second story was written one line further down the income statement. Net loss: ¥158.4 billion. The largest annual loss in ROHM's sixty-eight-year history, driven almost entirely by a ¥193.6 billion impairment charge taken against production facilities tied, above all, to the silicon carbide business.¹²

That is the paradox at the centre of this story. ROHM's operating business is healing. Its balance sheet just formally acknowledged that the strategic bet the operating business was built to serve — a two-decade, crystal-to-module vertical integration in silicon carbide, one of the most complete such positions anywhere outside Wolfspeed — cost far more than it will earn on the timeline management assumed.

ROHM is not a household name. It is a roughly ¥480–510 billion revenue analog and discrete semiconductor maker headquartered in Kyoto, and its parts are inside an enormous share of the cars, phone chargers, industrial motor drives, printers and — increasingly — AI server racks sold today. It sells resistors, transistors, diodes, LED drivers, motor drivers, power management ICs, laser diodes and, at the sharp end, silicon carbide MOSFETs.

And as of this writing, it is at the centre of the most consequential corporate restructuring in Japanese semiconductors in a decade. In March 2026, デンソー Denso — Japan's largest auto-parts maker and already a ROHM shareholder — put a takeover proposal on the table reported at roughly ¥1.3 trillion, or around $8–9 billion.³ ROHM's board convened a special committee of independent outside directors, deliberated for six weeks, and did not support it. Denso withdrew on April 28, 2026.⁴ In parallel, ROHM had already signed a memorandum of understanding with 三菱電機 Mitsubishi Electric and Toshiba Electronic Devices & Storage, a subsidiary of 東芝 Toshiba, to explore combining their power-semiconductor businesses⁵ — a combination that would create, on Omdia's numbers, roughly 11% of the global power-device market against Infineon's 24%, and the world's second-largest power-chip supplier.⁶

Those talks are still open. They have not closed. That distinction matters enormously to anyone trying to value this company today.

Here is the route. We start in a Kyoto workshop in the 1950s, where a student with a utility-model patent on a small resistor built a components empire on an obsessive belief in reasonable profit. We trace how that boring, cash-generative base funded a move into ICs, sensors and automotive discretes. We then follow the silicon carbide bet from its 2009 origins to the moment it consumed the company's balance sheet. We examine the less-discussed capital-allocation decision that ran alongside it — ¥300 billion committed to Toshiba's privatization. We work through the two loss years, the CEO change, and the credit downgrade in forensic detail, including one accounting change that deserves far more attention than it has received. We look at which segments actually earn money. We war-game the competitive landscape, where Chinese capacity has done more damage to ROHM's thesis than any Western rival. And we end on the 2026 endgame, current management, and the specific evidence that would confirm or break the case from here.

↑ Back to Top

II. Kyoto Origins: From One Resistor to a Components Empire (1958–1990s)

The origin story begins with boredom.

Kenichiro Sato — known inside the company as Ken Sato — took a part-time job repairing radios while a university student in Kyoto. He found the repair work dull. What interested him was making the parts. So he set out to develop a resistor, on the reasoning that resistors were indispensable to every vacuum-tube radio then in existence. In 1954, still a student, he obtained utility-model rights for a "parallel-lead fixed resistor" — described by the company as the first small resistor in Japan. As soon as he graduated, he founded Toyo Electronics Industry. The corporation was formally established on September 17, 1958.⁷

The timing was extraordinary. Japan's transistor-radio boom was about to detonate, and every radio needed resistors. Sato's product rode that wave to a roughly 60% share of the domestic resistor market.⁷ For a first product from a one-man operation, that is close to the maximum possible outcome.

What happened next is the more instructive part, and it is the reason this history is worth telling at all.

In 1964, at the absolute peak of the resistor business, ROHM's chief technology officer attended a lecture on integrated circuits. The lecturer suggested that in the near future, ICs might replace discrete resistors entirely. Sato heard the report and read it as a threat to his entire company. Rather than harvest the cash from a dominant position in a product he had invented, he committed to entering ICs while continuing the resistor business. The first semiconductor product followed in 1967. By 1971 ROHM had become the first Japanese company to set up shop in Silicon Valley — an unusual move for a mid-sized Kyoto components maker at a time when the phrase "global supply chain" did not yet exist.⁷

The name itself is an engineer's joke: R for resistor, ohm for the unit of resistance. The trademark changed from "R.ohm" to "ROHM" in 1979; the registered company name followed in 1981. The company listed on the First Section of the Tokyo Stock Exchange in 1989, and moved to the Prime Market when the exchange restructured in 2022.⁷

ROHM grew up inside a specific regional culture. Kyoto's globally minded manufacturers — 京セラ Kyocera, 日本電産 Nidec, 村田製作所 Murata — shared a management style distinct from Tokyo's keiretsu-affiliated conglomerates: engineering-led, allergic to debt, willing to hold enormous net cash for decades, and organized around long-horizon technical bets rather than bank relationships. ROHM was the archetype. For most of its history it carried an equity ratio above 80% — a balance sheet so conservative it was arguably inefficient.

Keep that in mind. It is the single most important contrast in this story, because the ROHM of 2023–2026 behaved nothing like it.

Two other threads from this era matter now. First, in 2008 ROHM acquired OKI Semiconductor, which became LAPIS Semiconductor — bringing display-driver and low-power LSI technology that still ships today under the LAPIS TECHNOLOGY brand, and, more consequentially, bringing the Miyazaki fab that would later become the site of ROHM's largest silicon carbide investment.⁷ Second, the resistor business never went away. It is still there, folded into the "Others" segment, quietly earning money.

That is a Chekhov's gun, and we will fire it in Section VII.

↑ Back to Top

III. Diversifying the Core: Discrete, Sensors, and the Automotive Push (2000s–2019)

If Section II is about a founder's paranoia, this section is about a company methodically buying its way into the customers it wanted.

The first move was geographic and technical. In November 2009 ROHM acquired Kionix, a MEMS accelerometer maker founded in 1993 and based in Ithaca, New York.⁸ Kionix gave ROHM motion sensors — accelerometers and gyroscopes — and a US engineering site. It has never been a headline business; ROHM does not break it out. It is best understood as an option on sensing that ROHM has held for seventeen years without it becoming a pillar. That is worth remembering when management talks today about optical devices and LiDAR laser diodes becoming "the next-generation pillar":⁹ the historical base rate at which ROHM converts a bolt-on technology acquisition into a reported growth engine is not encouraging.

The second move was pure customer arbitrage. On December 1, 2019, ROHM completed the acquisition of part of Panasonic's diode and transistor business — bipolar, built-in-resistor and JFET small-signal transistors, plus Schottky, TVS, Zener, switching and fast-recovery diodes.¹⁰ The technology was not exotic. What ROHM bought was Panasonic's automotive customer book and the scale to defend a top position in small-signal discretes against Nexperia. It was a deliberate repositioning around automotive power electronics, executed a full three years before silicon carbide became a mainstream investor conversation.

By the fiscal year ended March 2026, that repositioning had thoroughly reshaped the revenue mix. Automotive accounted for 49.2% of sales and industrial another 12.7% — so roughly 62% of ROHM's revenue now depends on two end markets that share a common characteristic: long design-in cycles, high qualification barriers, and brutal multi-year price-down expectations from the customer. Consumer contributed 21.9%, computer and storage 12.3% (and growing fast on AI servers), communications just 3.8%.²

The geography is more subtle than it first appears, and it is where a lot of casual analysis of ROHM goes wrong.

Measured by where product ships, China is roughly a third of ROHM's revenue. Measured by the nationality of the customer placing the order — the cut ROHM itself discloses — Japan is 52.4% of sales and China only 14.6%, or ¥70.4 billion.² Both facts are true. Chinese contract manufacturers assemble products for Japanese, American and European brand owners, so shipment destination overstates ROHM's exposure to Chinese demand and understates its exposure to Japanese OEM decision-making. For the fiscal year ending March 2027, ROHM guides China customer-nationality sales to fall 11.7% while every other region grows — with "other Asia" up 19.7% and the Americas up 13.0%.²

That single guidance line is the clearest statement management has made about the China problem. ROHM is not planning to win in China. It is planning to grow around it.

