The network is what lets all those processors talk without choking.
As European manufacturing grows ever more dependent on simulation and artificial intelligence to design the products of tomorrow, two infrastructure companies have formalized a partnership aimed at removing the hidden friction that slows such ambitions down. Cornelis and xFusion, meeting this week at the International Supercomputing Conference in Hamburg, have joined their networking and server technologies into a single platform built for the scale and speed that automotive, aerospace, and advanced engineering workflows now demand. Their collaboration reflects a quiet but consequential truth: raw computing power alone is not enough when the connections between machines cannot keep pace with the data flowing through them.
- European manufacturers face a compounding crisis of scale — crash simulations, fluid dynamics models, and AI-assisted design are outgrowing the infrastructure meant to support them.
- The bottleneck is not processing power but the network itself, which can strangle an entire server cluster when data cannot move fast enough between machines.
- Cornelis and xFusion are pairing the CN5000's low-latency, high-bandwidth networking with purpose-built FusionServer hardware to eliminate that chokepoint across large HPC deployments.
- Liquid-cooling technology in both the servers and networking hardware addresses the heat and energy costs that come with running thousands of processors simultaneously at industrial scale.
- The combined platform is being demonstrated in Hamburg this week, targeting automotive and aerospace sectors where simulation complexity and AI-driven design cycles are accelerating fastest.
A partnership announced this week in Hamburg is targeting one of European manufacturing's quieter but growing crises: the infrastructure bottleneck that emerges when engineering simulations and AI workloads scale faster than the systems meant to run them.
Corneis, a maker of high-performance networking technology, and xFusion, a server infrastructure provider, have combined their products into a unified platform. The pairing links xFusion's FusionServer hardware with Cornelis' CN5000 networking system — designed together for the kind of computational intensity that automotive engineers, aerospace designers, and advanced manufacturers now depend on daily. Crash simulations, fluid dynamics modeling, digital twin creation, and AI-assisted design all place extraordinary demands not just on processors, but on the connections between them.
That distinction matters. A cluster of powerful servers can still grind to a halt if its network cannot move data quickly or smoothly enough. Cornelis CEO Lisa Spelman framed the collaboration as a question of utilization: combining intelligent networking with high-performance compute so that large-scale infrastructure environments can run at their actual potential, not the reduced pace imposed by internal friction.
Both companies have also invested in liquid-cooling technology — in servers and networking hardware alike — recognizing that heat management is as much a competitive factor as processing speed. Denser, cooler infrastructure means lower operating costs and more computing power in the same physical footprint, a meaningful advantage for manufacturers optimizing large deployments.
The announcement arrives as electric vehicle development, autonomous systems, and AI-powered design tools are dramatically expanding the computational demands on European engineering teams. Companies that can simulate faster and more efficiently gain real ground on competitors. This partnership is, at its core, a bet that those companies will invest in infrastructure built to keep up.
Two infrastructure companies announced a partnership this week aimed at solving a problem that's becoming urgent across European manufacturing: how to run massive engineering simulations and AI workloads without the whole system grinding to a halt.
Cornelis, which makes networking technology for high-performance computing, and xFusion, a provider of server infrastructure, have combined their products into a single platform. The pairing connects xFusion's FusionServer hardware with Cornelis' CN5000 networking system. The target customer is anyone running the kind of computational work that automotive engineers, aerospace designers, and advanced manufacturers depend on—crash simulations, fluid dynamics modeling, structural analysis, digital twin creation, and increasingly, AI-assisted design processes.
The problem they're addressing is real. As companies push to process larger datasets faster and scale their computing environments, the infrastructure itself becomes a bottleneck. A server farm can have plenty of raw processing power, but if the network connecting those servers can't move data quickly enough or smoothly enough, the whole operation slows down. Lisa Spelman, CEO of Cornelis, framed it as a question of efficiency: "This collaboration combines high-performance compute and intelligent networking to help customers maximize utilization, accelerate engineering and AI-driven workflows, and improve efficiency across large-scale infrastructure environments."
The CN5000 networking platform is designed to deliver low-latency, high-bandwidth connections—the kind of throughput required when you're running fluid dynamics calculations or training machine learning models on massive datasets. xFusion's servers are built specifically for these kinds of demanding deployments, with architecture that can handle the density and complexity of modern HPC clusters. Frank Qin, CEO of xFusion Europe, emphasized that the FusionServer platforms are "purpose-built for high-density HPC deployments."
There's also an energy angle. Both companies have invested in liquid-cooling technology—xFusion in its servers, Cornelis in its networking hardware. When you're running thousands of processors simultaneously, heat becomes a serious problem. Liquid cooling is more efficient than traditional air cooling, which means lower operating costs and the ability to pack more computing power into the same physical space. For large manufacturing operations trying to optimize their infrastructure spending, that matters.
The partnership isn't entirely new ground. Cornelis and xFusion have been working together across multiple generations of hardware and software, so there's existing experience with how these systems perform under real-world conditions. The announcement comes as both companies are exhibiting at the International Supercomputing Conference in Hamburg, where they're showing the combined platform to potential customers.
The timing reflects a broader shift in European manufacturing. As automotive companies race to develop electric vehicles, autonomous systems, and AI-powered design tools, the computational demands on their engineering teams have exploded. A single crash simulation can require hours of compute time. Digital twin modeling—creating a virtual replica of a physical product to test designs—demands constant, massive data processing. Companies that can run these simulations faster and more efficiently gain a real competitive advantage. This partnership is essentially betting that European manufacturers will pay for infrastructure that helps them do exactly that.
Notable Quotes
This collaboration combines high-performance compute and intelligent networking to help customers maximize utilization, accelerate engineering and AI-driven workflows, and improve efficiency across large-scale infrastructure environments.— Lisa Spelman, CEO of Cornelis
Our FusionServer platforms are purpose-built for high-density HPC deployments. Together with Cornelis, we are helping customers build infrastructure environments optimized for large-scale modeling, simulation, and other data-intensive workloads.— Frank Qin, CEO of xFusion Europe
The Hearth Conversation Another angle on the story
Why does a partnership between a networking company and a server company matter? Aren't those just commodities?
They're not commodities when you're trying to run a crash simulation that takes 48 hours and you need it done in 12. The server has the power, but the network is what lets all those processors talk to each other without choking. Pair them wrong and you've wasted half your computing capacity just waiting for data to move.
So this is about removing a bottleneck that already exists?
Exactly. Companies are already buying powerful servers. What they're struggling with is making sure the infrastructure actually works at scale. When you go from 100 processors to 10,000, suddenly the network design matters as much as the hardware.
Why Europe specifically? Why automotive?
Europe has a massive automotive and manufacturing base that's under pressure to innovate faster. Electric vehicles, autonomous driving, advanced materials—all of that requires simulation and testing that's computationally intense. And European companies tend to be willing to invest in infrastructure if it gives them a real edge.
The liquid cooling angle—is that just marketing or does it actually change the economics?
It changes the economics significantly. If you can cool more densely, you can fit more computing power in the same building footprint. That means lower real estate costs, lower power distribution costs. For a large manufacturing operation, that adds up quickly.
What happens if this works? What's the next problem these companies will have to solve?
Probably software. You can have perfect infrastructure, but if the engineering software can't actually use it efficiently, you're still leaving performance on the table. The real win is when the hardware, the network, and the software all work together seamlessly.