FuriosaAI, the Seoul based maker of AI inference accelerators, has closed a Series D round of 620 million dollars, or 820 billion won, at a post-money valuation of approximately 2.9 billion dollars. The valuation is nearly four times the level of its previous round in July 2025.
South Korea’s National Growth Fund is the largest investor, contributing 280 million dollars as part of the government’s programme to build a domestic AI semiconductor industry. International investors contributed 110 million dollars in total, including Cisco Investments, Abu Dhabi based NeoFort Holdings, Singapore based QUDA Capital Partners, and US firms Infinitum, Jericho Capital and Worldline Capital. Existing backers including the Korea Development Bank, Kakao Investment and NAVER also participated.
The money funds production of RNGD, FuriosaAI’s second generation inference accelerator, which entered mass production in January 2026 and is built on its Tensor Contraction Processor architecture. The company expects to produce 20,000 RNGD accelerators in 2026 and more than 40,000 in 2027, and will use part of the round to develop a third generation accelerator, being built with Broadcom on a 2 nanometre process. Customers named by the company include LG, which runs its EXAONE models on RNGD, Upstage and Samsung SDS.
The pitch is power. Inference, the running of trained models, is the fastest growing part of data centre compute, and FuriosaAI sells efficiency per watt against Nvidia’s general purpose GPUs. A state anchored cap table and a funded production ramp do not by themselves win sockets against Nvidia’s software ecosystem, but they remove the financing question that has ended earlier challengers, and they put a Korean national champion into the inference market at production scale.