In a recent development, European startup Vybium is working on developing an AI accelerator chip using the open RISC-V instruction set architecture. The company aims to compete with Nvidia A100 GPU in the data center by offering RISC-V AI accelerators for various verticals. Vybium, founded by VRULL in Austria and Software Ecosystem Solutions in Albania, has licensed the NPU IP from Stream Computing to speed up chip development. They are also incorporating new data types, support for sparsity, and higher-bandwidth memory solutions. The initial focus of Vybium's products will be on AI/ML accelerator cards to compete with Nvidia A100 family. They also have plans to develop systems that combine RISC-V general-purpose compute with AI/ML for embedded, industrial, and edge use cases [ba77faad].
This development adds to the ongoing advancements in AI hardware, with companies like NeuReality and Ceva introducing innovative solutions for AI processing. Vybium's use of the open RISC-V instruction set architecture provides an alternative to traditional AI hardware solutions and offers the potential for increased competition in the market. These developments contribute to the progress of AI technology and its applications in various sectors [ba77faad] [bbd8aad9] [3cbe8719].