💸 A Japanese chipmaker has admitted, in unusually plain language, that it bet too big. Rohm posted its worst-ever net loss, 158.4 billion yen, after writing down 193.6 billion yen on power chips it had built factories to ship. The buyers, electric vehicles, did not arrive fast enough. Chinese rivals undercut wafer prices by two thirds. And the CEO reached for the Japanese idiom for squeezing out pus.

A $1 billion confession, in six characters

On May 12, 2026, Rohm Co., a Kyoto-based maker of power semiconductors founded in 1958, released earnings that closed a chapter on what now reads as a strategic miscalculation. The headline number: a 158.4 billion yen net loss for the fiscal year ending March 2026, the largest in the company's history and its second consecutive annual loss. Dollar figures throughout use a rate of about 158 yen, which puts that loss near $1 billion.

The cause was concentrated in a single line item: 193.6 billion yen of impairment charges on power semiconductor manufacturing equipment, almost all of it tied to silicon carbide (SiC), the wide-bandgap material that was supposed to be the chip industry's next gold rush, especially for electric vehicles.

What made the press conference unusual was the language CEO Katsumi Azuma chose. He used the Japanese phrase "uminoshikitta" (膿み出し切った), which literally means "we have squeezed out the pus." It is the sort of vivid, almost medical idiom you reach for when you want to be heard saying that the bad news ends here. He paired it with an explicit admission: in the previous five-year plan, when "there was wind in our sails," Rohm "did simply too much investment." That kind of self-criticism is rare from Japanese executives, who tend to default to softer phrasing.

The bet that didn't pay off

To understand the loss, it helps to back up to roughly 2021–2022, when nearly every power semiconductor company on the planet was racing to expand SiC capacity. The story automakers were telling was straightforward. EV powertrains would migrate from silicon-based chips to silicon carbide, which handles higher voltages with less heat loss, extending range by single-digit percentages and making fast charging more efficient. Tesla had already adopted SiC in the Model 3. European and Chinese automakers were following. Demand projections from market research firms looked vertical.

Rohm responded the way Japanese manufacturers historically respond to a perceived once-in-a-generation opportunity. It built. Between fiscal 2021 and fiscal 2025, the company spent a cumulative 608.2 billion yen on capital expenditure, a substantial portion of it on SiC. New cleanrooms went up at the Chikugo plant in Fukuoka Prefecture for 200-millimeter SiC wafer production. In April 2026, Rohm announced it had achieved 200mm SiC MOSFET mass production technology two years ahead of its original schedule, a genuine engineering achievement.

The problem was that by the time the factories were ready, the market was no longer.

Three things broke at once

EV demand stopped accelerating. Through 2023 and into 2024, European and U.S. EV growth visibly decelerated. Tier-1 automotive suppliers started pushing out delivery schedules. Tesla cut prices and slowed its capacity additions. Volkswagen and General Motors trimmed EV production targets. For SiC suppliers, this translated into the brutal arithmetic of expensive fabs running below planned utilization.

China industrialized SiC faster than expected. According to market research firm TrendForce, Wolfspeed held the top spot in SiC substrates with a 33.7% share in 2024, with China's TanKeBlue at 17.3% and SICC at 17.1% in second and third. Pricing pressure followed. Six-inch SiC wafers that once sold for $1,500 each are now offered by Chinese suppliers for $500 or less. A threefold price collapse on the underlying substrate makes it very hard for anyone running a Western or Japanese cost structure to recover invested capital on schedule.

Wolfspeed cracked first. The U.S.-based SiC pioneer, which had been the loudest investor in the new generation of 200mm wafer fabs, filed for Chapter 11 bankruptcy in June 2025. It emerged in late summer after cutting debt by roughly 70%, about $4.6 billion, and reducing headcount by 20%. The bankruptcy was a clarifying event for the rest of the industry. It confirmed in legal documents what spreadsheets had been hinting at: the timing assumptions on SiC capacity build-outs had been wrong.

Rohm's writedown is, in essence, the Japanese version of that same recognition, taken through impairment accounting rather than Chapter 11.

What the numbers actually say

Underneath the headline loss is a more nuanced story. Operating profit, the result of actual business activity before one-time writedowns, was 10.8 billion yen in fiscal 2026, a swing from roughly 40 billion yen of operating loss the year prior. Sales rose 7.3% to 481.1 billion yen. EBITDA, which strips out depreciation, rose 56.6% to 67.8 billion yen.

In other words, the underlying business is recovering. What the impairment does is acknowledge that some of the SiC capacity Rohm built will not generate the cash flow originally projected, at least not on the original timeline, so its book value has to be written down to match reality.

For fiscal 2026 (ending March 2027), Rohm forecasts sales of 510 billion yen, operating profit of 30 billion yen (nearly triple the previous year), and a return to net profit of 29 billion yen. Critically, having written down the assets, depreciation expense drops sharply, mechanically improving reported profitability even if revenue grows only modestly.

