FuriosaAI raises $620 million in Series D

FuriosaAI said it closed a $620 million Series D at an approximately $2.9 billion post-money valuation.[1] That is about four times the $735 million valuation in its previous round last year.[2] The National Growth Fund invested $280 million, making it the round’s largest investor.[1] Infinitum was the largest international investor, while other international participants included Cisco Investments, NeoFort Holding, Jericho Capital and Worldline Capital.[1] Existing shareholders added $125 million in follow-on investments.[1]

FuriosaAI’s three original seed investors—NAVER, DSC Investment and KDB Capital—also participated in the financing.[1] The Export-Import Bank of Korea selected FuriosaAI for its first direct investment in a startup.[1] UAMCO committed $30 million.[1] Together, these details show participation from existing backers and Korean institutions alongside international investors; the evidence does not specify the investment amounts for most participants.[1] For financing teams, the announced amounts distinguish the National Growth Fund’s position as largest overall investor from Infinitum’s position as largest international investor.[1]

FuriosaAI says it will use the proceeds to expand RNGD production, strengthen its software ecosystem and accelerate development of its third-generation accelerator.[1] The company expected to produce 20,000 RNGD accelerators in 2026 and planned to increase production to more than 40,000 units in 2027.[1] RNGD entered mass production in January 2026.[1] These figures are company expectations and plans, not reported production results.[1] Teams assessing supply should distinguish the stated production targets from confirmed output.

FuriosaAI develops AI accelerators for data-center inference.[1] Samsung SDS launched a commercial NPU-as-a-service offering powered by RNGD in July, while LG Group runs its EXAONE deployments on the accelerator.[1] Upstage uses RNGD to provide live search summarization to millions of users.[1] FuriosaAI says RNGD matches the throughput of Nvidia’s RTX PRO 6000 at the same card count and serves twice as many users per kilowatt of power.[3] Those performance comparisons are company claims; the evidence does not provide independent test results.[3]