FuriosaAI, a Seoul-based designer of chips that run artificial-intelligence models, raised $620 million in a Series D round that values the company at about $2.9 billion, roughly four times its valuation a year ago.
South Korea’s National Growth Fund was the largest investor, committing $280 million. Foreign investors put in about $110 million, led by U.S. firm Infinitum and including Jericho Capital, Worldline Capital, Cisco Investments, Abu Dhabi’s NeoFort Holding and Singapore’s QUDA Capital Partners. Existing shareholders added $125 million, among them seed investors Naver, DSC Investment and KDB Capital. The Export-Import Bank of Korea made its first direct investment in a startup, and asset manager UAMCO committed $30 million.
An inference play
FuriosaAI’s previous round valued it at $735 million. The jump reflects investor appetite for alternatives to Nvidia in inference, the work of serving trained models to users, where cost and power use matter more than raw training speed. The company’s RNGD chip, which entered mass production in January, draws 180 watts and runs on air cooling. Taiwan Semiconductor Manufacturing Co. is producing about 20,000 of the chips this year, and FuriosaAI says it has secured memory and factory capacity for more than 40,000 in 2027.
Customers include Samsung SDS, which launched a cloud service on RNGD in July, and LG Group, which runs its EXAONE models on the chips. The company is also courting buyers in the Asia-Pacific region and the Middle East.
Next chip in the works
Proceeds will fund chip production, software and a third-generation processor being developed with Broadcom Inc. and other partners using a 2-nanometer, multi-die design with HBM4 memory. A worldwide shortage of high-bandwidth memory remains the main constraint on how quickly smaller chipmakers like FuriosaAI can scale output.
