Preferred Networks (PFN) is a Japanese deep-learning company founded on 26 March 2014 by Toru Nishikawa and Daisuke Okanohara. It became Japan’s most visible private AI lab, originally focused on applying deep learning to the industrial internet of things in transportation, manufacturing, and healthcare, working with partners such as Toyota and the machine-tool maker FANUC.
PFN is best known internationally for Chainer, an open-source deep-learning framework released in 2015 that introduced the “define-by-run” approach to building neural networks, an idea that influenced later frameworks. The company has since pursued vertical integration, operating its own supercomputers and co-developing the MN-Core AI processor, and building Japanese-capable large language models in its PLaMo line.
Why a business reader should care: Preferred Networks shows that serious AI capability has long existed outside the US and China, and its move into custom silicon and sovereign Japanese models reflects the same drive for technological independence now seen across many national ecosystems.