Which brings us to the reason the China question exists at all.

↑ Back to Top

IV. The Silicon Carbide Bet: Building Vertical Integration (2009–2019)

Start with the physics, because everything downstream depends on it.

A power semiconductor is a switch. Its job is to turn a large flow of electrical current on and off, very fast, with as little waste as possible. Every time it switches, some energy leaks out as heat. In an electric vehicle, that heat is range you paid for and did not get, plus a cooling system you had to build to carry it away.

Silicon has been the material for these switches for sixty years. Silicon carbide is a compound — silicon bonded to carbon — with a much wider "bandgap," which is the energy required to knock an electron loose. The practical consequences are three: a SiC device can hold off far higher voltages across a much thinner layer of material, it can switch faster, and it tolerates much higher temperatures before it fails. Thinner means less resistance when the switch is on. Faster means smaller magnetics and capacitors around it. Hotter means less cooling hardware.

For an EV traction inverter — the box that converts DC from the battery into AC for the motor — swapping silicon IGBTs for SiC MOSFETs has historically bought several percent of range at the same battery size, or a smaller battery at the same range. In a car, several percent of battery is worth real money. That is why SiC became the single biggest structural growth story in power electronics of the 2020s.

There is a catch, and it is the whole story. Silicon carbide is savagely difficult to make. Silicon wafers are grown from a melt; SiC crystals must be grown from vapour at temperatures above 2,000°C, over days, and they come out full of defects that kill device yield. The material is also nearly as hard as diamond, so slicing and polishing it is slow and expensive. For most of SiC's history, the substrate — the raw wafer — was the bottleneck and the cost.

ROHM's strategic answer was to own the bottleneck.

In 2009 it acquired SiCrystal, a German SiC single-crystal wafer maker spun out of research at the University of Erlangen-Nuremberg.⁷ That gave ROHM in-house crystal growth years before most competitors and, unusually, a merchant wafer business selling to third parties — including a multi-year supply agreement with STMicroelectronics.¹¹ The device milestones followed quickly: in 2010 ROHM started what it describes as the world's first mass production of SiC transistors, and in 2015 the world's first trench-type SiC MOSFET.⁷ Fourth-generation double-trench devices followed, and on April 21, 2026 ROHM announced its fifth generation, with roughly 30% lower on-resistance at 175°C junction temperature than the fourth.¹²

Let us be precise about what that record does and does not prove. It proves ROHM's device engineering is genuinely front-rank. Being first to commercialize trench SiC MOSFETs — ahead of Infineon and STMicroelectronics on that specific structure — is not a marketing claim; it is a documented sequence of product launches. It does not prove the technology lead converted into profit, and Section VI will show, from ROHM's own filings, that for at least two fiscal years it did not.

The manufacturing footprint was built to match the ambition. ROHM's SiC front-end process lines run at Miyazaki and at the Apollo plant in Chikugo, Fukuoka, with LAPIS's Miyazaki Plant No. 2 designated for SiC devices and wafers.²¹² The roadmap ran from six-inch to eight-inch wafers — the same jump silicon made decades ago, roughly 1.8x more usable area per wafer and a proportional cost reduction if yields hold. ROHM has since reported that its NEDO Green Innovation Fund-subsidised next-generation eight-inch SiC MOSFET project completed two years ahead of schedule, and that mass-production shipments of fifth-generation eight-inch devices have begun.²

The investor pitch was clean and, on its face, compelling: own the crystal, own the device, own the module, capture the whole margin stack, and use substrate self-supply as a cost shield against the Chinese entrants everyone could see coming.

The claim under test is therefore not "does SiC work." It does. The claim under test is that crystal-to-module vertical integration produces a durable cost advantage. Hold that thought — because as recently as the Integrated Report published in 2025, management was still asserting it explicitly, telling investors there would be "no major changes" to the SiC strategy and that ROHM believed "we can maintain a cost advantage in the manufacturing of SiC substrates."⁹ Twelve months later, the company impaired the assets and told the market its external SiC substrate business was being downsized because six-inch substrate demand had collapsed into oversupply.²

We will hold that contradiction up to the light in Section VI. First, there is a second bet to account for — one that had nothing to do with silicon carbide, and everything to do with the balance sheet that had to absorb both.

↑ Back to Top

V. The Alliance Era: Betting Big Alongside Toshiba (2023–2024)

On September 22, 2023, ROHM disclosed a transaction that had almost nothing to do with making chips.

The company had invested ¥100 billion in TB Investment Limited Partnership, a fund managed by Japan Industrial Partners, and had underwritten ¥200 billion of non-voting preferred stock issued by TBJ Holdings — the parent of the special-purpose vehicle JIP had created to take Toshiba private. Total commitment: ¥300 billion. It was funded with debt from financial institutions. The stated purpose was to "participate in the privatization of TOSHIBA and help to resolve its issues."¹³

Read that structure carefully, because the details are the argument. ROHM put ¥300 billion of borrowed money into a leveraged buyout of another company, and two-thirds of it went into a security with no votes attached. Whatever ROHM bought, it was not control, and it was not governance rights.

Japan Credit Rating Agency published a note the same week. Its assessment was that the borrowing raised concerns about deterioration in ROHM's financial structure, but would not have an immediate rating impact — because ROHM's equity ratio at the end of the June 2023 quarter had been "an extremely high 83.0%," interest-bearing debt was low, and liquidity was ample. JCR closed with a warning it would repeat in different words three times over the following thirty months: the amount of investment had been increasing in line with aggressive growth spending, and the burden of fixed costs was becoming heavier.¹³

Ten weeks later, on December 7, 2023, came the operating half of the Toshiba relationship: a ¥388.3 billion joint capital investment in power-semiconductor manufacturing. ROHM and LAPIS committed ¥289.2 billion, chiefly for LAPIS's Miyazaki Plant No. 2 producing SiC devices and wafers; Toshiba Electronic Devices & Storage and Kaga Toshiba committed ¥99.1 billion for silicon power devices at Ishikawa. Japan's Ministry of Economy, Trade and Industry agreed to subsidise up to ¥129.4 billion — a third of the total. The logic was division of labour: ROHM specialising in silicon carbide, Toshiba in silicon, each able to use the other's capacity.¹⁴

This was not a merger. It was two companies agreeing to build complementary capacity with a third of the bill paid by the Japanese state, at the top of the SiC capex cycle.

Then came the financing. On April 8, 2024, ROHM priced ¥200 billion of zero-coupon euro-yen convertible bonds in two ¥100 billion tranches, issued April 24 and maturing in 2029 and 2031. The stated purpose was explicit: to partially repay the bridge loan taken out in connection with participation in Toshiba's privatization. The closing price on the determination date was ¥2,431.5; conversion prices were set at ¥3,063 for the five-year tranche and ¥2,943 for the seven-year — premiums of 26.0% and 21.0%. ROHM's own supplementary material put potential dilution at 17.26% if the cash-settlement clause were not exercised.¹⁵

The company's defence of the structure was that zero-coupon CBs cost nothing on a cash basis, and that a conversion-restriction clause plus a par-value cash-settlement option would suppress most of the dilution. On the numbers, that is a real feature, not spin. But the market's objection was never about coupon. It was about what the money was for. Investors were being asked to accept up to 17% dilution risk to finance a minority, largely non-voting position in a private-equity buyout of a company ROHM did not control — at exactly the moment ROHM was also running capital expenditure at ¥186.7 billion in the year ended March 2024, the highest in its history.²

Here is the part that makes this more than a financing footnote. ROHM's own chief financial officer has since said it in print. Writing in the 2025 Integrated Report, Peter Kenevan — who joined as CFO in June 2025 — described the Toshiba investment's purpose as generating synergies, noted the company was "cautiously moving forward with deliberations because of strong market volatility," and then wrote: "I think that we should have made the decision from a slightly broader perspective, such as risk of dilution and capital cost if bonds are converted, which many investors are concerned about." He added that if it were determined the partnership would not create shareholder value, "we will immediately clean up both sides of the B/S."⁹

That is a sitting CFO stating, on the record, that the ¥200 billion financing decision was made too narrowly. It is the strongest single piece of evidence available on ROHM's capital-allocation discipline in this period, and it comes from inside the company.

