
FuriosaAI is accelerating the commercialization of its second-generation neural processing unit (NPU) "RENEGADE" through partnerships with Samsung SDS and others. Fellow NPU startups DeepX and Mobilint are also pursuing revenue by building on real-world deployment cases. Industry observers say Korean startups entering mass production of second-generation NPUs now face a critical test of product commercialization this year.
FuriosaAI held the "RENEGADE 2026 Summit" in Gangnam, Seoul, on Monday, unveiling the chip's mass production status and collaboration progress with clients. "The key to future AI data center design lies in how much we can reduce repetitive inference costs," CEO Baek Jun-ho said. "We will achieve large-scale adoption and expand global revenue to meet the explosive market demand for high-performance, high-efficiency AI semiconductors."

RENEGADE, which FuriosaAI began mass-producing in January, is a chip that delivers lower power consumption and higher efficiency compared with graphics processing units (GPUs). Tests showed it can handle up to 7.4 times more simultaneous users than Nvidia's RTX Pro 6000 at the same power level. FuriosaAI plans to produce 20,000 RENEGADE-equipped cards this year.
At the event, FuriosaAI also revealed that Samsung SDS will launch a subscription-based NPU-as-a-Service (NPUaaS) powered by RENEGADE in July. This marks the first time a domestic cloud service provider has offered a Korean-made NPU as a cloud subscription product. Subscribing customers will be able to use RENEGADE on a pay-as-you-go basis. For FuriosaAI, the deal opens a channel to reach enterprise clients through Samsung SDS. The move is seen as a strategy to expand its customer base beyond existing partners such as LG AI Research and LG Uplus.
Other NPU startups are making similar moves. Mobilint received a 3 billion won investment in February from the POSCO DX Corporate Venture Capital New Technology Investment Fund, co-established with POSCO DX and POSCO Technology Investment. Following the investment, Mobilint plans to integrate its NPU into "PosMaster," an industrial control system developed by POSCO DX.
DeepX is pursuing the deployment of its NPU in Lotte Innovate's intelligent transportation system (ITS). The plan involves installing AI edge cameras equipped with NPUs in high-traffic zones to verify real-time detection of abnormal situations on-site. DeepX aims to expand pilot deployments of its chips into other Lotte business areas. Rebellions, which secured 250 billion won from the National Growth Fund, is deploying its NPU in SK Telecom's "A." (A-dot) service.
Industry officials agree that as revenue competition among domestic NPU startups intensifies this year, real-world usage experience has become critical. While NPU startups previously validated performance through first-generation chips, they have now reached the stage where they must generate revenue from actual industrial deployments. "Domestic NPU startups have received significant investment through government support and venture capital," an industry official said. "Now is the time for these companies to show results from that investment rather than continuing to rely on government backing."
