SAN JOSE, Calif., Oct. 14, 2026 /PRNewswire/ -- UfiSpace, a global leader in open disaggregated networking solutions, today announced its technology showcase for the OCP Global Summit 2026 (Booth D19). As next-generation AI workloads push hyperscale data centers toward extreme power densities and strict thermal constraints, UfiSpace is featuring a live engineering demonstration of its 1.6T liquid-cooled AI networking platform, proving thermal feasibility and system-level readiness for future high-density AI fabrics.
Solving the 1.6T Thermal Challenge: Dual-Track Tomahawk 6 Innovations
At the center of UfiSpace's booth showcase is a live demonstration featuring the S9331-64HO3L, a 102.4Tbps 2OU ORv3 liquid-cooled platform powered by the Broadcom Tomahawk 6 ASIC. Equipped with 64 x 1.6T OSFP ports, the platform operates with a Cooling Distribution Unit (CDU) to deliver real-time cooling and validate system-level thermal feasibility for next-generation AI fabrics. Positioned as a validation-ready early access architecture, the S9331-64HO3L provides hyperscalers and system integrators with a live, operational proof-of-concept for the thermal and mechanical architectures required to build 2027/2028 AI clusters.
Recognizing the transitional realities of data center infrastructure, UfiSpace is also exhibiting the S9331-64HO, a 3RU air-cooled 102.4Tbps platform. Designed for hyperscalers with advanced, highly optimized air-cooled infrastructures, it offers a high-performance alternative to scale bandwidth in targeted zones. By adopting this open, disaggregated architecture, operators can maximize bandwidth density and optimize Total Cost of Ownership (TCO) while establishing a unified operational blueprint for broader liquid-cooled rollouts.
Enabling Scale-Across AI Interconnects with Deep-Buffer Routers
To address lossless packet delivery across distributed AI data centers (Scale-Across / DCI), UfiSpace is highlighting its next-generation router portfolio powered by Broadcom Qumran ASICs:
- S9630-32HO (Qumran 4D): Powered by next-generation routing silicon, this 3RU router delivers 51.2Tbps switching capacity with 32GB of embedded deep packet buffering and 32 x 1.6T OSFP ports. Designed to eliminate buffer bloat and packet drops in long-distance AI cluster communications, it serves as an ultra-high-capacity gateway for disaggregated AI fabrics.
- S9620-32E (Qumran 3D): A 2RU 25.6Tbps deep-buffered platform featuring 32 x 800G QSFP-DD interfaces with OpenZR+ coherent optics support, providing versatile edge and spine routing for multi-site AI infrastructures.
Dedicated AI Fabric 800G AI Foundation
Rounding out the showcase is UfiSpace's S9322-64EO, a high-density 800G switching platform engineered specifically for demanding AI workloads. Powered by the Broadcom Tomahawk Ultra ASIC, the S9322-64EO (64 x 800G OSFP) delivers enhanced cognitive routing and ultra-low latency performance tailored for AI backend fabric interconnects. By adhering strictly to open networking specifications, the S9322-64EO streamlines the deployment of high-radix GPU clusters, ensuring uncompromised throughput for intensive AI training workloads.
Integrated Ethernet–Optical Switching Solution for AI Scale-Out and Scale-Across Infrastructure
UfiSpace and GeNopsys Technologies are joining forces to advance AI scale-out and scale-across infrastructure. The solution integrates UfiSpace's high-performance S9321-64E, an 800G Tomahawk 5 Ethernet switch featuring 64 × 800G QSFP-DD ports, with GeNopsys' dynamic OCS technology, creating a complete vertical networking platform for next-generation AI infrastructure. UfiSpace's hardware-driven lossless engine, leveraging SmartBuffer and Dynamic Load Balancing, provides Ethernet interoperability and zero-packet-drop performance. By eliminating O-E-O conversion, GeNopsys' all-optical switching enables high throughput, shorter AI training time, and lower total AI data center energy consumption. By combining Ethernet switching with the fundamental benefits of optical switching, the integrated solution delivers ultra-low latency, significant power savings, massive bandwidth, and seamless scalability for AI scale-out and scale-across networks, including long-distance DCI.
Maria Yuang, Founder and CEO of GeNopsys, said: "Our collaboration with UfiSpace brings together the flexibility of Ethernet and the transformative impact of optical switching for AI scale-out and scale-across. Together, we can accelerate AI training and significantly reduce total data center energy consumption—two critical challenges facing today's AI data centers."
Driving Open Ecosystem Interoperability
Vincent Ho, CEO of UfiSpace said, "The transition to 1.6T AI networking requires more than just raw bandwidth; it demands complete ecosystem alignment across silicon, optics, thermal management, and NOS partners. At OCP Global Summit 2026, we are demonstrating that our open, disaggregated platforms are engineered to solve the real-world thermal and architectural bottlenecks of next-generation AI infrastructure. By validating liquid cooling feasibility today, we enable our ecosystem partners and customers to design their future AI data centers with absolute confidence."
Event Details & Live Demonstration
UfiSpace invites industry analysts, partners, and customers to experience the TH6 Liquid Cooling Live Demo and explore the full AI networking portfolio at Booth D19, Expo Hall 3, during the OCP Global Summit, October 12–15, 2026, in San Jose, California.
For more information about UfiSpace AI networking solutions or to schedule an on-site meeting, please visit www.ufispace.com.