xFusion Advances AI-HPC Convergence with Full-Stack Innovation at ISC 2026

Hamburg, Germany, June 30, 2026 — xFusion returned to the International Supercomputing Conference (ISC 2026) in Hamburg, Germany, June 22–26, marking its fourth consecutive year at the event. From Booth D20, the company presented its latest liquid-cooled computing platforms, enterprise AI appliances, and full-stack HPC solutions, while delivering keynote presentations at both the Vendor Showdown and HPC Solutions Forum alongside global industry experts, technology pioneers, and researchers.

"Over four years of deep engagement at ISC, xFusion has evolved from a participant into an active contributor to the global HPC ecosystem," said Frank Qin, CEO, xFusion Europe. "Our FusionServer platforms are purpose-built for high-density HPC deployment — and Europe remains one of the world's most important markets for scientific and high-performance computing. We look forward to working with local research institutions, industrial customers, and ecosystem partners to build next-generation infrastructure for large-scale simulation, modeling, and data-intensive workloads."

Technical Thought Leadership on the Main Stage

At the Vendor Showdown, Arthur Wang, Global Director of Computing Infrastructure Product Management, presented xFusion's strategy around three pillars — technology leadership, ecosystem collaboration, and business growth — and outlined how xFusion helps customers turn infrastructure investment into measurable outcomes.

"The future isn't a choice between HPC or AI — it's about making them work together," said Wang. "The real question isn't how much compute you have, but how much of it is actually driving results."

At the HPC Solutions Forum, xFusion HPC Solution Architect Liu Jierui presented the company's end-to-end HPC architecture — a layered design spanning compute, networking, and storage at the infrastructure level, converged HPC-and-AI scheduling at the software level, and workload-specific tuning at the service level. He detailed how the SuperPOD rack-scale platform unifies CPU nodes, GPU nodes, and networking equipment through tri-bus blind-mating for power, network, and liquid cooling, while supporting all three major interconnects — InfiniBand, RoCE v2, and Omni-Path — to match each customer's cluster scale and workload profile.

"Most clusters are over-provisioned on hardware and under-invested in architecture," said Liu. "We designed ours from the bottom up — compute, network, storage, scheduling, tuning — so every layer talks to the next and nothing is wasted."

Deepening the European HPC Ecosystem with Cornelis

During the conference, xFusion and high-performance intelligent networking provider Cornelis announced a deepened collaboration targeting the European automotive and advanced manufacturing sectors.

The partnership brings together xFusion's FusionServer compute platforms and Cornelis' CN5000 intelligent networking technology to address demanding engineering workloads such as CFD simulation, crash analysis, structural modeling, and digital twin applications. Certification covers xFusion servers across multiple mainstream CPU architectures and supports a broad range of enterprise and HPC configurations. Paired with xFusion's liquid-cooled server designs and Cornelis' end-to-end liquid-cooled networking, the joint solution helps customers improve system-wide efficiency across both traditional HPC simulation and AI-assisted engineering workflows.

Innovation Showcase: Data Center to Edge

xFusion presented a converged AI-HPC cluster solution alongside a full product lineup spanning the complete compute continuum:

FusionOne HPC Solution — A full-stack offering covering compute, storage, networking, and application tuning. Compatible with heterogeneous architectures, the solution features the proprietary FusionOne DFS for high-performance storage and a unified converged platform for AI-HPC workloads. Professional planning services, a five-step tuning methodology, and dedicated optimization experts help customers accelerate cluster deployment and AI-for-Science innovation.

FusionPoD Rack-Scale Liquid-Cooled Server — Supporting up to 240 kW power density per cabinet with native cold-plate liquid cooling and a pPUE as low as 1.06. The tri-bus blind-mating design eliminates internal cabling and enables robotic, unmanned operations. To date, xFusion has deployed more than 100,000 liquid-cooled nodes in production environments worldwide.

TokenBox™ — Token-Native Enterprise AI Production Platform — Making its ISC debut, TokenBox is purpose-built from hardware to software for enterprise-scale token production. A single appliance supports local deployment of full-parameter flagship models up to 1.6T — including DeepSeek V4 — giving organizations the ability to host, serve, and scale frontier-class models entirely on premises for the first time. Liquid cooling keeps noise below 35 dB under typical workloads. TokenFabric™ provides seamless interconnection from a single accelerator pack to a full system to multi-node clusters, delivering true node-level scalability.

FusionXpark Portable AI Agent Development Platform — Extending AI capability from the data center to the edge, FusionXpark gives developers a portable environment for building and deploying AI agents locally.

From Watts to FLOPS: A Maturing Global HPC Strategy

Since its ISC debut in 2023, xFusion has steadily expanded its global HPC footprint — progressing from product-level capability to deeper ecosystem collaboration. As AI continues to reshape the computing infrastructure landscape, xFusion will keep working with partners worldwide to build more efficient, open, and sustainable infrastructure — ensuring every watt of computing power delivers greater business value.