UfiSpace and GIGABYTE to Showcase Broadcom's Next-Generation 800G AI Networking Solutions at AMD Advancing AI Day 2026
SAN FRANCISCO, CA — July 23rd, 2026 — UfiSpace, a leading provider of open networking solutions, and GIGABYTE, a global leader in high-performance server technology, today announced an expanded strategic collaboration to deliver next-generation, high-bandwidth networking infrastructure engineered for large-scale AI and agentic workloads. The joint solutions will be featured at AMD Advancing AI Day 2026 in San Francisco, held from July 22–23, 2026.
As enterprise AI transitions from standard inference toward highly parallelized, complex agentic AI workloads, data centers require a tighter integration between next-generation compute and high-density networking fabric. To maximize cluster efficiency and eliminate interconnect bottlenecks, the joint architecture pairs GIGABYTE’s high-density AI servers—powered by AMD's newly introduced 6th Gen AMD EPYC™ Servers CPUs built on TSMC's advanced 2nm process technology—with UfiSpace's flagship open networking platforms of 800G architectures.
At the core of the standard 800G deployment is the UfiSpace S9321-64E data center switch. Built on the 5nm Broadcom® Tomahawk 5 chipset, the S9321-64E delivers a massive 51.2 Tbps of switching capacity across 64 high-speed 800G ports. The platform features advanced, hardware-level congestion control capabilities that are fully optimized to slash Job Completion Times (JCT) for the most demanding enterprise AI, machine learning, and multi-node orchestration tasks running across GIGABYTE's EPYC-based server ecosystem.
Further pushing the boundaries of AI cluster density, UfiSpace and GIGABYTE have also introduced an ultra-high-density 64-port 1.6T OSFP direct liquid-cooled switch powered by the state-of-the-art Broadcom® Tomahawk 6 silicon at Computex 2026. Delivering a staggering 102.4 Tbps of switching capacity in a single platform, this innovative 64-port 1.6T OSFP switch tackles the extreme thermal and bandwidth demands of next-generation data center scaling. By integrating advanced liquid cooling directly into the fabric tier, the joint platform dramatically improves Power Usage Effectiveness (PUE) and maintains operational stability for kilowatt-range accelerated nodes.
"Deploying agentic AI at scale requires absolute synchronization between elite compute capabilities and massive networking throughput," said Daniel Hou, General Manager of Giga Computing. "By showcasing our latest rack-server infrastructure with UfiSpace's 64-port 800G OSFP switch at AMD Advancing AI Day, we will demonstrate a validated, future-proof blueprint. Combining the core density of the new 6th Gen AMD EPYC Server processors with an optimized 102.4 Tbps fabric eliminates cluster data starvation while addressing critical green data center power targets."
"We are proud to expand our collaboration with GIGABYTE to optimize fabric efficiency for the modern AI data center," said Vincent Ho, Chairman and CEO of UfiSpace. "Whether deploying our S9321-64E open networking switch for high-performance 800G fabrics or utilizing our Broadcom® Tomahawk 6-powered 1.6T OSFP liquid-cooled switch for next-generation clusters, this unified ecosystem provides the ultra-low latency, cognitive routing, and thermal management required to ensure 2nm CPU clusters operate at peak capacity."
"As AI workloads evolve from traditional inference to complex, multi-modal agentic AI, the network becomes a key factor in overall cluster performance," said Hasan Siraj, VP of Product Management, Core Switching Group, Broadcom. "Our Tomahawk 5 and Tomahawk 6 silicon are built specifically to provide the massive bandwidth, scale, and advanced congestion management required to lower Job Completion Times. We welcome UfiSpace and GIGABYTE’s new platform based on our merchant silicon to deliver open, highly efficient, and scalable 800G and 1.6T networking architectures that unleash the full potential of next-generation AI data centers."
Technical Highlights of the Joint Portfolio:
● High-Density 800G Performance: The UfiSpace S9321-64E data center switch provides 64 ports of 800G connectivity in a compact form factor, delivering a total fabric capacity of 51.2 Tbps to handle intensive inter-node communication.
● Next-Generation 1.6T Capabilities & Liquid Cooling: Powered by Broadcom® Tomahawk 6, the 64-port 1.6T OFSP direct liquid-cooled data center switch doubles total fabric capacity to 102.4 Tbps. Advanced direct liquid cooling mitigates extreme thermal thresholds, lowering PUE while enabling dense scale-up topologies.
● Minimized Job Completion Time (JCT): Advanced congestion management and adaptive routing across both platforms mitigate network tail-latency, directly accelerating iterative AI training, checkpointing, and dense agentic AI data flows.
● AMD Venice Ecosystem Optimization: Engineered specifically to match the ultra-high throughput and expanded PCIe/CXL lane counts introduced by the AMD EPYC™ Servers CPUs, ensuring efficient data orchestration across the rack.
● Open & Disaggregated Architecture: Offers cloud service providers and enterprise data centers a flexible, future-proof alternative to proprietary fabrics, preventing vendor lock-in while maintaining maximum scalability.
Showcase Information
The GIGABYTE rack-scale solution with integrated UfiSpace 64-port 800G OSFP switch was on display during AMD Advancing AI Day from July 22–23, 2026, in San Francisco, CA. Technical representatives were available to discuss deployment architectures and performance benchmarks.
For more information regarding UfiSpace open networking switches and GIGABYTE server compatibility, please visit www.ufispace.com or www.gigacomputing.com
About UfiSpace
UfiSpace is a leading provider of open networking and modular switch platforms, enabling the transformation of telecom, cloud, and data center infrastructure. Backed by deep technical expertise and a global partner ecosystem, UfiSpace delivers scalable, energy-efficient solutions designed to meet the performance and sustainability demands of next-generation networks. As a certified contributor to the Open Compute Project with telecom grade capabilities, UfiSpace plays a key role in accelerating the adoption of open, disaggregated architectures worldwide.
Learn more at www.ufispace.com
Follow us on our social media channels: LinkedIn, Facebook, and YouTube
MEDIA CONTACT
Name: Betty Wu
E-mail: betty.wu@ufispace.com