By 
SEOUL, South Korea — South Korean semiconductor startup FuriosaAI has secured $620 million in Series D funding to expand production of its energy-efficient artificial intelligence inference processors and accelerate international deployment. Announced on October 28, 2026, the financing values the company at approximately $2.9 billion after the investment. The round includes participation from international investors and South Korean financial institutions, reflecting growing interest in alternatives to conventional graphics processing units for large-scale AI computing.
FuriosaAI Expands Production of RNGD Processors
FuriosaAI develops specialized processors designed for AI inference, the process of using trained artificial intelligence models to generate responses and perform tasks. Its second-generation accelerator, known as RNGD, is designed to deliver high computing performance while limiting electricity consumption. The company says its processors are particularly suited to enterprise applications that require repeated execution of AI models.
The RNGD accelerator is available as an air-cooled PCIe card with a stated power consumption of approximately 180 watts. FuriosaAI also offers a server configuration containing eight RNGD cards, designed to operate within existing enterprise data center infrastructure. The company argues that energy efficiency can help customers reduce the operating costs associated with large-scale AI inference deployments.
Funding Supports Manufacturing and Global Expansion
FuriosaAI says RNGD entered mass production in January 2026 and expects to manufacture approximately 20,000 accelerator cards during the year. The company has secured manufacturing and memory capacity intended to more than double production in 2027. Expanding supply will be important as the startup seeks to serve customers operating AI cloud platforms, private data centers, and sovereign computing infrastructure.
South Korea’s National Growth Fund committed approximately 370 billion won, equivalent to about $280 million, making it a significant participant in the financing. International investors include Cisco Investments, Infinitum, Jericho Capital, NeoFort Holding, QUDA Capital, and Worldline Capital. Additional participation came from established Korean investors and financial institutions.
AI Inference Creates Demand for Specialized Chips
The growth of artificial intelligence has increased demand for computing hardware capable of operating trained models efficiently. While model training requires substantial computing resources, inference workloads can generate recurring infrastructure costs because deployed applications may serve millions of requests. Companies therefore have incentives to improve the amount of useful computing work performed for each unit of electricity consumed.
FuriosaAI positions RNGD as a processor optimized for these requirements. According to company-reported performance figures, its server configuration can deliver competitive throughput while serving more users per kilowatt under selected workloads. Actual performance and economic benefits will depend on the models, software configurations, and deployment conditions used by customers.
Startup Builds Enterprise Customer Relationships
FuriosaAI reports that Samsung SDS has introduced an AI processing service powered by RNGD, while LG Group companies are deploying the processors in hybrid cloud environments. These deployments provide opportunities for the company to demonstrate its technology in production settings rather than relying exclusively on laboratory benchmarks. Continued customer adoption will be important as the startup competes with established semiconductor suppliers.
The latest financing provides FuriosaAI with additional resources to scale manufacturing, expand its software ecosystem, and develop future processor generations. Its progress reflects growing demand for energy-efficient AI infrastructure as enterprises seek to manage the costs and power requirements associated with deploying advanced artificial intelligence systems.
Comments