The arithmetic of what the two bets did together is stark. ROHM's equity ratio was 81.4% at the end of the year to March 2023 and 83.0% at June 2023. It was 65.3% at March 2024, 61.7% at March 2025, and 59.1% at March 2026.⁹¹ Roughly twenty-four points of balance-sheet cushion, built over sixty-five years, spent in under three. Free cash flow, positive at ¥29.2 billion and ¥18.2 billion in the two years to March 2022 and March 2023, went to negative ¥59.2 billion and then negative ¥69.3 billion.⁹

A company that had been criticised for decades for hoarding cash converted itself, in thirty months, into a leveraged bet on two things: that EV silicon carbide demand would compound, and that a stake in a private Toshiba would eventually pay.

The first of those assumptions was about to be tested.

↑ Back to Top

VI. The Reckoning: Two Loss Years and a Broken Growth Story (FY2024–FY2026)

The first crack

The fiscal year ended March 2024 still looked fine on the surface. Revenue was ¥467.8 billion, operating profit ¥43.3 billion, net profit ¥54.0 billion.¹⁶ But return on equity had fallen to 5.7% from 9.2% the year before⁹ — the classic signature of a company whose asset base is growing faster than its earnings. Capital expenditure that year hit ¥186.7 billion, more than double the ¥79.9 billion of three years earlier.² The capacity was being built. The demand to fill it was not arriving on schedule.

In November 2024, ROHM did two things in the same month. It announced a structural reform programme.⁹ And it delayed its SiC plant expansion, pushing out the Chikugo and Miyazaki eight-inch wafer timeline in response to the EV slowdown.¹⁷ The vertical-integration roadmap from Section IV was visibly slipping in real time, and the company was, for the first time, publicly conceding it.

The year everything broke

The fiscal year ended March 31, 2025 was ROHM's first annual loss in over a decade. Revenue fell 4.1% to ¥448.5 billion. Operating profit went from +¥43.3 billion to a loss of ¥40.1 billion — an ¥83.4 billion swing on a ¥19 billion revenue decline. Ordinary loss was ¥29.7 billion; net loss attributable to owners of parent was ¥50.1 billion. ROE was minus 5.4%.¹⁶

The mechanism matters more than the magnitude. Cost of sales rose 16.2% while revenue fell 4.1%.¹⁶ That is not a demand problem; that is an under-absorption problem. ROHM had built factories for volumes that never came, and the fixed cost of those factories landed in cost of goods sold whether or not the wafers moved.

The segment detail is where the story becomes specific. In the year to March 2025, the discrete semiconductor devices segment — the home of both silicon and silicon carbide power devices — swung from a ¥13.0 billion profit to a ¥45.9 billion loss on ¥187.1 billion of sales. The IC segment, historically the reliable earner, also flipped, from ¥21.3 billion of profit to a ¥0.8 billion loss. Only Modules (¥2.7 billion) and Others — the resistor business — (¥2.5 billion) stayed profitable.¹⁶ EBITDA fell to the lowest level in the company's history.⁹

At the May 13, 2025 results briefing, management said the year had been the worst possible outcome, and admitted it had misjudged the market shift and failed to adjust production and forecasts in time.¹⁸

That admission is the pivot of this entire analysis, and it deserves to be quoted at its most precise. In the Integrated Report published later that year, ROHM answered an investor question about missing every financial target in its first medium-term plan this way: "while the selection of markets and products to focus on was not mistaken, our delayed response to rapid changes in market conditions led to excessive capital expenditure and an increase in inventory, resulting in a deterioration in profitability and asset efficiency."⁹

Note the construction. The strategy was right; the execution was late. That framing is not obviously wrong — SiC content in EVs did grow — but it is a convenient boundary, and investors should notice where management chose to draw it. When ROHM formally announced its second medium-term plan on November 6, 2025, it stated plainly that it expected to meet none of the financial targets of the first plan, and reminded readers that in May 2022 it had revised those targets upward mid-plan.¹⁹ Raising guidance into the top of a cycle and then missing every target is a specific, dateable failure of target-setting discipline, not an act of God.

The response was hard and fast: a voluntary retirement programme costing ¥2.1 billion in special losses, exit from the silicon wafer materials business with a ¥30.3 billion impairment, and a capital expenditure cut to ¥133.0 billion and then ¥82.4 billion.¹⁸² ROHM designated the three years from the fiscal year ending March 2026 through March 2028 a "structural reform period" and, notably, described itself as "partially moving away from IDM" — meaning it would no longer insist on manufacturing everything in-house.⁹ For a company whose entire investment thesis had been vertical integration, that is a significant philosophical concession, buried in an FAQ.

The impairment year

The fiscal year ended March 31, 2026 was, operationally, a recovery. Revenue rose 7.3% to ¥481.1 billion. Every end market except communications grew. Operating profit turned positive at ¥10.9 billion; ordinary profit reached ¥19.2 billion.¹

Then the write-down. ROHM recorded ¥193.6 billion of impairment losses, principally on domestic and overseas production facilities related to the power-device business and above all silicon carbide. Net loss attributable to owners of parent: ¥158.4 billion. Against ROHM's own initial guidance of ¥10 billion of net profit for the year, that is a ¥168 billion miss.¹²

ROHM's stated reasons, in its own presentation, are unusually candid and worth taking one at a time.²

First, BEV market forecasts peaked in 2023 and have been revised down continuously since — ROHM's chart shows the 2023-to-2030 compound growth assumption falling from 19% in its 2024 view to 15% in its 2026 view, citing GlobalData. The company cited reduced EV incentives in the United States and a review of regulations in Europe.

Second — and this is management describing its own capital allocation — "excessive investment due to upfront investment." ROHM wrote that while SiC remains a growth driver, "the gap between the amount invested to date and market growth has widened, making the payback period longer."

Third, the rise of Chinese SiC devices: "Local Chinese SiC device manufacturers are aggressively pushing prices down, causing our market share in China to decline."

Fourth, shrinking external sales of SiC substrates, with six-inch demand down on oversupply.

Set that fourth point against the Integrated Report claim from twelve months earlier that ROHM "can maintain a cost advantage in the manufacturing of SiC substrates."⁹ The merchant substrate business — the very asset that was supposed to prove the cost advantage of owning crystal growth — is the piece being downsized. The vertical-integration moat claim is not refuted in its entirety; ROHM still grows its own wafers and still has genuine device leadership. But the version of the claim that said substrate self-supply confers a durable cost edge against Chinese entrants is now contradicted by the company's own disclosure, and should be retired.

The accounting change nobody is talking about

There is one more item in the fiscal 2026 filing that changes how the recovery should be read.

Effective for the year ended March 31, 2026, ROHM changed its depreciation method for tangible fixed assets from the declining-balance method to the straight-line method. The stated rationale was that the business mix had shifted from consumer to automotive, that large-scale capacity had been built for automotive demand, and that stable long-run utilization made straight-line a better reflection of actual use.¹

The effect: depreciation expense for the year was ¥17.1 billion lower than it would have been under the old method, and operating profit and ordinary profit were each ¥15.554 billion higher.¹

Reported operating profit for the year was ¥10.864 billion.

Under the prior depreciation method, in other words, ROHM would have reported an operating loss of roughly ¥4.7 billion — a third consecutive year without an operating profit. The entire headline swing from loss to profit at the operating line is attributable to an accounting policy change, not to trading. The segment note makes the distribution explicit: the change added ¥6.2 billion to IC segment profit and reduced the discrete segment's loss by ¥7.8 billion.¹

To be fair to ROHM: the change is disclosed clearly, quantified precisely, and the stated justification is defensible — assets built for steady automotive volumes genuinely are better matched by straight-line depreciation, and the auditors, Deloitte Touche Tohmatsu, issued no going-concern note.⁹¹ This is not an accounting scandal. But it is a change of estimate that flattered the single number management most needed to improve, in the same year it took the largest write-down in company history, and its effect on the year-ahead guidance is larger still: the FY2027 operating-profit bridge attributes the improvement substantially to lower depreciation following the impairment.² Anyone using "ROHM returned to operating profit" as evidence that the turnaround is working is, on the arithmetic, mostly observing a depreciation schedule.

