SUPX•SuperX will unveil its 1.6T optical module solution and full-stack AI compute platforms at Interop Tokyo from June 10–12, including NVIDIA HGX B300 and Grace Blackwell Ultra systems. The modular AI factory cuts deployment time to 6–9 months and the Tsu City center produces up to 20,000 servers annually.
SuperX will debut its 1.6T optical module solution built on a full digital signal processor architecture and silicon photonics integration, supporting both Ethernet and InfiniBand networks for AI-scale connectivity. This follows the formation of SuperX Optical Communications joint venture to deliver end-to-end optical modules for large-scale training and inference environments.
The company will display its XN8161-B300 servers powered by NVIDIA HGX B300 for intensive training, high-density platforms with NVIDIA RTX PRO 6000 Blackwell GPUs for versatile AI workloads, and rack-scale GB300 NVL72 systems with the NVIDIA GB300 Grace Blackwell Ultra Superchip for large-scale model development.
SuperX’s modular AI factory integrates compute, power, cooling and networking in a pre-validated full-stack design, reducing deployment time to 6–9 months and enabling scalable expansions from 2.5MW pods up to 80MW computational campuses.
The 800V DC Medium Voltage Rectifier architecture enhances power distribution efficiency and supports flexible configuration for high-density AI data centers, while the Tsu City supply center offers localized delivery and can produce up to 20,000 AI servers annually to accelerate customer deployments.