SuperX will introduce its 1.6T optical module solution at Interop Tokyo 2026 and plans to show a broader set of AI data center (AIDC) infrastructure products spanning optical interconnects, AI compute, modular “AI factory” designs, and an 800 V DC power architecture. The company is tying the 1.6T optics to the bandwidth requirements of large-scale AI training and inference networks, where fabric capacity and link density can become the limiter before GPU supply does.
The 1.6T optical module solution is built on a full DSP architecture with silicon photonics integration. SuperX says it’s intended for deployment across both Ethernet and InfiniBand AI infrastructures, targeting high-bandwidth connectivity in AI-scale clusters. The optics effort follows the establishment of SuperX Optical Communications, a joint venture focused on end-to-end optical solutions for next-generation AI data centers.
On the compute side, SuperX will show several systems built around NVIDIA platforms: the SuperX XN8161-B300 AI servers based on the NVIDIA HGX B300 platform; a high-density compute platform based on NVIDIA RTX PRO 6000 Blackwell GPUs; and a rack-scale SuperX GB300 NVL72 System based on the NVIDIA GB300 Grace Blackwell Ultra Superchip. The mix matters operationally because it implies SuperX is aiming to cover everything from large training footprints to inference-heavy deployments where capacity planning often looks more like “many smaller pools” than one monolithic cluster.
For facility design and deployment, SuperX’s modular AI factory concept centers on pre-validated, full-stack reference architectures integrating compute, power, cooling, and networking. The company says those reference designs are compatible with the NVIDIA Blackwell Ultra GB300 NVL72 platform, and that the modular approach can shorten time-to-market to 6–9 months. SuperX also describes an expansion path from a 2.5 MW initial pod to 20 MW clusters and up to 80 MW campuses.
Power is another focus: SuperX will showcase its Medium Voltage Rectifier (MVR) 800 V DC power architecture for high-density AI data centers. The company describes the design as modular and highly integrated, with goals including more efficient power distribution, lower conversion complexity, reduced power loss, stable operation, and flexible configuration for different sites and expansion plans. 800 V DC architectures are getting more attention as rack densities rise, but the details—where rectification happens, how distribution is protected, and how the plant is maintained—tend to decide whether the theoretical efficiency wins show up in real operations.
SuperX also highlighted its local operations in Japan, including a Global Supply Center in Tsu City, Mie Prefecture. The company lists an annual production capacity of up to 20,000 AI servers for that facility and says its Japan presence supports localized delivery, faster response, and long-term operational support.
“With our established presence in Japan, including our Global Supply Center in Tsu City, Mie Prefecture, we are well positioned to provide localized delivery, faster response, and long-term operational support,” said Aiko Furukawa, CEO of SuperX Industries Co. Limited, SuperX’s wholly owned subsidiary in Japan.
Source: SuperX AI Technology Limited