The strategic reversal: from "build" to "live within means"

The most consequential disclosure was not the loss itself but the capital expenditure plan that goes with it. After spending 608.2 billion yen over the previous five-year plan, Rohm now intends to spend approximately 150 billion yen over the first three years of the next plan, a roughly 75% reduction in pace.

This is a different company than the one that broke ground in Chikugo. Azuma's framing was that Rohm will return to being "a company that, after letting go of what's not working, generates large improvement and profit again." The mechanism is straightforward: stop building capacity ahead of demand, run the existing fabs harder as orders return, and let depreciation declines do the rest.

It is also a tacit admission that the era of Japanese semiconductor companies trying to win SiC market share through unilateral capacity investment is over.

The bigger reshuffle: three Japanese players, one combined company?

Rohm's earnings sit in the middle of a fast-moving industrial reorganization. It began on March 6, 2026, when reports emerged that Denso, the Toyota-affiliated tier-1 supplier, had proposed acquiring all of Rohm's shares through a tender offer valued at around 1.3 trillion yen, motivated by securing automotive chip supply. Denso already held about 5% of Rohm. Rohm set up a special committee of outside directors to evaluate it. Azuma described the approach as the arrival of the black ships.

That proposal unstuck a separate negotiation. On March 27, Rohm, Toshiba and Mitsubishi Electric announced a basic agreement to begin talks on combining their power semiconductor businesses, alongside Japan Industrial Partners and TBJ Holdings. The three bring non-overlapping strengths: Rohm leads in SiC, Toshiba in silicon MOSFETs, Mitsubishi Electric in IGBTs and power modules. A combined entity was reported as ranking second globally behind Germany's Infineon Technologies.

Denso withdrew its proposal in late April without Rohm's support, and both companies said their semiconductor partnership would continue. Internal backing for management is not unanimous either. At the June 29 shareholder meeting, support for Azuma's reappointment came in at 76.8%, down 12.5 points from the previous year.

In that context, the impairment is doing political work as well as accounting work. By taking the writedown now, Rohm enters consolidation talks with cleaner financials and the standing that comes from having confronted its own mistakes. It is easier to negotiate merger terms when you have already written down your own balance sheet.

The contrast with Wolfspeed is instructive

The two companies took roughly the same bet at roughly the same time. Both built ahead of demand. Both got caught when EV growth softened and Chinese supply scaled. But the resolutions diverge.

Wolfspeed went through Chapter 11, restructured its debt, and emerged smaller. It also lost CHIPS Act funding under the new U.S. administration, leaving its long-term capital position fragile. Renesas Electronics, which had prepaid Wolfspeed roughly $2 billion on a 10-year wafer supply contract, ended up converting some of that exposure into Wolfspeed equity and seating a director on the board. It was an awkward outcome for a Japanese chipmaker that had simply hoped for reliable wafers.

Rohm did not need bankruptcy court because it had a healthier balance sheet to begin with, operating profit was already positive, EBITDA was strongly positive, and the underlying business kept generating cash. The impairment is painful but voluntary, and it leaves the company free to focus on integration rather than restructuring negotiations with creditors.

The two paths illustrate the same lesson at different intensities. In SiC, being first did not protect you, and neither did scale.

What to watch from here

A few things will determine whether Azuma's "we've cleared out the pus" reads, in a year, as honest reckoning or as wishful thinking.

The first is whether EV demand actually recovers on the timeline Rohm now assumes. If Chinese EV exports continue to displace Japanese and European production, the addressable market for SiC outside of China may remain structurally smaller than 2022's projections.

The second is whether the three-way merger with Toshiba and Mitsubishi Electric actually closes, and on what terms. A combined entity needs to be more than the sum of its current shares; it needs to make integrated design wins at the inverter level, where automakers and industrial buyers are increasingly making purchase decisions on multi-chip systems rather than individual MOSFETs.

The third is whether Denso's tender offer for Rohm advances. A Rohm that becomes part of Toyota's industrial group has different incentives than a Rohm operating as an independent merchant supplier. Both paths are coherent. They are not the same path.

Finally, there is the question of gallium nitride (GaN). Some analysts now argue GaN will take share from SiC in the lower voltage segments (consumer chargers, data center power supplies, some EV applications). If that view proves right, Japan's SiC consolidation will need to make room for a GaN strategy too, and Rohm, having just written down a generation of SiC capacity, will face the question of how aggressively to invest in the next material before it has fully digested the last one.

For now, the headline is straightforward. The Japanese power semiconductor industry is reorganizing under financial pressure, the SiC gold rush is over for at least this cycle, and Rohm has chosen, in a culturally noteworthy way, to say the hard parts out loud.

In your country, how is your domestic chip industry handling the EV slowdown and Chinese competition? Are companies still investing aggressively, or pulling back?

References