Rebellions said on the 16th that it has formed a partnership with Japanese neocloud provider ai& to supply more than 100 neural processing unit (NPU) racks to ai&'s data center in Tokyo.
ai& is a neocloud provider that supplies artificial intelligence computing resources such as graphics processing units (GPUs). The company decided to adopt Rebellions' "REBEL Rack" as it builds heterogeneous AI computing infrastructure at its Tokyo data center. REBEL Rack is a rack-level product built around servers equipped with Rebellions' NPU, the REBEL100. Japanese public institutions and companies will be able to use AI inference servers based on Rebellions' NPUs. ai& plans to operate five data centers by the end of this year and expand total capacity at the Tokyo data center to 40 megawatts by the end of next year.
ai& customers will be able to use Rebellions' NPUs within their existing ai& development environment. The ai& technical team uses open-source software widely adopted in GPU environments, and Rebellions also supports major open-source frameworks, allowing customers to keep their current development setup. The two companies expect that customers will be able to use NPU inference services without switching environments, meaning inference services can go live as soon as the REBEL Racks are delivered.
"This partnership allows Rebellions to broaden its business base in Japan," Rebellions CEO Park Sung-hyun said. "We will provide more efficient inference infrastructure to Japanese public institutions, companies and developers."
