Japanese startup seeks to go public to keep up in AI chip race

By the Editorial Team · Wednesday, September 9, 2026

Tokyo-based startup Preferred Networks is planning an IPO to fund the scaling of production for its MN-Core L series AI chips, aiming for sample delivery in early 2027 and full commercial launch by the end of that year. The company seeks to challenge Nvidia’s dominance in AI accelerators by targeting low-latency inference applications, but faces manufacturing capacity constraints amid global chip supply shortages.

Tokyo-based startup Preferred Networks is preparing to go public as it seeks to scale up production of its custom-designed AI chips amid intensifying competition in the global semiconductor market. The company, valued at over $1 billion and one of Japan's few closely held tech firms of its size, is targeting the first half of 2027 to deliver samples of its latest semiconductor technology to customers, with full commercial launch planned by December of that year.

CEO Daisuke Okanohara outlined the company’s strategy in an interview, emphasizing the need to raise capital through an initial public offering (IPO) to meet the substantial financial demands inherent in AI hardware development. This marks a shift from Preferred Networks’ previous reliance on private funding from strategic investors such as Toyota Motor and Fanuc, highlighting the growing costs and scale required to compete, particularly against dominant players like Nvidia.

Preferred Networks aims to address customer demand for alternatives to Nvidia’s AI accelerators by offering its MN-Core L series chips, specifically designed for generative AI inference—focusing on executing rather than training AI models. The chips target low-latency AI processing suitable for applications in devices and robotics. While the company has developed AI hardware for roughly a decade, its semiconductor products have mainly been limited to research institutions and university labs until now.

A critical challenge for the company lies in securing manufacturing capacity as global foundries face strained resources amid a surging demand for AI-related chips. Preferred Networks has encountered significant difficulties obtaining sufficient production volumes, with many firms globally competing for priority at chip fabrication facilities. The company plans to manufacture the MN-Core L series using an established semiconductor process node at Taiwan Semiconductor Manufacturing Co. (TSMC). Okanohara noted that the risk is not market acceptance but the ability to produce enough units to satisfy demand.

In addition to semiconductor development, Preferred Networks is advancing large-language models tailored for industrial applications. It collaborates with major Japanese corporations including SoftBank, Sony Group, and Honda Motor through a joint venture called Noetra, focused on building a domestic multimodal foundational AI model.

As the AI hardware sector expands rapidly, Preferred Networks’ move to seek public funding reflects broader trends in the industry, where escalating capital investment and competitive pressures are reshaping the market landscape for semiconductor startups worldwide.

This article was reported by synthesizing coverage from multiple independent news outlets to present a balanced, unbiased account. Read about our editorial standards. Spotted an error? Let us know.