For 14 months, Japan's most-watched factory had a problem: it kept losing money. Now, finally, it doesn't.

TSMC's Kumamoto fab — the symbol of Japan's bid to claw back relevance in the semiconductor world — posted its first quarterly profit between January and March 2026. The number itself is modest: NT$951 million, or roughly $30 million (about ¥4.78 billion at the current rate near ¥158 to the dollar). But the symbolism is heavier than the figure. A plant that consumed roughly ¥476 billion ($3 billion) of public money before it even produced a single chip has now, by one measure at least, started to justify the bet.

The question that follows is the one that has shadowed the project from the beginning: was it worth it?

The 14-month climb out of the red

The Kumamoto plant — run by Japan Advanced Semiconductor Manufacturing Inc. (JASM), TSMC's Japanese subsidiary — began mass production in December 2024. Its first full year was rough. JASM posted a net loss of NT$3.25 billion in Q1 2025, then another NT$4.52 billion across the first half of the year, with utilization sitting at roughly 50%, according to TrendForce reporting cited by Commercial Times. In Q4 2025 the loss narrowed to NT$1.39 billion. Then came the swing: a NT$951 million profit in Q1 2026, confirmed by TSMC's financial filings reported by Taiwan's Central News Agency on May 16.

The driver, analysts say, is straightforward — utilization is rising. The fab makes chips on mature process nodes (12, 16, 22, and 28 nanometers) for image sensors and automotive applications. These aren't the kind of cutting-edge chips that grab AI headlines. They're the workhorses inside your smartphone camera and the brake control module of a Toyota. When those orders fill the line, the fab makes money. For most of 2025, they didn't fill the line.

There's a structural reason for that. The Kumamoto Fab 1 was originally conceived around 2019 as a captive supplier for Sony, which needed the logic chips that sit underneath its CMOS image sensors (the things that turn light into the data in your photos). Sony's image sensors are the eyes of iPhone cameras and most of the smartphone industry. Around that anchor, Japan's government layered in Denso and Toyota as minority investors, hoping the same fab could also stabilize the auto sector's chip supply — still scarred by the 2021–2022 shortage that idled assembly lines worldwide.

The shareholding reflects that hierarchy: TSMC roughly 86.5%, Sony 6%, Denso 5.5%, Toyota 2%, according to the joint announcement from the partners.

Tokyo paid a lot for this win

Japan's investment in TSMC's Kumamoto presence is one of the most aggressive industrial policy plays the country has run in decades. The government committed up to ¥476 billion ($3 billion) in subsidies for Fab 1 and up to ¥732 billion ($4.6 billion) for the second fab now under construction next door. Add them together and Japanese taxpayers are on the hook for more than ¥1.2 trillion ($7.6 billion). Total project investment, including TSMC's own capital, exceeds $20 billion.

For perspective: the entire 2025 budget of Japan's Ministry of Economy, Trade and Industry for chip and AI support was about ¥1.23 trillion, roughly four times the previous year — and the FY2026 number is bigger still. A meaningful share of that has been chasing the goal of putting cutting-edge chip manufacturing back inside Japan's borders.

That spend has been controversial. Critics — including some semiconductor analysts writing in Japanese trade press — have argued that the mature-process market for which Kumamoto was originally designed is already oversupplied, especially after the auto-chip crunch eased. Why pour public money into yet another 28-nanometer line when TSMC's own Taiwan facilities at that node have been running below capacity? The Q1 2026 profit doesn't refute the question. It does, however, take some of the sting out of it.

A reality check from Arizona

The cleanest comparison sits across the Pacific. TSMC's Arizona Fab 21 — its first U.S. facility, anchored by Apple and NVIDIA orders — began producing 4-nanometer chips in late 2024 and turned profitable faster than Kumamoto did. According to Commercial Times reporting on TSMC's filings, the Arizona operation posted around NT$4.2 billion in net income in Q2 2025 alone, and roughly NT$4.7 billion across the first three quarters of 2025, while JASM was still bleeding. By Q2 2025 it had also begun contributing investment income back to the TSMC parent. Apple has committed to buying more than 100 million chips made in Arizona during 2026.