Leadership, credit, and the first checkpoint

The leadership change came before the second loss was even reported. On January 17, 2025, ROHM's board resolved that Katsumi Azuma would become President and Representative Director, Chief Executive Officer effective April 1, 2025, and that Isao Matsumoto — CEO through the capex build and the Toshiba commitments — would move to executive advisor and resign as a director effective March 31, 2025. The stated reason was "to accelerate the establishment of a robust management foundation for enhancing corporate value."²⁰ Announced in January, effective in April, on the eve of the first annual loss in twelve years: this reads as a change forced by results, not a planned succession.

The credit trajectory followed the same arc. Japan Credit Rating Agency affirmed AA-/Stable in January 2024, revised the outlook to Negative in January 2025, and downgraded ROHM to A+/Stable on January 16, 2026 — and has since flagged that it is watching the Toshiba and Mitsubishi Electric integration talks for further rating implications.²¹

Guidance for the year ending March 2027, issued May 13, 2026: revenue ¥510.0 billion, up 6.0%; operating profit ¥30.0 billion, up 176%; net income ¥29.0 billion.² The operating-profit bridge attributes roughly ¥28.9 billion to sales growth and ¥15.4 billion to lower depreciation, partially offset by ¥11.0 billion of higher material costs — driven substantially by surging gold prices, which ROHM is attacking both by raising prices and by switching from gold to copper bonding wire.²

The first checkpoint has now been passed, and it was good. For the quarter ended June 30, 2026, reported August 5, ROHM posted revenue of ¥135.7 billion, up 16.8%, and operating profit of ¥9.6 billion against ¥0.2 billion a year earlier — roughly double management's internal expectation and comfortably ahead of the ¥6.1 billion analysts expected.²²²³ Management credited recovering automotive power demand, AI-server-linked data centre business, improved SiC profitability, price increases and structural reform — and kept the full-year ¥30 billion operating profit guidance unchanged.²³

One quarter at nearly a third of the full-year target, with guidance untouched, is either conservatism or caution about the back half. Given this management team's recent history with mid-cycle upward revisions, the caution is arguably the right instinct.

Where this leaves the thesis. The SiC moat claim is not rejected. It is narrowed, substantially. ROHM's device technology leadership is real and continuing. Its capacity position is real. What the record rejects is the specific proposition that vertical integration would deliver a cost and margin advantage durable enough to survive a demand deceleration plus Chinese entry — because ROHM built to a 19% BEV growth assumption, the market delivered something closer to 15%, and the company had to write off ¥193.6 billion of the difference while conceding its share in China was falling on price. The forward version of the claim now has to be conditioned on three things: Chinese SiC pricing stabilising rather than continuing to fall; the ¥30 billion operating profit target being met at the Q2 and Q3 checkpoints without further help from accounting changes; and the discrete segment reaching sustained operating profitability.

That last one is the cleanest test available, and it takes us to the segment data.

↑ Back to Top

VII. Segment Reality Check: Who Actually Makes Money at ROHM

Every investor conversation about ROHM is about silicon carbide. Almost none of ROHM's profit comes from it.

ROHM reports four segments: Integrated Circuits, Discrete Semiconductor Devices, Modules, and Others. For the year ended March 31, 2026, the split of external sales was ICs ¥218.4 billion, Discrete ¥205.3 billion, Modules ¥31.6 billion, and Others ¥25.9 billion.¹

Now the profit, which is where the picture inverts.

Integrated Circuits earned ¥24.5 billion — an 11.2% segment operating margin, up from a small loss the year before. Modules earned ¥3.5 billion, an 11.1% margin. Others — the descendant of Ken Sato's 1954 resistor, now selling shunt resistors and high-power resistors into automotive and industrial applications — earned ¥4.1 billion on ¥25.9 billion of revenue, a 15.8% margin and the highest in the company. Discrete Semiconductor Devices, the segment that contains the entire silicon carbide business, lost ¥22.7 billion: a margin of roughly minus 11%.¹

Strip out the depreciation-method change and the picture sharpens further. Of the ¥15.6 billion aggregate benefit, ¥7.8 billion landed in the discrete segment and ¥6.2 billion in ICs.¹ On a like-for-like basis with the prior year, ICs earned closer to ¥18 billion and Discrete lost closer to ¥30 billion.

Read that plainly: in the fiscal year just reported, the analog IC business, the printhead and optical module business, and a resistor line older than most of the company's employees collectively earned about ¥32 billion of segment profit, and the power-device business gave back ¥23 billion of it.

The Q1 data through June 2026 shows the gap narrowing but not closed. ICs generated ¥7.3 billion of operating profit on ¥62.1 billion of sales, an 11.8% margin. Discrete generated a ¥0.9 billion operating loss on ¥58.9 billion of sales — better than the ¥6.3 billion quarterly loss a year earlier, but still a loss, in a quarter management described as roughly double its internal plan.²²

There are two honest ways to read this.

The bearish read is that ROHM is a mid-single-digit-growth analog company subsidising a capital-intensive power-device business that has not earned its cost of capital, and that the market is valuing the SiC option rather than the cash flows.

The more balanced read is that the discrete segment's loss is dominated by fixed cost absorption on capacity built for volumes that have not yet arrived, not by an inability to sell profitably at the unit level — which is why ROHM's path back is yield, the eight-inch transition, and volume, rather than exit. ROHM's own roadmap targets SiC break-even during the year ending March 2027, with fifth-generation eight-inch devices carrying most of the improvement.² Its stated SiC device and module sales growth was 41% in the year to March 2026, with a plan of more than 55% for the year ending March 2027 — meaningfully faster than the total SiC business including the shrinking substrate line, which grew 14% and is planned above 30%.²

Both reads share one implication for anyone following this company: the segment operating result for Discrete Semiconductor Devices is the number that decides whether the last decade of capital spending was a mistake or a delay. It is disclosed quarterly. It is currently negative.

Myth versus reality

Three consensus statements about ROHM circulate widely. Each is worth checking against the filings.

Myth: the FY March 2026 loss was a non-cash accounting event, and the underlying business returned to profit. Reality: the impairment was indeed non-cash, but the "return to operating profit" was itself substantially an accounting outcome. Reverse the depreciation-method change and the operating line was negative.¹ Both halves of the headline were shaped by accounting decisions, in opposite directions.

Myth: ROHM's vertical integration gives it a structural cost advantage in silicon carbide. Reality: ROHM asserted exactly this in its 2025 Integrated Report⁹ and, one year later, disclosed that its external substrate business was being downsized on oversupply and that its share in China was falling to Chinese price competition.² Owning crystal growth remains a genuine supply-security advantage. As a cost weapon in a glut, the company's own disclosure does not support it.

Myth: rejecting Denso proves the board is confident in the standalone plan. Reality: it proves the board did not support the proposal. With a founder-linked foundation holding more than a tenth of the register,⁹ the board did not need broad shareholder support to decline, and the valuation analysis behind the special committee's conclusion has not been published. Confidence is one explanation; it is not the only one the disclosed facts permit.

Which raises the question of who is standing on the other side of the trade.

↑ Back to Top

VIII. Industry Structure: Who ROHM Is Actually Fighting

The instinctive framing of the SiC market is a duel: the incumbent Western leaders versus a Japanese challenger. The data says something less flattering and more interesting.

