Preferred Networks Targets IPO to Finance AI Chips and Japan’s Industrial Computing Push

Preferred Networks is pursuing a public listing to finance mass production of its own artificial-intelligence processors, CEO Daisuke Okanohara said. The move places a Japanese AI company’s capital needs at the center of a chip race dominated by enormous spending from global technology firms.

The company remains privately held, and the timing, exchange and valuation of a listing were not disclosed. Its strategy links fundraising directly to manufacturing scale: PFN wants its MN-Core processor family to support generative AI and industrial workloads while reducing dependence on outside accelerator suppliers. Bloomberg described the IPO effort as a response to the rising cost of staying competitive in AI hardware.

Preferred Networks began developing AI chips in 2016 to power its supercomputers and optimize energy consumption, according to Reuters. That effort has grown into a broader domestic computing stack spanning semiconductors, infrastructure, foundation models and applications.

The chip strategy addresses a basic constraint in generative AI: access to capable processors is expensive, capacity is limited, and software performance depends heavily on hardware design. PFN’s low-power MN-Core line, including the L1000 processor, is intended to support the company’s own systems and commercial deployments. A 2024 financing announcement said new capital would fund talent, development, production and sales for the processor family.

PFN has built its position through backing from Japanese industrial companies and financial institutions. Toyota and Fanuc are among its investors, giving the startup links to automotive manufacturing and robotics markets where AI must operate inside factories, machines and supply chains.

SBI Holdings planned an investment of about 10 billion yen, or $69 million, and a collaboration on AI chip design, Nikkei reported. PFN later announced 19 billion yen in equity and debt financing, with SBI Group leading the equity round. That funding provides a bridge to larger production, while an IPO would offer a broader pool of capital and a public market valuation for a business spanning chips, models and applications.

Preferred Networks has positioned AI as an industrial technology rather than a consumer chatbot business. Its stated focus covers manufacturing, robotics, materials, transportation and other real-world problems, with the company seeking to combine applications, machine-learning models and computing hardware.

That approach gives PFN a distinct route into the market. Toyota and Fanuc provide access to demanding industrial use cases, while the company’s chip work gives it control over more of the computing stack. CNBC described PFN as a Toyota-backed Japanese unicorn seeking global growth, and NextPlatform highlighted its role as an underrecognized Asian AI processor developer. The model also raises the cost of execution because PFN must develop silicon, software, manufacturing relationships and customer applications at the same time.

The IPO push reflects a capital race as much as a technology race. Training and serving advanced AI models require large processor deployments, and custom chips can improve efficiency only when a company can fund design, fabrication, software support and volume production.

PFN’s financing partners give Japan a domestic participant in that contest, but the startup faces competitors with deeper resources and established manufacturing ecosystems. Its advantage rests on specialization, low-power designs and industrial relationships rather than raw scale. SBI’s semiconductor entry, PFN’s work with major manufacturers and the company’s decision to develop a full stack show how Japan is assembling AI capacity around corporate partnerships.

The immediate test is whether PFN can convert private financing and strategic partnerships into repeatable chip production and customer revenue. An IPO would provide funds for that expansion, but public investors would also scrutinize manufacturing costs, processor adoption, software compatibility and the economics of competing with established accelerators.

PFN’s broader business will matter as much as the listing itself. The company is extending into security through a joint venture with GMO Internet Group and GMO Cybersecurity by Ierae, while its founders describe a strategy covering applications, models, infrastructure and semiconductors. If PFN proves that a Japanese startup can sell an integrated industrial AI stack, the listing could become a test of whether specialized hardware companies can scale without matching hyperscaler budgets.