Panasonic Industry to Invest About ¥92 Billion in AI Server Materials and GPU Power Capacitors

Panasonic Industry, the electronic components and materials arm of the Panasonic Group, said on October 30 that it will invest about ¥92 billion in businesses tied to AI infrastructure, according to the company’s announcement and a report by Nikkei. The money goes into three areas where AI servers are driving demand: circuit-board materials, capacitors for GPU power circuits and next-generation materials for semiconductor packaging.

The largest share, about ¥67 billion, will expand capacity for MEGTRON, the company’s low-loss multilayer circuit-board materials used in AI servers and high-speed network equipment. A second building at its plant in Suzhou, China, costing about ¥30 billion, is scheduled to start operating in August 2028, and a second building in Ayutthaya, Thailand, costing about ¥37 billion, is set for April 2029. The first new plants are due to begin production in December 2026 in Suzhou and November 2027 in Ayutthaya.

Another ¥17 billion will raise output of SP-Cap conductive polymer capacitors, which keep power stable in GPU circuits. Capacity will grow at Uji in Kyoto Prefecture and on Batam in Indonesia, where a new five-storey building of about 50,000 square metres is to be built from May 2027 and start operating in April 2029. A further ¥8 billion will pay for a research building at the Yokkaichi South factory in Mie Prefecture to develop packaging materials for larger AI chips and chiplet designs, which combine several smaller chips in one package.

What it means in practice

The plan shows how Japanese suppliers are positioning themselves in the AI hardware boom beyond the chips themselves. Other domestic companies are working on the same buildout, as when Hitachi unveiled a GPU cold plate meant to cut data-centre cooling power by 30% and when Sharp opened domestic orders for Foxconn-built Nvidia AI servers.

Most of the new production capacity sits outside Japan, close to the board and server makers that buy the materials, while the packaging research stays in Mie. The timetable is long, with key phases running into 2029, which leaves the plan exposed if spending on AI servers cools before the new buildings are ready.

What to watch next is whether the first Suzhou plant reaches mass production on schedule in December, how quickly customers adopt materials from the new research building, and whether Panasonic gives sales targets for its AI-related business.