On TrendForce's 2023 numbers, the top five SiC power-device suppliers held roughly 91.9% of market revenue. STMicroelectronics led with about 32.6%. onsemi was second, Infineon third, Wolfspeed fourth — and ROHM fifth.²⁴ ROHM is a credible participant in a concentrated market, not a co-leader of it. In the broader power-semiconductor market, Omdia puts Infineon at roughly 24% globally, while ROHM, Toshiba and Mitsubishi Electric together account for about 11%.⁶ That gap is the entire stated rationale for Section IX.

But the competitor that actually broke ROHM's thesis was not in that ranking.

The China mechanism

Here is the sequence, and it is worth being concrete because "Chinese competition" is otherwise just a phrase.

Two years ago, a mainstream six-inch silicon carbide wafer from Wolfspeed sold for around $1,500. Chinese suppliers now offer comparable wafers for as low as $500, or less. TrendForce attributes this to state-backed subsidies, the fact that most SiC production equipment falls outside US export controls, and Chinese electronics and automotive companies integrating backwards into chip manufacturing.²⁵

The capacity numbers explain the price. Yole Group's analysis is that the 2019–2024 capital expenditure boom created substantial upstream overcapacity just as the automotive market slowed. By 2024, Chinese players held roughly 40% of global SiC wafer and epiwafer capacity and were expanding fast into devices. Utilization has fallen to around 50% for upstream substrate processes and 70% for device lines. Yole expects the downturn to persist until 2027–2028, before renewed growth from eight-inch platforms and next-generation trench and superjunction MOSFETs takes the device market toward nearly $10 billion by 2030.²⁶

Now apply that to ROHM specifically. ROHM's central cost argument was that owning crystal growth would let it undercut competitors who had to buy substrates. That argument works when substrates are scarce and expensive. It inverts when substrates are abundant and cheap: a merchant buyer can simply purchase $500 wafers on the spot market, while ROHM carries the fixed cost of its own crystal-growth capacity whether it runs or not. That is precisely the dynamic ROHM described in its impairment disclosure — external substrate sales shrinking, six-inch demand falling on oversupply — and precisely why the ¥193.6 billion write-down landed on production facilities.²

Vertical integration did not fail because it was a bad idea. It failed as a moat because the input it was designed to secure stopped being scarce faster than anyone modelled.

Porter, and what the five forces actually say

Supplier power: low, and this is genuinely ROHM's strength. With SiCrystal, ROHM does not depend on Wolfspeed, Coherent or a Chinese substrate house for its most critical input. In a shortage, that is decisive. In a glut, it is a liability. Both statements are true at different points in the cycle, and 2026 is the second point.

Buyer power: rising, and structurally. Automotive customers qualify parts over years and are slow to switch, which is a real switching cost. But they also multi-source deliberately, and they negotiate annual price reductions as a matter of policy. Denso's approach in Section IX is the cleanest possible evidence of how much leverage sits on the customer side: the world's largest automotive inverter supplier decided that buying its power-semiconductor supply outright was preferable to procuring it. When a customer would rather own you than buy from you, your pricing power over that customer is not high.

Substitution: capped from below. SiC does not compete only with other SiC. It competes with silicon IGBTs, which remain cost-competitive at lower voltages and in cheaper vehicle segments, and increasingly with gallium nitride in lower-voltage, high-frequency applications like server power supplies. ROHM sells all three, which hedges the outcome but also means the SiC ramp partly cannibalises its own silicon power business. The practical effect is that SiC penetration proceeds at the pace at which its total system cost falls below silicon's — not at the pace at which SiC technology improves.

Rivalry: intensifying sharply, on the numbers above.

Barriers to entry: the claim that has aged worst. Crystal growth was supposed to take a decade to master. Chinese producers went from a standing start to roughly 40% of global wafer capacity in a few years, with state support and unrestricted equipment access. The barrier was real; it was just lower and shorter than the incumbents' models assumed.

Seven Powers

Under Hamilton Helmer's framework, ROHM's plausible sources of power are process power — accumulated manufacturing know-how in crystal growth, trench device fabrication and automotive-grade quality — and, prospectively, scale economies from vertical integration.

Process power is the more defensible claim. The trench-MOSFET record, the fifth-generation yield improvement, and the two-years-early completion of the NEDO eight-inch programme are evidence of genuine accumulated capability that cannot simply be bought.²¹² Cornered resource is weaker: SiCrystal was a real asset, but the resource it corners is no longer scarce. Scale economies is, for now, refuted by the direct evidence — a business with a segment operating margin of minus 11% does not have a scale advantage over competitors earning positive margins.

In the passive-component and small-signal discrete businesses, ROHM is routinely described as a top-tier global supplier alongside Yageo, Vishay, KOA, Panasonic and Samsung Electro-Mechanics. No independently verifiable numeric market-share ranking for ROHM's chip resistor position was located in the sources reviewed for this piece, so it should be treated qualitatively. What can be verified is more useful anyway: those businesses earn double-digit margins today while the glamorous one does not.¹

The scale gap against Infineon is not closing organically. Which is why, in March 2026, ROHM found itself with two very different offers to close it.

↑ Back to Top

IX. The 2026 Endgame: Denso's Bid and the Three-Way Power Chip Merger

On March 6, 2026, ROHM issued the kind of terse statement Japanese listed companies issue when a leak has run ahead of them: media had reported that ROHM and Toshiba were negotiating a power-semiconductor integration; ROHM had not announced it. The follow-up on March 13 confirmed the outline of the truth. ROHM had been in discussions since July 2024 with Toshiba and Japan Industrial Partners about strengthening business collaboration in semiconductors, including the possibility of a capital partnership — a process that traced back to a proposal ROHM had disclosed on March 29, 2024.²⁷

What the market did not know, and learned within days, was that a second party had made its own move.

Denso's approach

Denso was not a stranger. The two companies had announced a basic agreement on a strategic semiconductor partnership on May 8, 2025, working primarily on analog ICs — combining Denso's system-integration expertise in automotive with ROHM's device technology.⁴ By the September 30, 2025 shareholder record, Denso held 4.98% of ROHM, its third-largest shareholder.⁵

In March 2026 Denso proposed to acquire the whole thing, at a reported valuation of approximately ¥1.3 trillion.³ The strategic logic was easy to state. Denso holds the world's leading position in automotive inverters. Its exposure is that the highest-value content inside those inverters — the power semiconductors — is bought from third parties, ROHM among them. Owning ROHM would secure that supply chain and give Denso a semiconductor business with external customers beyond automotive.²⁸

ROHM's board established a special committee composed of independent outside directors and others to evaluate the proposal, and, per the company's disclosure, conducted written question-and-answer exchanges with Denso and held discussions on multiple occasions. On April 28, 2026, ROHM announced that it had not reached a conclusion to support the proposal, and that Denso — having concluded that continuing at that time would not necessarily contribute to enhancing Denso's own corporate value — had withdrawn it. Both companies agreed to deepen their existing partnership instead, centred on analog ICs, extending it beyond automotive into consumer and industrial equipment.⁴

Now, the honest question, which this article will pose rather than resolve.

The case that this was disciplined judgment: ROHM's business model depends on selling to everyone. Roughly half its revenue is automotive, and its customers include Denso's direct competitors and their tier-one suppliers. Becoming a Denso subsidiary would have compromised that neutrality and, quite possibly, cost ROHM business at rival inverter makers — destroying part of the value the acquirer was paying for. Preserving independence also preserved ROHM's ability to pursue the three-way integration, which addresses the scale problem more completely than a single-customer tie-up would.

The case that this was a board protecting itself: the approach arrived weeks before ROHM reported the largest loss in its history, tied to a strategic bet the board had approved. A full-price acquirer was standing at the door offering shareholders certain value at approximately ¥1.3 trillion, and the board declined without putting it to a vote. ROHM's ownership structure made that easy: the Rohm Music Foundation, a founder-linked entity, holds 10.76% of the shares⁹ — a block large enough to make any hostile approach structurally difficult regardless of price. An activist would ask exactly one question: was the special committee's rejection tested against the risk-adjusted value of management's own turnaround plan, and if so, on what discount rate, given that the same management had just missed every target of its previous five-year plan?

