Preferred Networks Files for Initial Public Offering in Tokyo

Strategic Shift Toward Public Markets

Preferred Networks Inc., the Tokyo-based AI startup, is preparing for an initial public offering (IPO) to finance the massive capital requirements of its semiconductor ambitions. The move represents a significant strategic pivot for the company, which has historically prioritized engineering independence over aggressive capital raising.

According to CEO Daisuke Okanohara, the company intends to go public once it achieves profitability, a goal set for three years from now. The capital raised through the offering is intended to fund the large-scale production of proprietary silicon, a necessity driven by the rising costs of staying competitive in the global AI hardware race. While a specific listing date or target capital raise has not been finalized, the company’s leadership indicated that an IPO will be necessary over the next few years to meet the scale required for mass manufacturing.

Technical Focus and Competitive Positioning

The company’s hardware strategy is centered on its MN-Core L series, a line of semiconductors specifically designed for generative AI inference. Unlike general-purpose graphics processing units (GPUs) produced by Nvidia Corp., which are widely used for both training and inference, Preferred Networks is focusing on low-latency AI processing optimized for running models on devices and robots.

By specializing in inference-optimized silicon, Preferred Networks aims to provide an alternative to Nvidia’s dominant AI accelerators. The company seeks to appeal to global customers looking to reduce procurement costs and lower their reliance on U.S.-based hardware providers. Okanohara noted that the primary challenge facing the company is not market demand, but rather securing sufficient manufacturing capacity. The MN-Core L series is slated to use a legacy node at Taiwan Semiconductor Manufacturing Co. (TSMC), with the company planning to deliver samples to customers in the first half of next year and a full commercial launch by December 2027.

Company Background and Market Context

Founded in 2006 by Daisuke Okanohara and Toru Nishikawa following their studies at the University of Tokyo, Preferred Networks has spent much of its history focused on research and industrial AI applications. The company has maintained a high valuation—reaching the billion-dollar mark—while relying on a concentrated group of strategic investors, including Toyota Motor Corp. and Fanuc Corp.

The shift toward a public market debut underscores the intensifying capital intensity of the AI sector. As the industry moves toward specialized hardware, Preferred Networks is also expanding its software ecosystem. Through the joint venture Noetra Corp., the company is working with major Japanese entities, including SoftBank Corp., Sony Group Corp., and Honda Motor Co., to develop domestic multimodal foundational models. This dual focus on hardware and software aims to provide a comprehensive domestic alternative within the AI ecosystem, mitigating risks associated with global supply chain volatility and geopolitical restrictions on hardware access.