AINEWS
Japan's Ministry of Economy, Trade and Industry is co funding Tower's bet that a shuttered Panasonic era plant can return as a silicon photonics line (chips that move data with light rather than electrical signals) by late 2027, faster than
Tower Semiconductor is committing up to $3 billion to reopen a Panasonic-era chip plant that has been dark since 2022. The Israeli specialty foundry and Japan's METI are jointly making a bet on a faster path to AI-era chip capacity than building from scratch.
The shuttered plant sits in Arai, in Niigata Prefecture, and will be reborn as "Fab 6": a 300mm production line for silicon photonics, a category of chips that move data with light rather than electrical signals, and for advanced optical packaging. Production is targeted for the fourth quarter of 2027. The investment is net of grants from METI, Japan's Ministry of Economy, Trade and Industry, which is co-funding the buildout to anchor specialty capacity on Japanese soil.
The bet rests on a claim from Tower CEO Russell Ellwanger. Greenfield construction or a fab acquisition, he argued, "typically require years of process development, customer [qualification]." Reopening an existing plant avoids that timeline. The plant already exists. What is missing is the front-end equipment and the customer pipeline.
That arbitrage is only possible because of a separate deal Tower closed in March 2026. The Israeli foundry took full ownership of the running 300mm line, Fab 7 in Uozu, by restructuring its TPSCo joint venture with Nuvoton Technology Corporation Japan, the 200mm operation's new owner. NTCJ paid Tower $25 million in the deal and absorbed the 200mm business; the joint venture is scheduled to close on April 1, 2027. Without that restructuring, Tower could not have committed to a $3 billion buildout on land it did not solely control.
The plant ran exclusively for Nuvoton Technology Corporation Japan before it went dark in July 2022. With NTCJ now redirected to the 200mm business, the 300mm footprint is free to take new specialty customers. Silicon photonics is the load-bearing bet. A domestic Japanese source is, for METI, a strategic asset as well as an industrial one.
The numbers Tower is publishing are aggressive. The company raised its 2028 business-model targets to roughly $3.6 billion in revenue and $1.2 billion in net profit, both figures company guidance rather than third-party validated. Both rest entirely on the "Track One" plan: the Arai restart plus expansion at the existing Uozu line. A second, "Track Two" project, a brand-new 300mm fab adjacent to Fab 7 that Ellwanger described as "highly accretive beginning in 2029," has no signed agreements and is excluded from the 2028 numbers. Calling this a two-track expansion overstates the commitment. Track Two is a placeholder, not a contract.
The risk is that silicon-photonics demand is more concentrated than the company is crediting, and the 2028 revenue and net-profit targets depend on those buyers qualifying Tower's new line on the timelines the company has set. Ellwanger's argument, that brownfield reuse buys speed, is testable. Tower's 2028 numbers will show whether the bet held.
Tower's Japan footprint dates to 2014, when it bought a 51% stake in Panasonic's three-fab operation, with Panasonic selling the remaining stake to Nuvoton in 2020. The current buildout is not a return to that relationship. It is the foundry taking direct ownership of the 300mm line it had been sharing. The Arai restart is the first major use of that ownership.