Both readings are consistent with the disclosed facts. The disclosure does not include the special committee's valuation work, so a definitive answer is not available from public sources.

The three-way integration

On March 27, 2026, Mitsubishi Electric announced that it had signed a memorandum of understanding to begin discussions on a business and management integration of its power-device business with the semiconductor businesses of ROHM and Toshiba Electronic Devices & Storage. Five parties signed: Mitsubishi Electric, ROHM, Toshiba, Japan Industrial Partners, and TBJ Holdings.⁵

The presence of JIP and TBJ Holdings in that list is not a technicality. Those are the vehicles through which Toshiba was taken private — and, as Section V established, the vehicles into which ROHM placed ¥300 billion. ROHM is simultaneously a merger counterparty, a creditor-adjacent preferred shareholder, and a limited partner in the fund that controls the other side of the negotiation. That is an unusually entangled position, and it is why ROHM's capital policy explicitly contemplates additional shareholder returns "upon realization of returns from LP investment in Toshiba."² The Toshiba stake and the merger are the same trade.

The complementarity argument is sound on paper: ROHM brings SiC and automotive, Toshiba brings a broad silicon power customer base, Mitsubishi Electric brings high-voltage industrial devices for rail, grid and industrial drives.²⁸ Japan's government has an explicit incentive structure pushing in this direction, requiring power-chip companies to make investments of at least ¥200 billion involving other companies to qualify for subsidies.²⁹

As of this writing, the deal is not done, and the sticking point is scope. Reporting in July 2026 indicated that Mitsubishi Electric CEO Kei Uruma was targeting an agreement around September 2026, but that the parties disagreed on what the combined entity should sell: Toshiba and ROHM want it to include a wide range of analog chips such as converters and drivers, so as to keep serving existing customers; Mitsubishi Electric wants a tighter focus on power devices alone.²⁹³⁰ Working-level talks stalled to the point that the three chief executives met in person. Uruma's reported comment — "There's only so much you can achieve through endless discussions" — is not the language of a process running smoothly.³⁰

ROHM has confirmed that due diligence has begun on the Toshiba D&S semiconductor business, with Mitsubishi Electric discussions proceeding in parallel.²

For investors, the scope disagreement is not a detail; it is the deal. ROHM's analog IC business is the segment currently earning an 11% margin and carrying the company. If it goes into the combined entity, ROHM shareholders are exchanging their profitable business for a share of a much larger, more complex, and less profitable one. If it stays out, ROHM contributes the loss-making discrete business and retains the earner — a cleaner outcome for ROHM shareholders, and precisely why Mitsubishi Electric might resist it. Whichever way that resolves, it will tell you more about the relative bargaining power of the three parties than any press release will.

And the whole thing carries a base rate that should not be ignored: Japanese consensus-driven industrial mergers have a long history of taking longer, delivering less, and preserving more overlapping structure than announced. Elpida, Renesas and Japan Display are all cautionary precedents for combinations assembled from national champions under government encouragement.

↑ Back to Top

X. Current Management, Ownership & Capital Allocation

The man in the seat

Katsumi Azuma was born on November 10, 1964 in Aichi Prefecture, graduated from the Nagoya Institute of Technology in March 1989, and joined ROHM the following month. He has worked nowhere else.²⁰

His career is a manufacturing career. He joined the board in June 2013 as director of the Discrete Production Headquarters, took charge of discrete and optical modules in 2017, moved to business and strategy in 2019, became chief operating officer in 2020 with responsibility for production, quality and sales, and in June 2023 became president of ROHM Apollo — the Chikugo subsidiary at the centre of the silicon carbide build. He held 27,416 ROHM shares at appointment.²⁰

That biography cuts two ways. Azuma is not an outsider brought in to clean house; he was COO through the capex surge and ran the SiC plant subsidiary during the ramp. He was in the room. Against that, he is a factory operator by training being asked to fix a factory problem, which is a better match of skill to task than a finance or sales background would have been.

His public statements are unusually specific, and specificity is the raw material of accountability. In the 2025 Integrated Report he wrote that his priority was to "return quickly to a corporate culture with a 20% profit margin," described the restructuring as "crouching before jumping," and said his top priority was improving profit margin "even if that means scaling down temporarily." He also volunteered a term limit: "I think my term of office is a maximum of six years, and ideally, I would like to pass the role of President to my successor at the end of the next Medium-Term Management Plan." He proposed a 1:1 ratio of cash to stock in his own remuneration, "demonstrating my determination to take responsibility for management results."⁹

Those are checkable commitments with dates attached. That is more than most Japanese industrial CEOs offer, and it should be held against the record as it accumulates. Azuma has been in the seat seventeen months. The evidence so far is one guidance year set conservatively and one quarter beaten decisively — a start, not a track record.

The CFO hire

The more revealing appointment may be the other one. In June 2025 ROHM appointed Peter Kenevan as chief financial officer. Kenevan had served as a ROHM outside director from June 2022, made partner at McKinsey's Tokyo office in 2000 after several years in China, and subsequently ran PayPal's Japan business. ROHM's own explanation of the hire is unusually blunt: "feedback from outside the company called for installing an appropriate CFO, and ROHM also felt that was necessary."⁹

Kenevan's stated agenda — raise ROIC through strict valuation of investments and assets, shrink assets that are not earning, lift ROE toward 10%, reduce cash on hand to roughly three months of sales, broaden restricted stock units beyond management to employees — reads like an investor's letter written from inside the company. He has also stated the dividend policy plainly: ROHM has "adopted a policy of maintaining our dividend level even when struggling with performance," with a payout-ratio target of 30% that he explicitly does not consider "etched in stone."⁹

That last point answers a question the market has reasonably asked. The dividend was not cut through the loss years. ROHM paid ¥50 per share annually in each of the years ended March 2023, 2024, 2025 and 2026, and plans ¥50 again for the year ending March 2027 — meaning the payout ratio was undefined in both loss years and works out to roughly 67% on the FY2027 plan.²⁹ Total dividends ran about ¥19.2 billion a year. In a year with negative free cash flow of ¥69.3 billion, a maintained dividend is funded from the balance sheet, not from operations. Whether that is admirable consistency or capital indiscipline is a genuine judgment call, and it is fair to note ROHM chose it deliberately and disclosed the reasoning.

Governance and ownership

The board structure is better than the Japanese average. Independent outside directors constitute 54.5% of the board, a majority achieved in 2022; an outside director chairs the board, and the Audit and Supervisory Committee and Officer Nomination Council are chaired by outside directors.⁹ The auditor is Deloitte Touche Tohmatsu, and the fiscal 2026 report carries no going-concern note.⁹¹ No activist stake and no auditor qualification was identified in the ROHM filings, JCR releases and press coverage reviewed for this article; that is a bounded observation about specific documents, not a general assurance.

The ownership register is where the interesting tension sits. As of March 31, 2025, the Master Trust Bank of Japan held 14.51%, the Rohm Music Foundation 10.76%, Custody Bank of Japan 5.97%, with State Street, the Bank of Kyoto and various global custodians below that.⁹ By September 30, 2025, Denso had appeared at 4.98%.⁵ ROHM had 403,760,000 shares issued, including 17,769,970 treasury shares, and 107,900 shareholders.⁹

The Rohm Music Foundation position deserves a moment. It is a founder-legacy charitable foundation — Sato was a serious patron of classical music in Kyoto — and it holds more than a tenth of a company that has just fielded a ¥1.3 trillion takeover approach. Whatever its intent, a stable double-digit friendly block is a takeover defence. Investors evaluating the Denso decision should factor in that ROHM's board could decline without meaningful risk of being overridden.

The capital allocation record, held up to the light

Bring Section V forward and set it beside the current plan.

