FuriosaAI targets up to $500m funding for AI chip expansion

FuriosaAI, a South Korean startup specialising successful AI spot design, is reportedly aiming to unafraid betwixt $300m and $500m successful a Series D backing round, arsenic it prepares for an archetypal nationalist offering (IPO).

This backing circular has Morgan Stanley and Mirae Asset Securities arsenic its co-advisers, according to a study successful Bloomberg. It aims to bolster FuriosaAI's efforts to vie with manufacture person Nvidia.

The funds volition enactment the wide accumulation of its second-generation RNGD chip, grow its planetary concern operations, and assistance successful the improvement of a third-generation chip. The startup is considering a nationalist listing arsenic aboriginal arsenic 2027.

FuriosaAI expects the archetypal mass-produced shipment of its RNGD AI chips from TSMC aboriginal this month.

The company, founded successful 2017 by June Paik, who has acquisition with Samsung Electronics and Advanced Micro Devices (AMD), focuses connected high-efficiency AI inference chips.

FuriosaAI claims that its RNGD spot provides 2.25 times the inference performance-per-watt of accepted graphics processing units (GPUs).

In a erstwhile backing effort successful July 2025, FuriosaAI raised $125m successful a Series C round, expanding its full backing to $246m astatine that time. Participants successful this circular included Korea Development Bank, Industrial Bank of Korea, Keistone Partners, PI Partners, and Kakao Investment.

The Series C round’s nonsubjective was to conscionable increasing request for FuriosaAI's flagship AI spot and accelerate its roadmap.

The company's ngo is to marque AI computing much sustainable by addressing the inefficiencies of power-hungry graphic processing units (GPUs). FuriosaAI claims that its exertion offers a much businesslike alternate for AI compute scalability.

In July 2025, LG AI Research adopted the RNGD spot for its EXAONE instauration models, reporting a 2.25 times amended performance-per-watt compared to bequest GPUs.

The RNGD chip's show stems from its Tensor Contraction Processor (TCP) architecture, co-designed with a bundle stack specifically tailored for tensor contraction, which is indispensable to heavy learning computations. This attack provides businesslike customized dataflow paths for computations, reducing unnecessary information question and resulting successful little costs and much flexible deployments.

In March 2025, FuriosaAI reportedly declined an $800m acquisition connection from Meta. According to a study successful South Korean work Maeil Business Newspaper, the negotiations broke down owed to disagreements implicit post-acquisition concern strategy and organisational structure, alternatively than issues related to the offer's price.