FuriosaAI, founded in 2017, designs AI chips built for efficiently running advanced models. The Seoul-based company specialises in AI chip design, Tensor Contraction Processor architecture, and deep learning hardware acceleration, with a focus on sustainable AI computing. Its team includes engineers with backgrounds at Samsung, AMD, and other leading semiconductor firms.
The company's product lineup spans hardware and software. Its second-generation RNGD accelerator targets high-performance large language model and multimodal inference, operating at just 180W and currently sampling with global technology partners. A first-generation Vision NPU has shipped into production. FuriosaAI also provides a software stack comprising a compiler, runtime, and integrations with PyTorch and Kubernetes.