Over the five years of the first medium-term plan, ROHM invested a total of ¥608.2 billion in capital expenditure — itself already curbed from an original ¥700 billion — while separately committing ¥300 billion to Toshiba's privatization, funded with roughly ¥400 billion of financing cash flow.²⁹ Annual capex peaked at ¥186.7 billion in the year ended March 2024 and has been cut to a planned ¥60.0 billion for the year ending March 2027.²

The second medium-term plan, "MOVING FORWARD to 2028," announced November 6, 2025, promises roughly ¥150 billion of capital expenditure over three years — less than one peak year — alongside operating cash flow above ¥300 billion, ¥100 billion of debt repayment, a dividend payout ratio above 30%, a target debt-to-equity ratio of 0.5 with temporary tolerance to 0.7, and cash on hand cut from approximately ¥430 billion to approximately ¥150 billion.²¹⁹

The financial targets for the year ending March 2029 are net sales above ¥500 billion, an operating margin above 20%, and ROE above 9%.¹⁹

That operating margin target requires arithmetic. ROHM earned a 2.3% operating margin in the year just reported. A 20% margin on ¥500 billion of sales implies operating profit above ¥100 billion — which is exactly what the structural reform programme is scoped to deliver, broken into ¥34–40 billion from SiC profitability, ¥18–22 billion from site reorganisation and portfolio optimisation, ¥9–12 billion from manufacturing and procurement cost, ¥7–10 billion from pricing, and ¥14–20 billion from non-SiC sales growth.² The plan is at least internally consistent and specific enough to track.

It is also, on the historical record, extremely ambitious. ROHM's operating margin was 9.3% in the year to March 2024 — a good year, before the losses.¹⁶ The 20% target is roughly double anything the company has achieved recently, and it is being set by an institution that raised targets in May 2022 and then missed every one of them.

A skeptical investor's summary would be short. This is a company that spent its balance sheet on two bets in thirty months, wrote down ¥194 billion of one and holds the other in non-voting preferred stock of a private company, maintained its dividend through two loss years on negative free cash flow, saw its credit rating cut, and is now asking the market to underwrite a doubling of its best-ever margin. The response — a genuinely credible one — is that the same management has cut capex by two-thirds, brought in an outside CFO with an explicit ROIC mandate, published a quantified reform bridge, and just beaten its own first quarterly checkpoint by roughly 100%.

The honest position is that this is a turnaround with a specific, testable plan and a management team whose credibility is currently a hypothesis rather than a finding.

↑ Back to Top

XI. Bull vs. Bear: The Investment Case

Why ROHM wins from here

The bull case does not depend on ROHM beating Infineon. It depends on four narrower propositions.

First, the profitable core is real and improving. The IC segment, Modules and Others together generate double-digit operating margins today, and they are not commodity businesses — automotive-qualified analog and motor-driver ICs carry multi-year design-in cycles and meaningful switching costs.¹ This is the cash flow that funds the turnaround, and it grew 7.1% last year.

Second, the cost programme is quantified, not aspirational. ROHM has published a ¥100 billion operating-profit bridge to the year ending March 2029 with named workstreams, and several items — electricity procurement renegotiation, warehouse consolidation, gold-to-copper wire conversion, price pass-through — are already in execution with disclosed status.² Capital expenditure has been cut from ¥186.7 billion to a planned ¥60.0 billion. Depreciation, the largest fixed cost, falls mechanically after the impairment.

Third, the AI server business is a genuine and underappreciated second growth vector. ROHM's server-related sales were ¥17 billion in the year to March 2026, are planned at ¥25 billion for the year ending March 2027, and management now targets over ¥100 billion by the year ending March 2031 — against a target of ¥30 billion set only in November 2025.²¹⁹ The mechanism is real: as AI rack power consumption scales toward the megawatt level, the number of power and analog devices per rack rises by more than an order of magnitude, and ROHM sells SiC, GaN, silicon MOSFETs, DrMOS and the surrounding analog content. Note carefully, though, that this target was tripled within six months by a management team with a documented history of raising targets into a cycle. Treat it as an option with real underlying demand and an unproven size.

Fourth, the technology lead is continuing. Fifth-generation eight-inch SiC MOSFETs are in mass production, the NEDO programme finished two years early, and the yield curve management has published shows fifth-generation devices materially above fourth.²¹² If SiC demand recovers on Yole's 2027–2028 timeline, ROHM enters that recovery with capacity already built and largely written down — meaning incremental volume carries very high contribution margins.

That last point is the strongest version of the bull case, and it is worth stating precisely: the impairment, however painful, permanently lowered the depreciation burden on assets ROHM still owns and still operates. The capital is spent. The capacity is real. What was destroyed was the accounting value, not the productive capability.

Why it breaks

The growth thesis has already been walked back once, by management, in writing. ROHM built to a 19% BEV growth assumption and now models 15%.² If that number is revised again — and US incentive rollbacks and European regulatory review are live, not hypothetical — the ¥34–40 billion of profit improvement earmarked for SiC does not appear, and roughly a third of the path to the FY2029 target goes with it.

Chinese pricing shows no sign of bottoming on any independent forecast. Yole's base case is overcapacity through 2027–2028.²⁶ ROHM has conceded its share in China is falling on price.² A company cannot cost-cut its way through a multi-year deflation in its own growth product.

The loss-making segment is the growth segment. Discrete was still in operating loss in the June 2026 quarter, in a quarter that was otherwise excellent.²² Break-even is a plan, not a result.

The merger is unresolved with a self-imposed deadline that has already been extended once in practice. Discussions began in July 2024, were formalised in March 2026, and are targeted for September 2026 with the fundamental question of scope still open.²⁷³⁰ Integration risk here is not abstract — merging three power-device organisations with overlapping fabs, three quality systems and three customer books, under Japanese consensus governance, is exactly the situation where announced synergies evaporate into preserved duplication.

The financial cushion is materially thinner than it was. Equity ratio at 59.1%, a credit rating cut to A+, ¥200 billion of convertible bonds maturing in 2029 and 2031 with up to 17.26% potential dilution, and a maintained dividend through two loss years.¹²¹¹⁵ ROHM is still a financially sound company by most measures. It is no longer a fortress, and the difference matters if the recovery slips.

And the CEO has seventeen months in the job. Every element of the bull case above is an execution claim.

The stress test an activist would run

Four questions, none of which have public answers.

What is the current mark on the ¥300 billion Toshiba position, and what does ROHM expect to realise from it and when? Management has tied incremental buybacks to that realisation.² The size, timing and certainty of it are not disclosed.

Why was a ¥1.3 trillion approach declined without shareholder consultation, and what discount rate did the special committee apply to management's own plan — a plan produced by the team that missed every target of its predecessor?

Why maintain a ¥19.2 billion annual dividend across two years of negative free cash flow while simultaneously repaying ¥100 billion of debt and asking for patience on returns?

And are executive incentives tied to the FY2029 targets management is now setting, or to the ones it missed? ROHM has disclosed that restricted stock units are granted to management and corporate officers, that it is revising the officer remuneration system, and that the CEO has proposed a 1:1 cash-to-stock ratio for his own pay.⁹ The specific vesting metrics were not located in the sources reviewed here, and they are the detail that would show whether the incentive structure has actually changed.

The three KPIs that matter

Everything above collapses into three numbers, all of which ROHM discloses quarterly and none of which require calculation.

One: the Discrete Semiconductor Devices segment operating result. This is the single cleanest test of whether the SiC bet works. It was minus ¥22.7 billion for the year ended March 2026 and minus ¥0.9 billion in the June 2026 quarter.¹²² Sustained positive quarters mean the capacity is absorbing and the yield curve is real. Continued losses through a strong demand environment mean the capital is impaired in economics as well as in accounting.

Two: consolidated operating profit against the ¥30 billion full-year guidance, tracked at the Q2 and Q3 checkpoints. This is the management-credibility test. Guidance was set in May 2026 and reaffirmed in August. Meeting it would be the first time this company has hit a number it set since before the downturn.

Three: China customer-nationality revenue against the guided 11.7% decline. This is the competitive-pressure gauge. Guiding a region down while every other region grows is management's own statement of where it is losing. A worse outcome than guided means Chinese price competition is spreading beyond where ROHM has modelled it; a better one means the price war is finding a floor.

