The South Korean AI chip designer says it has about 20 billion won in orders, while analysts say more contracts and repeat buying are needed to turn pilot deployments into durable sales.
FuriosaAI, a South Korean AI semiconductor fabless company, is accelerating commercialization of its second-generation Renegade inference accelerator (chip for running AI models) with a plan to produce 20,000 units this year and generate 80 billion to 100 billion won in annual revenue. The company said on the 5th that it has secured about 20 billion won in purchase orders since mass production began five months ago and expects to ship 7,000 to 8,000 units to customers, while using the rest to build its ecosystem and reference base. Renegade has already been deployed by Samsung SDS Cloud, LG CNS, LG Uplus and Daum AI summarization service, and FuriosaAI also disclosed participation in a 15MW AI data center project in Sweden. It is also working with Broadcom on a third-generation AI accelerator using TSMC's 2-nanometer process and HBM4, targeting sample supply in the first half of 2028. Analysts and industry insiders say the harder test is winning additional contracts and repeat orders in a market still anchored by Nvidia's CUDA-based software ecosystem (Nvidia's AI development platform), even as demand for power-efficient NPUs (AI processing chips) grows with the expansion of AI inference. Separately, FuriosaAI is pursuing a total 800 billion won through South Korea's National Growth Fund and a pre-IPO round, though the funding process has drawn fairness criticism after policy funds were approved before private investment commitments reached target levels.