Why the difference? Utilization, mostly. The Arizona fab landed at the right time — early in the AI-driven scramble for 4-nanometer capacity — and it ran near-full from the start. Kumamoto, by contrast, opened straight into a soft patch in the auto-chip cycle. Same operator, same engineering discipline, very different demand environment.

The German operation tells yet another story. ESMC, TSMC's Dresden joint venture with Bosch, Infineon, and NXP, is still under construction; equipment installation is slated for the second half of 2026, with production targeted for 2027. It's the EU's first FinFET-capable foundry — the kind of plant that uses 3D transistor structures (the technology that took chipmaking past the 22-nanometer plateau) — and Berlin and Brussels are jointly putting €5 billion into the €10 billion project. Compared with Dresden, Kumamoto is far ahead in time. Compared with Arizona, it's been the late bloomer. Both comparisons are fair.

Where Rapidus fits in this picture

While TSMC was sweating utilization in Kumamoto, an entirely different Japanese bet was taking shape 1,800 kilometers north. Rapidus — a domestic startup backed by Toyota, Sony, NTT, and six other Japanese companies, plus the government — is building a 2-nanometer fab in Chitose, Hokkaido, using process technology licensed from IBM. The company says it produced its first working 2-nanometer transistors in July 2025 and is targeting mass production in 2027.

Rapidus is the more spectacular gamble. The government has already committed roughly ¥1.7 trillion (about $11 billion) in support, and the economy ministry has pledged another ¥1 trillion across fiscal 2026 and 2027, bringing cumulative public backing toward ¥2.9 trillion. The total project bill, including private capital still to be raised, is estimated at around ¥7 trillion ($44 billion).

The strategic logic of running both bets at once is that they cover different ground. TSMC Kumamoto secures Japan's automotive and image-sensor supply chains using proven technology and a proven operator. Rapidus aspires to put Japan back on the leading edge — the territory it ceded to Taiwan, South Korea, and the United States after the 1990s. They're not competitors so much as two different answers to two different anxieties.

The Kumamoto profit gives the more conservative half of that strategy its first concrete evidence of working. Rapidus, by contrast, won't have its own profit-and-loss verdict for years.

What Kumamoto actually means for the chip map

The Q1 profit is one data point, not a vindication. JASM still has to demonstrate that utilization can stay up through the rest of 2026. The second Kumamoto fab — originally planned for 6-nanometer production but now reportedly being upgraded to 3-nanometer to chase AI demand — has had its construction schedule disrupted, with equipment vendors reportedly told no new tool deliveries would be needed in 2026. The leap from a 12-nanometer specialty line to a 3-nanometer leading-edge operation is not a minor renovation. It is a different kind of factory.

There's also the geopolitical layer. TSMC's overseas expansion — $165 billion in Arizona, €10 billion in Dresden, $20 billion in Kumamoto — has been accelerated by Washington's pressure on Taiwan to diversify away from the cross-strait risk that hangs over every leading-edge chip on Earth. JASM's profitability matters less as a financial event and more as evidence that this dispersion is actually viable. If overseas fabs can stand on their own, the global chip map looks different. If they can't, the world stays as dependent on Taiwan as it was five years ago.

For Japan, the local payoff is already visible. Kumamoto Prefecture's economic research group estimates roughly ¥6.9 trillion ($44 billion) in regional economic impact over a decade. Around 90 companies have set up or expanded operations in the prefecture since JASM arrived. Land prices in Kikuyo, the small town that hosts the fab, have risen by roughly 50%, and rental rates for single-occupancy housing have climbed past Kumamoto City's. JASM has set itself a target of sourcing 60% of its semiconductor materials domestically by 2030, up from 46% in 2025 — a number that ripples through the equipment and chemicals industries that were arguably Japan's last remaining strength in semiconductors.

Whether the Kumamoto Q1 profit is the start of a sustained run or a one-quarter blip is the question that will define the next year of Japan's chip story. For now, after fourteen months of red ink and very loud doubt, the world's biggest contract chipmaker has at least one Japanese fab that pays its own way.

How does the picture look from your side of the world? Is your country putting public money behind domestic chip production — and if so, would you rather see it backing a foreign operator like TSMC, a national champion like Rapidus, or both? We'd like to hear how this debate is playing out where you are.

References