How Long Will Kumamoto's Aftershocks Last? Japan's Official Forecast — and the One Chip Product With No Backup Factory

[July 29 update] The day-after status — TSMC gradually resuming, physical damage confirmed at Renesas — is covered in our day-2 report. Tokyo Electron — the single-site risk this article highlights — has since officially reported no major damage and expects only a minor earnings impact. Japanese-language coverage is also available: the day-1 report and the day-2 update.

[July 30 update] Three days on, the single-site risk this article flagged — Tokyo Electron's coater/developer tools — is still not back to full operation. See our day-3 report for why, and for the 2016 precedent it draws on.

This article does two things. First, it translates the official aftershock outlook from the Japan Meteorological Agency (JMA) — the government body that issues Japan's earthquake assessments — into plain terms for readers outside Japan. Second, it maps every major Kumamoto semiconductor product to its backup production sites, based on company disclosures and verified reporting. A companion piece covers the hour-by-hour status of each plant.

What exactly did Japan's Meteorological Agency say?

The JMA held two press conferences on July 28, at 5:30 p.m. and 8:30 p.m. local time. Its assessment was direct: seismic activity in the region is "considerably active." The agency warned that earthquakes registering up to intensity 7 — the same maximum level as the main shock — could occur for about one week, and that the risk is highest in the next two to three days (NHK, FNN, July 28, 2026). The warning was not hypothetical: by 8 p.m. on July 28, four aftershocks of intensity 5- or greater had already struck.

Two details help international readers calibrate this forecast. First, the "one week, especially 2–3 days" formula is the JMA's standard post-quake advisory; the agency issued the same language after the 2016 Kumamoto and 2024 Noto Peninsula earthquakes. Second, and more importantly, the 2016 Kumamoto sequence is the reason Japanese companies treat this warning seriously: on April 14, 2016, a M6.5 quake hit Kumamoto, and the true main shock — M7.3 — arrived two days later. In other words, the first big quake is not always the biggest. Because seismologists describe the July 28 rupture as a strike-slip event similar to 2016 (Nikkei, July 28, 2026), plant managers will not treat a quiet first night as an all-clear.

Therefore, the practical meaning of the JMA forecast for the chip industry is this: safety inspections and line requalification will proceed under the assumption that another violent quake may come within days. Restart timelines announced before the 2–3 day window closes should be read as provisional.

Which Kumamoto products can be made elsewhere?

Redundancy — whether the same product can be built at another site — differs sharply by company. The table summarizes the position as of July 28.

(Compiled from Nikkei xTECH, Tokyo Electron and company disclosures)

The pattern is counterintuitive. JASM attracts the most headlines because it carries the TSMC name, yet it is the easiest to substitute: the 22/28nm and 12/16nm processes it runs are mature nodes, and TSMC operates several times JASM's capacity on those nodes in Taichung and Nanjing. Consequently, if JASM stays down for weeks, TSMC's customers will get their chips — from Taiwan. What cannot be replaced is the strategic point of JASM itself: chips made in Japan, next to Sony's sensor fabs and Japan's automotive supply chain.

At the other extreme sits Tokyo Electron Kyushu. Its coater/developers apply and develop photoresist — the light-sensitive coating — around every lithography step, and Tokyo Electron holds roughly 90% of the global market for these machines, including near-total share for EUV-compatible models. However, the company's domestic production network divides cleanly by tool type: Miyagi makes plasma etchers, Yamanashi and Iwate make deposition systems, and coater/developers come only from Kumamoto. Therefore, a prolonged stoppage in Kumamoto would not merely dent one company's revenue; it would delay tool deliveries to fab construction projects in Taiwan, Korea, the US, Europe and Japan simultaneously. This is the single point of failure the industry should watch.

Sony occupies the middle, and its position is best understood through what happened in 2016. Sony's sensor fabs divide work by application: Nagasaki builds the high-volume smartphone sensors, while Kumamoto builds sensors for digital cameras, surveillance and industrial equipment, plus microdisplays for viewfinders. When the 2016 earthquake shut Kumamoto, the undamaged Nagasaki fab could not absorb Kumamoto's product mix, because transferring a sensor line requires requalification that takes months. As a result, camera makers worldwide ran short of sensors while Sony needed about three months for full recovery — and Sony itself later admitted its business-continuity plan had been insufficient (Kumamoto Nichinichi Shimbun). The same structural split exists today.

Two more single points deserve attention. Mitsubishi Electric's 8-inch SiC (silicon carbide — the material behind next-generation EV power chips) line in Kikuchi started up only in November 2025 and is the company's sole 8-inch SiC site. TOPPAN's Tamana plant makes on-chip color filters, a processing step that sits directly upstream of Sony's sensor output; TOPPAN's other plants make different products.

How long could recovery take, based on precedent?

The 2016 earthquake produced a clean natural experiment, because it hit the same companies in the same prefecture. The recovery times split by plant type, not by company size. Tokyo Electron Kyushu — an assembly operation without a wafer cleanroom — resumed production in stages about 10 days after the quake (Nikkan Kogyo Shimbun, 2016). By contrast, Sony's front-end fab needed roughly 1.5 months to restart wafer input and 3 months to recover fully. A front-end fab must recover cleanroom air quality, requalify hundreds of precision tools and scrap in-process wafers ruined by the power loss; an assembly plant must mainly confirm its building and supply of parts. Hence the 2016 rule of thumb: weeks for equipment plants, months for wafer fabs.

However, three factors argue against copying 2016's timeline onto 2026. First, Renesas showed in 2016 that a well-drilled business-continuity plan works — its Kawajiri fab recovered faster than peers because of procedures built after the 2011 Tohoku earthquake. Second, every company in the region has since reinforced buildings and raised buffer inventories, precisely because of 2016. Third, JASM's fab did not exist in 2016; it was built to post-2016 seismic standards. The 2016 precedent therefore defines the worst case, not the base case — provided the aftershock sequence does not deliver a second, larger shock.

What should global supply-chain readers watch this week?

Watch four things, in order of information value. First, company statements from July 29 onward: Mitsubishi Electric, Renesas and Sony had said nothing as of the night of July 28, and their first statements will reveal whether the region's least-substitutable lines — SiC and sensors — took damage. Second, the aftershock count: every intensity-5+ quake resets inspection clocks. Third, infrastructure restoration: the Kyushu Shinkansen stopped for the day, expressways closed in 23 sections, and about 48,000 households lost power; fabs cannot run without stable power, water and logistics (Nikkei, NHK, July 28, 2026). Fourth, Tokyo Electron's restart announcement specifically — for the reasons above, it matters more to the global industry than JASM's.

Key takeaways

• The JMA warns that intensity-7-class earthquakes could occur for about one week after the July 28 M7.1 quake, with peak risk in the first 2–3 days; four intensity-5- or stronger aftershocks had struck by 8 p.m. the same day.
• The 2016 precedent looms over every restart decision, because in 2016 the biggest quake came two days after the first. Early restart timelines should be read as provisional.
• TSMC's JASM output is the most substitutable: Taiwan and Nanjing hold several times its 22/28nm and 12/16nm capacity. The 'made in Japan' strategic value is what cannot move.
• Tokyo Electron Kyushu is the opposite: coater/developers with ~90% global share are made effectively only in Kumamoto, making it the industry's single point of failure in this event.
• Sony's 2016 experience — three months to full recovery, with no transfer to Nagasaki possible — shows why application-split fab networks provide less redundancy than they appear to.