FuriosaAI to attract investment for powerful AI chips
Seoul-based startup FuriosaAI, a developer of AI inference chips and data center solutions, plans to raise over $300 million in a Series D round ahead of its IPO.
As reported by Bloomberg, the company has already approached several international banks to prepare an optimal deal structure and is anticipating high interest from global institutional investors.
FuriosaAI was founded in 2023 by its CEO, June Paik, who previously worked at Samsung Electronics and AMD and has extensive experience integrating neural networks into corporate business processes.
Over the past two years, the startup has completed several funding rounds, raising $125 million in a Series C bridge round in July 2025. Total investment since the company's founding has reached approximately $246 million, with its last publicly announced valuation at about $735 million.
In early 2025, the company turned down an $800 million buyout offer from Meta Platforms, a move that underscores FuriosaAI's strategy to maintain independence and pursue its ambitious development plans.
The Seoul-based manufacturer's key product is the RNGD (Renegade) NPU accelerator, which is designed for computer vision, large language models (LLMs), and multimodal AI systems. Its performance reaches up to 512 teraflops in FP8 format, with a TDP of about 180 watts. These capabilities have attracted the interest of OpenAI, leading to a partnership agreement.
In 2025, FuriosaAI signed a landmark contract with LG AI Research. After a seven-month testing period, the RNGD chip was approved for LLM Exaone inference. The RNGD delivered a 2.25x advantage in performance per watt compared to the GPU solutions previously used by LG.
A single FuriosaAI server can house up to 8 RNGD accelerators in a 4U chassis. A standard 15-kilowatt rack can accommodate up to 5 of these servers, allowing clients to efficiently scale their existing capacities.
The Series D round and subsequent IPO are aimed at expanding FuriosaAI's global market presence, scaling production, and growing its client base among the largest technology and industrial companies.
As global competition in the AI infrastructure sector intensifies, FuriosaAI is betting on highly efficient, specialized chips for practical image and text recognition, tailored to the real-world tasks of large corporate clients.
