Networks that stay online without interruption while pushing hardware to full capacity
In Munich, Huawei unveiled the second generation of its Xinghe AI Fabric platform, offering European enterprises a rethinking of how data center networks must evolve to sustain the relentless demands of artificial intelligence. The announcement reflects a deeper tension in our technological moment: as AI workloads grow more complex and unforgiving, the infrastructure beneath them must become not merely faster, but fundamentally more resilient and self-aware. That Huawei chose Europe as the stage for this commitment speaks to both the commercial stakes and the geopolitical currents shaping where and how global technology partnerships are built.
- AI workloads are pushing enterprise data center networks toward their limits, creating urgent demand for infrastructure that can sustain constant iteration without faltering.
- Huawei's Xinghe AI Fabric 2.0 introduces a tenfold reliability improvement and an engine that watches 200,000 service flows at once, detecting faults in seconds and recovering in minutes — compressing the window of disruption dramatically.
- A three-layer architecture — orchestration, connectivity, and hardware — attempts to remove human bottlenecks entirely, letting operators drag services into place while the system handles the complexity beneath.
- Real-world deployments at Turkey's Migros retail chain and Italy's Cineca research center signal that the platform is already embedded in European operations, not merely promised.
- Huawei is positioning this launch as a long-term regional commitment, even as geopolitical friction continues to complicate how European institutions weigh partnerships with Chinese technology firms.
At its annual Network Summit in Munich, Huawei introduced Xinghe AI Fabric 2.0, the successor to a platform first launched in 2018 and now substantially reimagined for an era when AI models demand relentless compute and cloud architectures shift without warning. The company's central promise was straightforward: networks that stay online without interruption while driving computing hardware to full utilization.
The system is built across three interconnected layers. The first handles orchestration and automation, allowing operators to configure services through simple drag-and-drop interactions while the platform manages cross-network operations and security zones on its own. The second layer introduces iReliable technology, which Huawei claims multiplies reliability tenfold, anchored by an Eagle-Eye Engine that monitors 200,000 service flows simultaneously and can detect faults in seconds, restoring service within minutes. The third layer sits at the hardware level, using specialized switches and optical modules to give operators precise visibility into where traffic is being lost or delayed, while embedding security controls that keep user groups isolated from one another.
Among the hardware on display was the XH9230, a liquid-cooled switch offering 128 ports at 400 gigabits per second — a density designed for enterprises consolidating infrastructure at scale. Live demonstrations at the summit showed the three-layer system responding to simulated network stress in real time.
Representatives from Migros in Turkey and the Cineca research center in Italy shared their experiences with existing Huawei deployments, grounding the announcement in operational reality. An associate professor from Sapienza University of Rome offered broader perspective on where AI networking is heading. Throughout, Huawei framed the launch as part of a sustained commitment to European enterprise infrastructure — a signal of regional investment that carries weight precisely because geopolitical tensions continue to shadow technology partnerships across the continent.
In Munich this week, Huawei introduced the second generation of its Xinghe AI Fabric system, a network infrastructure platform built specifically for European enterprises managing the surge in artificial intelligence workloads across their data centers. The announcement came at the company's annual Network Summit, where the technology was framed around a central promise: networks that stay online without interruption while pushing computing hardware to full capacity.
The original Xinghe AI Fabric arrived in 2018. This update represents a substantial rethinking of how data center networks should behave in an era when AI models demand constant iteration and cloud architectures shift rapidly. Bingyu Zhou, vice director of Huawei's European ICT marketing and solution sales division, noted that the company has already deployed professional data center infrastructure across multiple European industries, and this new version extends that foundation with what Huawei describes as high-performance, reliable, and secure networking that balances innovation with operational stability.
The system rests on three interconnected layers. The AI Brain layer handles orchestration and automation—users can drag and drop services into place, and the system manages application-to-network operations across different network types and security zones without manual intervention. The AI Connectivity layer introduces what Huawei calls iReliable technology, which the company claims increases reliability by a factor of ten. A component called the Eagle-Eye Engine watches 200,000 service flows simultaneously, spotting problems in seconds and restoring service within minutes. The third layer, AI Network Elements, uses specialized switches and optical modules to track traffic patterns with precision, showing operators exactly where packets are being lost or delayed, while also embedding security controls that isolate groups of users from one another.
The hardware side includes a liquid-cooled switch called the XH9230, which packs 128 ports running at 400 gigabits per second each—a density that matters when enterprises are consolidating their infrastructure. Paired with optical modules called StarryLink, the system provides what Huawei describes as full visibility into network behavior.
At the summit, representatives from two European organizations spoke about their work with Huawei. Suleyman Hilmi Tekinture of Turkey's Migros retail chain and Matteo Turisini of Italy's Cineca research center both shared details of their deployments. Daniele De Sensi, an associate professor at Sapienza University of Rome, contributed perspective on where AI and high-performance computing networks are heading and what practices work best.
Huawei's booth displayed working demonstrations of the three-layer architecture in action, with live simulations showing how the system responds to network stress. The company positioned the announcement as part of a broader commitment to working openly with European partners and customers, signaling that continued investment in the region's enterprise infrastructure remains a priority even as geopolitical tensions complicate technology partnerships globally.
Notable Quotes
Huawei has already built professional data center network infrastructures across industries in Europe, enabling both business innovation and operational efficiency through high-performance, reliable, and secure network solutions.— Bingyu Zhou, Vice Director of Huawei European ICT Marketing & Solution Sales
The Hearth Conversation Another angle on the story
Why does a data center network need three separate layers? Why not just one intelligent system?
Because the problems are different at each level. You need something that understands what services you're trying to run and how to move them around—that's the brain. You need something that keeps the pipes themselves working reliably and fast—that's connectivity. And you need the actual hardware to be aware of what's flowing through it and to protect it. Trying to solve all three at once creates complexity. Separating them lets each layer do its job well.
The Eagle-Eye Engine monitors 200,000 flows. That's a specific number. What does that actually mean for an enterprise?
It means the network isn't blind anymore. Without that kind of monitoring, you know something broke, but you don't know what or why. With it, you see the problem the moment it starts, which means you can fix it before users notice. That's the difference between a five-minute outage and a five-second hiccup.
Why is this announcement happening in Europe specifically? Why not globally?
Europe has strict requirements around reliability and data sovereignty. European enterprises need networks they can trust to stay up and to keep data contained. Huawei is saying: we've built something that meets those demands. It's also a signal that despite everything else happening geopolitically, Huawei is committed to the European market.
The iReliable technology claims a tenfold increase in reliability. How do you measure that?
You measure it in mean time between failures and mean time to recovery. If your network used to go down once a month for an hour, now it goes down once a year for six minutes. That's what tenfold means in practice—fewer incidents, faster fixes.
Who actually needs this? Is this for every company or just the big ones?
The big ones first—Migros, Cineca, companies running serious AI workloads or research. But the architecture is designed to scale down. A mid-sized enterprise with growing AI needs could use the same principles. It's not a luxury feature; it's becoming table stakes.