FuriosaAI, a South Korean startup specialising in AI chip design, is reportedly aiming to safe between $300m and $500m in a Sequence D funding spherical, because it prepares for an preliminary public providing (IPO).
This funding spherical has Morgan Stanley and Mirae Asset Securities as its co-advisers, in response to a report in Bloomberg. It goals to bolster FuriosaAI’s efforts to compete with business chief Nvidia.
Entry deeper business intelligence
Expertise unmatched readability with a single platform that mixes distinctive information, AI, and human experience.
The funds will assist the mass manufacturing of its second-generation RNGD chip, increase its international enterprise operations, and assist within the improvement of a third-generation chip. The startup is contemplating a public itemizing as early as 2027.
FuriosaAI expects the primary mass-produced cargo of its RNGD AI chips from TSMC later this month.
The corporate, based in 2017 by June Paik, who has expertise with Samsung Electronics and Superior Micro Gadgets (AMD), focuses on high-efficiency AI inference chips.
FuriosaAI claims that its RNGD chip offers 2.25 occasions the inference performance-per-watt of conventional graphics processing models (GPUs).
In a earlier funding effort in July 2025, FuriosaAI raised $125m in a Sequence C spherical, growing its whole funding to $246m at the moment. Individuals on this spherical included Korea Growth Financial institution, Industrial Financial institution of Korea, Keistone Companions, PI Companions, and Kakao Funding.
The Sequence C spherical’s goal was to satisfy rising demand for FuriosaAI’s flagship AI chip and speed up its roadmap.
The corporate’s mission is to make AI computing extra sustainable by addressing the inefficiencies of power-hungry graphic processing models (GPUs). FuriosaAI claims that its know-how gives a extra environment friendly various for AI compute scalability.
In July 2025, LG AI Analysis adopted the RNGD chip for its EXAONE basis fashions, reporting a 2.25 occasions higher performance-per-watt in comparison with legacy GPUs.
The RNGD chip’s efficiency stems from its Tensor Contraction Processor (TCP) structure, co-designed with a software program stack particularly tailor-made for tensor contraction, which is important to deep studying computations. This method offers environment friendly customized dataflow paths for computations, lowering pointless information motion and leading to decrease prices and extra versatile deployments.
In March 2025, FuriosaAI reportedly declined an $800m acquisition supply from Meta. In keeping with a report in South Korean publication Maeil Enterprise Newspaper, the negotiations broke down as a consequence of disagreements over post-acquisition enterprise technique and organisational construction, fairly than points associated to the supply’s value.
In September 2025, FuriosaAI launched the NXT RNGD Server, a system that integrates its RNGD accelerators right into a turnkey resolution designed for environment friendly AI inference deployment in current information centres.
The NXT RNGD Server is engineered to supply excessive efficiency on key AI workloads whereas becoming seamlessly into present information centre setups.
It comes with the Furiosa SDK and Furiosa LLM runtime pre-installed, enabling rapid software activation. This method optimises customary PCIe interconnects, eliminating the necessity for proprietary materials or particular infrastructure, making it a sensible and cost-efficient choice for enterprises to scale their AI operations inside their current services.
Join our day by day information round-up!
Give your online business an edge with our main business insights.