↑ Back to Top

XII. Risk Radar

SiC oversupply persistence is the dominant mechanism risk, and it operates through fixed-cost absorption rather than through demand. ROHM's power-device factories were sized for a market growing at 19% a year. If Chinese capacity keeps expanding faster than EV production — Yole's base case through 2027–2028²⁶ — ROHM's core growth segment stays margin-impaired for years rather than quarters, and the ¥34–40 billion SiC contribution to the FY2029 profit bridge does not materialise.

Integration execution risk is live and near-term. The three-way talks could collapse over the analog-scope disagreement, or close on terms that hand ROHM shareholders a minority position in an entity whose power-device strategy they no longer control. The presence of Japan Industrial Partners and TBJ Holdings on the other side of the table, while ROHM holds ¥300 billion of exposure to those same vehicles, is a related-party complexity that has not been publicly resolved.⁵¹³

Automotive customer concentration and buyer power. With roughly half of revenue from automotive, and with the world's largest inverter maker having just attempted to buy the company outright, ROHM's power-device economics depend on a small number of large customers. Any one of them choosing to in-source, dual-source aggressively, or delay a platform materially moves ROHM's volumes.

Refinancing and cost of capital. The A+ rating with stable outlook is investment grade, but the direction of travel over three years has been down, and JCR has said the integration talks are a rating input.²¹ Two ¥100 billion convertible tranches mature in 2029 and 2031; ROHM has structured them to allow cash settlement of par, which protects EPS but converts a potential equity outcome into a cash outflow at maturity if the stock is below the conversion price.¹⁵

China exposure runs in both directions, which is unusual and worth naming. Chinese customers are roughly 15% of ROHM's revenue by customer nationality and roughly a third by shipment destination, while Chinese SiC makers are the primary source of price pressure on ROHM's growth product.² ROHM's customer base and its most dangerous competitor occupy the same geography, which limits how aggressively it can respond.

Input cost inflation is currently a real and specific drag, not a generic macro worry. ROHM has repeatedly cited surging gold prices — gold is used in bonding wire — as a material variable-cost headwind, costing roughly ¥26.8 billion of material cost increase in the year ended March 2026 and a further ¥11.0 billion planned in the year ending March 2027, and it is responding with price pass-through and a customer-approval process to switch from gold to copper wire.² That switch requires automotive customer requalification, which takes time ROHM does not control.

Restructuring execution risk is concentrated in the manufacturing footprint. ROHM is consolidating from nine front-end locations toward eight and reorganising back-end sites, with production of ICs and transistors at the Kyoto headquarters and ICs at Kasaoka scheduled to end and transfer elsewhere, and a transfer agreement for the Dalian plant finalised.² Moving qualified automotive production lines between fabs requires customer approval at every step — ROHM's own materials note that several reform items are "under customer review." Any of those approvals slipping pushes savings from the year ending March 2027 into March 2028.

Accounting judgment belongs on this list rather than in a footnote. The depreciation-method change is disclosed and quantified, but it means year-over-year operating profit comparisons across the March 2026 boundary are not like-for-like, and the impairment has permanently reduced the depreciation base against which future margins will be measured.¹ Investors comparing ROHM's future operating margin to its own history should adjust for both.

↑ Back to Top

XIII. Durable Lessons & What This Story Teaches

Vertical integration is a moat only when the input is scarce. ROHM's crystal-to-module strategy was not stupid; for roughly a decade it was correct. Owning SiCrystal genuinely protected ROHM through the substrate shortage. What the last three years demonstrated is that integration converts a variable cost into a fixed cost — brilliant when the variable cost is rising and supply is short, ruinous when supply floods and the market price of your self-supplied input collapses below your internal cost. The moat and the trap were the same asset.

Capital committed to relationships deserves the same underwriting as capital committed to capacity. ROHM's ¥300 billion Toshiba position and its ¥608 billion capex programme competed for the same balance sheet, were approved in the same eighteen months, and were stress-tested by the same downturn. One of them has been written down; the other is held in non-voting preferred stock and an LP interest whose realisation timing is not disclosed. The company's own CFO has said the financing decision should have been made from a broader perspective.⁹ The generalisable lesson is that strategic investments in other companies escape the ROI scrutiny applied to factories precisely because they are called "strategic."

Management's own words are the highest-quality disconfirming evidence available, and they are usually free. Two of the most important facts in this analysis came from ROHM's own documents: the admission that it expected to meet none of its first medium-term plan's financial targets,¹⁹ and the CFO's written regret about the convertible bond.⁹ Companies that disclose this clearly are, on balance, more trustworthy than those that do not — but the disclosure is only useful to investors who actually read the FAQ section of the integrated report.

Consolidation among mid-sized national champions is rational and historically hard. Combining ROHM, Toshiba and Mitsubishi Electric's power businesses is the only realistic route to competing with Infineon on scale, and Japan's subsidy structure is explicitly designed to encourage it.²⁹ It is also the same logic that produced Elpida and Japan Display. Strategic necessity does not confer execution ability, and the fact that the three CEOs had to meet personally because working-level talks were going nowhere³⁰ is a data point about how this will go, not just about how it started.

And credibility is rebuilt one guidance cycle at a time. Azuma inherited a company mid-collapse and has done, so far, the things a new CEO should do: cut capex hard, take the impairment in one year rather than three, hire an outside CFO with a mandate to shrink the balance sheet, publish a quantified bridge, and set a conservative first-year number. None of that is proof. It is a testable position, and the tests arrive every three months.

↑ Back to Top

XIV. Outro

Where ROHM stands today: a technically credible silicon carbide pioneer with a genuinely profitable analog core, mid-turnaround under a seventeen-month-old CEO, mid-negotiation on a transformative three-way merger that could reshape the global power-semiconductor league table, having just declined an approach valued at roughly ¥1.3 trillion, and carrying the largest write-down in its history on the strategy that was supposed to define its next decade.

The company Ken Sato built by pivoting out of a dominant resistor business at its peak, because he was afraid of what integrated circuits might do to it, has spent the last three years discovering how expensive it is to make that kind of bet when the market moves against you.

Two dates should anchor the next six months. The first is the targeted conclusion of the ROHM–Toshiba–Mitsubishi Electric integration talks, which Mitsubishi Electric's leadership has aimed at September 2026 and which remain unresolved on the fundamental question of whether analog chips are in or out.³⁰ The second is the next quarterly checkpoint against the ¥30 billion operating profit guidance for the year ending March 2027, and specifically whether the Discrete Semiconductor Devices segment — the one that holds the silicon carbide business, and the one that has now lost money for two consecutive fiscal years — finally posts a positive quarter.

Everything else in this story is commentary on those two numbers.

ROHM Co., Ltd.

Table of Contents

ROHM Co., Ltd.: The Kyoto Chipmaker's All-In Bet on Silicon Carbide

I. Introduction & Episode Roadmap

II. Kyoto Origins: From One Resistor to a Components Empire (1958–1990s)

III. Diversifying the Core: Discrete, Sensors, and the Automotive Push (2000s–2019)

IV. The Silicon Carbide Bet: Building Vertical Integration (2009–2019)

V. The Alliance Era: Betting Big Alongside Toshiba (2023–2024)

VI. The Reckoning: Two Loss Years and a Broken Growth Story (FY2024–FY2026)

The first crack

The year everything broke

The impairment year

The accounting change nobody is talking about

Leadership, credit, and the first checkpoint

VII. Segment Reality Check: Who Actually Makes Money at ROHM

Myth versus reality

VIII. Industry Structure: Who ROHM Is Actually Fighting

The China mechanism

Porter, and what the five forces actually say

Seven Powers

IX. The 2026 Endgame: Denso's Bid and the Three-Way Power Chip Merger

Denso's approach

The three-way integration

X. Current Management, Ownership & Capital Allocation

The man in the seat

The CFO hire

Governance and ownership

The capital allocation record, held up to the light

XI. Bull vs. Bear: The Investment Case

Why ROHM wins from here

Why it breaks

The stress test an activist would run

The three KPIs that matter

XII. Risk Radar

XIII. Durable Lessons & What This Story Teaches

XIV. Outro

References